2025 the year of the rise of open source / open weights LLMs and other models

Open source models became a major buzzword in the AI world with the official launch of Deepseek-R1 in January 2025. The model matched the performace of major proprietary or closed models of that time like OpenAI-o1, but at a fraction of the cost. Many good open source Large Language Models(LLMs) and Small Language Models (SLMs) by major tech companies existed before that , especially since 2023, like LLaMA and SAM series by Meta AI, Mixtral & Codestral by Mistral, Phi series by Microsoft, Dolly by Databricks, Falcon by TII UAE etc. But the launch of Deepseek-R1 was the moment that really put the spotlight on open source models as a viable option for AI implemenation for major use cases. In 2025, other major AI companies also joined the party by launching powerful open source/ open weights models like GPT-OSS by OpenAI, Gemma 3 by Google, Nemotron by NVIDIA, Qwen & Qwen-Image models by Qwen AI of Alibaba, Seed by ByteDance, Kimi series by Moonshot AI, GLM Series by z.ai, MiniMax series, etc . As 2025 comes to an end, many new AI labs with much smaller teams and budget have launched promising open source models like Rnj-1 coding model by Essential AI of Ashish Vaswani (lead author of “Attention is All You Need”), Chatterbox Turbo by Resemble AI, MiMo series by Xiaomi, etc.

What is an open source or open weights model?

Most of us got introduced to the world of LLMs and their amazing capabilities by interacting with models like GPT-3 in mid 2020 or later with GPT-3.5 via ChatGPT towards end of 2022. These are closed source or proprietary models that are owned and operated by the creator OpenAI. A user can interact with the model via OpenAI website or via API calls and each interaction is chargeable. Bascially the user can use the product as per pricing or limits, but will have no info about the weights or inner details of the model architecture. Users have ownership only to their chats / outputs.

On the other hand, an Open Source model is one for which the creator has publicly released the model weights along with details of the underlying architecture and training process. Anyone can freely download the model weights, customise it via fine-tuning if needed, run it on any infra of their choosing and can use it freely without paying any royalty or usage fee to the original creator. A user can even commercialise an open source model and charge others money for using it. Open Source models are typically released under Apache 2.0 or MIT licenses. Example : Deepseek-V3

Theoretically an open weights model means the creator has published the model weights in public and anyone can download the weights and run the model in their own infrastructure or via certain hosting services. But the creator has not published any further details of the model like the training process or achitecture. Open weights model can fall into three types:

1, With fully permissive licensing like Apache 2.0 or MIT that allows for broad commercial and research uses without any limits. Example : GPT-OSS 120B and GPT-OSS-20B by OpenAI.

2, With partially permissive licensing where users have full commerical access upto a very high limit and for most common use cases. Examples : 1, Llama models by Meta allow full commerical usage till a very high threshold of 700 Million MAUs (Monthly Active Users). 2, Stable Diffusion , the text-to-image model by Stability AI uses a CreativeML Open RAIL-M License till version 2.1 that allows free usage except for harmul purposes.

3, With non-commerical use license that allows user to download and try the open weights model only for non commerical purposes that are not revenue generating yet. Example : FLUX.2 [dev] by Black Forest Labs.

In this blog whenever we talk about open source or open weight models, we will be referring to all models with permissive licensing like Apache 2.0 or MIT . This will also include models like Llama that have a very high limit for any restrictions to apply for most us and also earlier versions of Stable Diffusion. Also for ease of understanding, the terms open source and open weights maybe used interchangeably.

Why to use Open Source or Open Weights models

Following are the major advantages of using open source models:

1, Full privacy for interactions : When you are using an open source or open weights model by downloading the model weights and running on a suitable hardware of your choosing, you have full ownership of your data. The original model creator or anyone else cannot see your data, thus ensuring full privacy. Whereas in proprietary model you are sending all data to the server of the model creator, who may have access to the data for monitoring purposes or even have rights to use it for training purposes.

2, Safest way to use proprieatary data : For many business use cases, the data that you want to process using an LLM could be your proprietary or domain-specific data that exists in your private databases and you cannot risk sending such data to a model owned and controlled by a thrid party. For such purposes, the best option is to operate an open source model by yourselves and then use options like Fine Tuning or Retreival Augmented Generation (RAG) to introduce proprietary data into the model.

3, Predicatable costs and control of usage : Once you download a model and host it, you are in full control of planning the cost of running the model. You don’t have to bother about usage restrictions or price changes that you may face in the case of proprietary models controlled by others.

Which are the current major Open Source Models

This is a short list of some of the major models that you can try for the various purposes like text , image , audio, video, etc. This is not any ranking based on performace benchmarks, but instead just an indicative list for you to get familiar with some of the popular models and learn about them.

Text Generation (LLM)

1, Deepseek V3, R1, V3.1 and V3.2

2, GPT-OSS 20B, 120B by OpenAI

3, Gemma initial release, 1.1, 2, 3, 3n by Google

4, Qwen initial relase, 2, 2.5, 3 by Qwen AI of Alibaba

5, Mistral Small, Large 3 by Mistral

7, Llama 1, 2, 3, 3.1, 3.2, 4 by Meta AI

8, Kimi K1.5, VL, Dev, K2, Linear by Moonshot AI

9, GLM 4.1, 4.5, 4.6 , 4.7 by Z AI

10, Nemotron 2, 3 by NVIDIA

11, Mimo 7B, V2-Flash by Xiaomi Mimo

Image generation

1, Qwen-Image by Qwen AI(Alibaba) : Launched in August 2025 by Qwen under Apache 2.0 license. This is a 20B paramters model that is good at image generation, editing and text rendering.

2, Stable Diffusion by Stability AI : Versions 1 and 1.5 are free for commerical use. Later versions are free to download and use only for non commerical purposes. Latest version is 3.5

3, FLUX by Black Forest Labs : Model weights can be downloaded freely, but only for non commerical usage. Versions 1 and 2 are available.

Video Generation

1, Wan by Wan AI of Alibaba : The models offer text or image to video and are free for commerical use under Apache 2.0 license.

2, FastVideo by Hao AI Lab of UCSD : These are open source models built upon Wan by using the lab’s FastVideo framework, which is an inference and post-training framework for diffusion models.

3, HunyuanVideo by Tencent : It is free to use commercially, but has few restrictions like maximum limit of 100 Million MAUs and excluded territories like European Union, UK and South Korea.

Automatic Speech Recognition (Audio to Text)

1, Whisper by OpenAI : Large (currently at Version 3) is a 2B parameters model, while Small is 200M parameters model.

2, Wave2Vec by Meta AI

3, Parakeet by NVIDIA : Latest version V3 supports English and 25 European language. This 600M parameters model comes under Creative Commons license meaning need to credit the creator for usage.

4, Speaker diarization by Pyannote : Distributed under MIT license. Allows to even seperate different speakers in an audio file.

Large Audio Language Models (LALM)

These models can process both audio & text inputs to produce text output. Qwen2-Audio by Qwen, Ultravox by Fixie AI, Audio Flamingo by NVIDIA.

How to quickly try out an open source model

Having read about the various open source models, it is time to learn how to quickly try out these models by getting their weights and running them on a suitable device. Following are some of the quick methods:

1, On the cloud

Step 1 : Get the model weights from HuggingFace

HuggingFace is the leading open source website and community dedicated to AI and is the best starting point for getting familiarised with open source models. Major AI companies and creators publish models (weights and other details), datasets, etc via this platform. HuggingFace also offers Spaces where you can host and share your ML models and works.

The best part about HuggingFace’s interface is that you can browse through a vast catalog of more than 2 Million models in a very easy to use manner by filtering based on tasks like text generation, automatic speech recognition, etc and further sorting by popularity, size, etc as seen below.

Once you do the tasks based selection or search for a model you are looking for, you can find all the details in the dedicated page for the model . You can also get further options to run the model from here on notebooks or inference providers. See example below of Deepseek V3.2 model page.

Now if you explore the Files and versions tab you can see the model weights store in .safetensors format. In this example Deepseek V3.2 is a 671 Billion parameters model. So the weights are split into 163 parts of less than 5GB and stored in .safetensors format.

Github is another platform where creators of open source or open weights models publish details about the model including code and documentation. Model weights being large files are not published directly on Github, but you can find deteailed guides to use a model like link to Colab notebook.

Step 2 : Run the model on notebooks like Google Colab or Kaggle

Google Colab is a product by Google that allows you to run python code blocks and html from your web browser. The website even provides you with free CPU and basic GPU to try out ML models. You can run different code blocks and see the output in the same page. Consider this as a resource for testing various models using sample codes available for generating output, finetuning, etc. The free tier allows you to even use NVIIDA’s Tesla T4 GPU for a limited amount of time. For extended compute time you can upgrade to a paid version.

You can get a Google Colab notebook for running any of the open source models . The below example shows the Colab notebook for GPT-OSS 20B

Kaggle , the online machine learning and data science community, is another online resource that allows you to try out ML models and even provides you with free GPU hours. You can get free GPU time of upto 30 hours of two T4s or one P100.

2, On your local machine

Another easy option to quickly try out open source models is to download the models and run them locally in your own laptops or desktops. There are several ready to use applications that make it easy with a GUI.

Ollama is one of the most popular options. As shown below it gives you option to download different open source or open weights models and then can interact with them. For running locally, the hardware requirements to keep in mind are : 1, Enough hard disk space in GBs to download the modes and 2, Basic CPU and RAM to run the models. Having GPU and more RAM increases the speed. As current model sizes keep increasing, downloadin them locally maybe a challenge. So Ollama now provides an option called cloud models where they have made 4 popular LLMs of large size available on the cloud, but you can interact with them as if running locally. Big models from Deepseek, Qwen, OpenAI, are available on cloud.

Few other options for downloading model weights and running them locally on your system are LMStudio, GPT4all, llama.cpp, vLLM, etc .

Customising an open source model with your own proprietary data

We have so far seen the advantages of using open source models, especially in ensuring the privacy of your proprietary or confidential data. But to get the best implementation of LLMs, you should be able to safely process your custom data to build meaningful applications. So the next step is about how to use your own proprietary data in the open source models that you have chosen to host yourselves . There are two main approaches to do it.

1, Fine tuning an open source or open weights model

Fine tuning is the process of modifying some of a model’s weights so that it can learn some new information from the data that is introduced during the finetuning process. Imagine a business with some very domain specific requirements and know-how. For example, a financial services company with some custom financial calculations or similarly a biotech company with deep knowledge bio research or a firm in supply chain domain. In such scenarios, a great option to utilise the power of LLMs, could be to take a well proven open source model and adapt it further to your domain requirement. This can be done by finetuning with domain specific data to make the model gain domain expertise on top of the general intelligence already present in the foundation model.

Unsloth is one of the best tools to get hands-on with fine tuning , even using a small dataset. Unsloth provides an open source framework for LLM finetuning and reinforcement learning. It provides an extensive set of free Colab and Kaggle notebooks for major open source or open weights model.

Inside a notebook, you get the option to test the functionality of the model and also to finetune it. Below example is the Colab notebook for Gemma 3n by Google.

Tinker by Thinking Machine Labs is a new API based approach that provides easy finetuning where you can run the python code in your basic CPU based enviornment, while their service takes care of the heavy lift of finetuning across distributed GPUs.

2, RAG (Retrieval Augmented Generation)

Imagine you have domain specific data that keeps getting updated frequently or you have some query or user specific data that is relevant only for a specific user oruser queryand not for all users. In all such scenarios, finetuning is not the suitable approach . Instead what you need is an option that allows you for each query to include the custom data relevant for a particular query or a user in the context along with the query text. This method is called Retreival Augemented Generation or RAG. Suitable documents are chunked into parts and stored in vector databases. At the time of user query, stored data with similartiy to user query is retreived and added to the query in the context and then send to the model for getting a much more relevant answer.

Implementing open source or open weights models via major cloud service providers

Anyone who has worked in web based applications, websites or backends for mobile apps will have some exposure to major cloud service providers like Amazon Web Services(AWS), Google Cloud Platform (GCP) and Micorosft Azure. It is possible to get started with the world of open source models via such cloud based options :

1, Huggingface’s Inference Providers

This enables you to choose from a set of cloud based inference providers that run various popular open source models on GPU hosted by them at a predefined usage rate. You can see the different models and running options by visiting Inference Providers section in Huggingface. Here you can see the inference providers for various major open source models and the input and output rates they charge.

You can also visit the individual service providers and choose from the model options. Seen below is the console of Groq.

2, AWS Bedrock and AWS SageMaker

Amazon Bedrock is a servie that provides users an easy option to run various foundation models(FMs) on Amazon’s cloud infrastructure and then access them via APIs. The catalogue of available models include both closed models and major open source models

Pricing is based on usage. Here is an example of the procing for using Deepseek models by AWS Bedrock.

AWS Sagemaker provides a more granular control to using machine learning models on the cloud and needs more infrastructure management and expertise than AWS Bedrock.

3, Google Vertex AI Model Garden

Google Cloud Platform (GCP) provides Vertex AI for choosing and running various machine learning models. The Model Garden is a section that allows a user to choose from various open source or proprietary models for running.

4, Microsoft Azure AI foundry

This interface allows you to choose various models across open source and proprietary and run on Microsoft’s infrastructure. One major advantage of Microsoft Azure AI foundry is that provides leaderboards for comparing models for different tasks and thus aids users in choosing models.

This concept suggests a way to use smart contracts and stablecoins to improve the payments between different parties in the FMCG supply chain

Current status : There are credit periods and payments between any two directly connected parties in the FMCG distribution network. FMCG companies provide goods to large distributors mostly on full payment or sometimes on short credits. Further flow of the goods from these distributors till they finally arrive at the retail stores are mostly based on credit periods and future payments. Most small retailers and small distributors struggle to make payments upstream before actually selling the goods to end customers. Because of this the process is haphazard and involves lots of manual negotiations and financial constraints.

Suggested methodology : When end customers buy products from a retailer and pay the total bill digitally , the payments to all the upstream partners for the various products will be initiated automatically. This is achieved by creating a smart contract involving the different parties and then settling the payments using stable crypto coins like Digital Rupee (e₹) , Tether (USDT), etc.

FMCG supply chain

First let us understand the typical FMCG supply chain in little more detail before going into more details of the suggested smart contract based payment network.

The typical parties and transactions in the FMCG supply chain are as follows :

0. Raw material suppliers : They are the providers of ingredients that are procured by the FMCG companies to make the final finished goods at the FMCG factories. This leg of the supply chain will not be covered in the scope of this smart contract payment mentioned in this concept note.

1. FMCG organizations typically operate to have a negative cash cycle. This means that they collect money from the market before paying their suppliers. They provide their goods to the distributors/wholesalers after taking upfront payments.
2. Distributors / Wholesalers are the ones who take the goods in bulk from the FMCGs and are in charge of distributing it further to a large area that each one covers. Most of the time, they have to pay upfront to the FMCGs for getting the goods and in some cases, they get a minimal credit time to pay after getting the goods. However, this is not true in the next leg of the product’s journey. They have to provide credit periods to the nodes below them to move the products through the supply chain. The distributor, in order to get the highest share of the shopkeeper’s wallet (and in turn the consumers) has to play out on multiple fronts : relationship and engagement , offers, returns and credit terms to name a few.

This node in the FMCG supply chain may have a further hierarchical structure where a distributor provides goods to smaller parties like sub distributors or stockists / agencies who finally supply the goods to the retail stores. For the ease of explanation of the smart contract concept, we will be considering all these as one node.

Cash and Carry is a form of wholesaler where the retailers can procure items in bulk from a typical warehouse kind of store at wholesale prices by settling the payment / cash on the spot and carrying the items by themselves. Since this does not involve any credit periods and retailers directly take the goods, for the sake of this concept note, we will be excluding this model from the distribution network being explained.

3. Retail stores / Supermarkets are the shops from which end consumers buy the products from. They can be classified into two categories.

A, General Trade : Mom and Pop stores, corner shops, kirana stores

B,Modern Trade : Supermarkets, Hypermarkets, brand outlets, etc.

These are the shops that procure a vast variety of goods of different FMCGs via many distributors and put them on shelves for the end users to buy. They typically receive credit periods from the distributors , but these credit periods vary based on the products and the distributors from 10 days to 20 days. In Indian context, most of the small retail shops give post-dated cheques for the owed amount to the distributors based on the credit periods.

National key accounts like D’Mart, Dunzo, Blinkit etc are a hybrid between the above 2 nodes (Distributors and Retail Stores) and are not considered as their operating structures are very dependent on the negotiations/contracts that may exist.

4. End consumers are the shoppers at the retail stores who buy the goods and make payments at the store counters. This is the last node in the FMCG supply chain. The consumers buy different products from the choices provided by a particular store and make payments. The retailers then use this money to pay the distributors / wholesalers at the end of the credit period.

Implementation of smart contracts and stablecoins for payments in FMCG supply chain

Example scenario 1 : Assume a product A of price X has a retailer margin of 10% and a distributor margin of 10%. In this example , let us assume that the FMCG company also gives a credit period to the distributor. When a user pays digitally on purchasing product A, the billing system at the retailer will ping the smart contract which will then pay the retailer 0.1X , the distributor 0.1X and the FMCG company 0.8X instantaneously using a stablecoin linked to the smartcontract

Solidity smart contract for this scenario will be as follows:

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.0;

contract PaymentSplitter {
    address payable public retailer;
    address payable public distributor;
    address payable public fmcgCompany;

    // Constructor to initialize the addresses
    constructor(
        address payable _retailer,
        address payable _distributor,
        address payable _fmcgCompany
    ) {
        retailer = _retailer;
        distributor = _distributor;
        fmcgCompany = _fmcgCompany;
    }

    // Function to split and distribute payment
    function pay() external payable {
        require(msg.value > 0, "Payment amount must be greater than zero");

        uint256 totalAmount = msg.value;

        // Calculate the split amounts
        uint256 retailerShare = (totalAmount * 10) / 100; // 0.1X
        uint256 distributorShare = (totalAmount * 10) / 100; // 0.1X
        uint256 fmcgShare = (totalAmount * 80) / 100; // 0.8X

        // Send the respective shares
        retailer.transfer(retailerShare);
        distributor.transfer(distributorShare);
        fmcgCompany.transfer(fmcgShare);
    }
}

Key Points of the Contract:

1. Constructor: Initializes the addresses for the retailer, distributor, and FMCG company.
2. Pay Function: Splits the payment into 10% for the retailer, 10% for the distributor, and 80% for the FMCG company.

Example scenario 2 : Assume a product B of price X has a retailer margin of 10% and a distributor margin of 15%. In this example scenario , let us assume that the FMCG company gives no credit to the distributor. When a user pays digitally on purchasing product B, the billing system at the retailer will ping the smart contract which will then use the linked stablecoin instantaneously to pay the retailer 0.1X , the distributor 0.90X ( 15% margin + 75% already paid to the FMCG company).

Solidity smart contract for this scenario will be as follows:

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.0;

contract PaymentSplitter {
    address payable public retailer;
    address payable public distributor;

    // Constructor to initialize the addresses
    constructor(
        address payable _retailer,
        address payable _distributor
    ) {
        retailer = _retailer;
        distributor = _distributor;
    }

    // Function to split and distribute payment
    function pay() external payable {
        require(msg.value > 0, "Payment amount must be greater than zero");

        uint256 totalAmount = msg.value;

        // Calculate the split amounts
        uint256 retailerShare = (totalAmount * 10) / 100; // 0.1X
        uint256 distributorShare = (totalAmount * 10) / 100; // 0.9X

        // Send the respective shares
        retailer.transfer(retailerShare);
        distributor.transfer(distributorShare);
    }
}

Advantages of smart contracts and stablecoins based payments in FMCG supply chain

1, Today, especially in the emerging markets, the point of sale is still un-organized and is serviced by multiple channels : Wholesalers/stockists, distributors and modern trade (DMarts/Metro’s of the world). The biggest advantage of such a technology innovation would be with the wholesalers and distributors as this is an unorganized space with a scale that is not easy to manage. Also, the challenges do not end at this node with collection of cash/money remaining a critical and most time consuming/loss generating activity for a distributor or a wholesaler. Another point to keep in mind is the fact of credit cycles that this sector operates at the grass root levels. This is an area where smart contracts can really be game changers by releasing cash tied up in inventory which can be put to better use.

2, Distributors are basically extensions of your supply chain who have sales targets to hit : your last mile to bring value back to the organization. Most distributors rely on their teams to collect money. A smart contract can enable flow of information and money seamlessly. They can then focus on building better retailer relations. For example : A typical salesman can cover 30 shops each day taking orders if they are not tasked with additional activities like continuous haggling for collection of payments and their efficiency drops to 23–25 shops if they have to do additional activities . That is close to 17% loss of efficiency and to some extent sales. Such a rule based payment with no need for manual back and forth will enhance the distributor and retailer relation.

3, Also, this automation will support improving the cost for the subsequent legs of the value chain and improve the ROI for all those invested in the journey of the product.

4, Another benefit that is hidden is that of range selling. Typically, distributors are too focussed on their cash cow SKUs for obvious reasons. With smart contracts the risk of new products can be fairly shared enabling the market to decide the destiny of the innovation vs. the distributor who wants his/her money to keep rotating the easiest way possible. Imagine the next innovation council meeting where you have very clear market feedback on why a product was/was not a success instead of rehashing what the distributor says and may/ may not be true. Remove the ambiguity. An example of this scenario is the sudden drop in the price of basic soap brands in rural / semi Urban pockets in India before the budget announcements. This is done by small stockists and distributors to get additional cash in hand to purchase cigarettes before the anticipated increased tax on cigarettes (a common trend) hits them : This helps them make some additional money but the disruption of this price drop takes time to get stabilised, having an impact across the supply chain.

5, End to end visibility of the products journey enabling unprecedented opportunity to optimize inventory and drive innovation and efficiencies without 3rd part engagements increasing the return on marketing/selling spends

New financial opportunities on implementing a smart contract based payment system

1, FMCGs can pump some money to the nodes in its distribution network for supplying their products and get paid back when the product finally sells and the end user pays. This will help FMCG companies that are looking for more distributor and retailer interests for their products

2, FMCG supply chain payments, though can be haphazard, are a kind of risk free cash flow since either when the product sells, there will be payment or when the product is returned from the shelf, there will be refund. By switching these payments to smart contracts and stablecoins, there is an opportunity to create a new kind of financial instrument. Investors can be allowed to buy off the pending payments from retailers and distributors at a discount and get paid when the end user buys, thus earning a profit.

Using smart contracts and stable coins for FMCG supply chain payments. was originally published in Talk about Tech on Medium, where people are continuing the conversation by highlighting and responding to this story.

In most markets world over users are bombarded with mutliple choices of apps for doing their online grocery shopping. In India, for grocery items that can be delivered quickly via bikes, there are 4 different players fighting to grab users in major cities — Dunzo, Swiggy’s Instamart , Zomato’s Blinkit and Zepto. This is further complemented by players like BigBasket (part of Tata group), Reliance’s JioMart and Amazon Fresh that offer deliveries of even larger items via mini trucks or three wheelers.

Most of these apps do not have major UI/UX differentiation as their designs mostly contain the same ingredients : a list of FMCG goods that are divided into relatable categories , options for users to build a shopping cart , option to pay digitally at checkout screen and option to pick a delivery time. Only lockin users have to any apps are the purchase of any subscription program for 1 month or upto 1 year to get free deliveries or item discounts.

This blogpost aims to reimagine the grocery shopping experience for the user by suggesting the design of a single super app for the user that is independant of all delivery apps. Delivery partners will become pure plugins into such an app. The same app can also have extended utitlity to aid users when they are shopping physically at a store with features like digital checkouts, AR based shopping assistant , etc.

1, Online shopping experience

Such a consumer super app can bring following new features to online grocery shopping :

i, List of items & SKUs published and maintained by FMCG brands directly

Each FMCG company has the most accurate details of their products including pictures, product details, etc. Currently most shopping apps take the pictures inhouse and manually enter the details. In most cases it end up being not fully legible and not in a digitally reusable format. This consumer app will instead allow FMCG companies to publish and update details of the products in the catalogue directly.

This will ensure following advantages :

• Most accurate and up-to-date product info directly from brand owners
• Digitally reusable product info which can further inputted to other apps used by the consumers like a calorie counter app or a recipe app
• Quickest update to a customer on a new or updated product thus enabling them to start shopping for latest products

ii, Delivery service automatically chosen for user’s online shopping basket

The consumer app is purely focussed on digital shopping part and does not maintain stock or provide any delivery service. It will instead allow all grocery delivery companies to provide their details to the consumer app When a customer completes a shopping cart, it will fetch the latest details from the different grocery delivery companies and allow customer to give a consent on which service provider to use.

Non-branded likes vegetables and fruits can be included without any brand mentioned and the list of available options will be an aggregrated list of the varieties stocked by nearby grocery delivery companies. In-house brands of different delivery comapnies can be included as an FMCG brand and included in the menu list.

The suggested ordering of delivery providers shown to customer on final checkout screen will be based on the following factors :

• Stock availability at the delivery company’s end for the items in the user’s basket.
• Delivery time guaranteed by a grocery delivery company for the items in the users basket
• Discount rates on the products and offers on various payment options
• User can be suggested deliveries from multiple service providers to cover all the items included in their shopping cart or to ensure maximum savings for a give set of items.

iii, User specific loyalty discounts by brands based on past digital bills

One major advantage of such a consumer shopping app is that it can finally provide the missing link between FMCG companies and their major customers. For shopping at physical retail outlets, the bills are not digitised or even if a digital copy is available , it does not link the FMCG companies to their consumers and are aimed for loyalty programs of the retail stores. For current online grocery shopping apps, though payments to receipts are all digital, their is still no link between FMCG companies and customers.

Imagine a user who has been buying a large SKU of a particular brand of detergent for many months in a row. The FMCG company would love to retain such a user by giving them some loyalty discounts irrespective of which platform the user shops from next.

The grocery shopping super app will provide following advantages :

• Enable FMCG companies to give loyalty discounts directly to the heavy shoppers of their brands instead of non-targeted general discounts for everyone. The app will user an intermediation logic that will collect the discount options from the FMCG companies for different usage patterns and publish the relevant ones to the user app based on past purchases.
• An instant digital user survery or feedback channel betweent FMCG companies and customers. Currently FMCG brands use the traditional user surverys, focus groups, market tests, etc to decide on new products or get feedback on existing products. This consumer super app could replace all that with a digital channel between brands and end users.

2, Physical store shopping experience

Online grocery shopping is still nascent in many markets and in India it is still less than 1 percent (Source : https://www.statista.com/topics/9572/online-grocery-shopping-in-india)

The above consumer app can also offer customers new experiences when they visit a physical store for doing their shopping.

i, Augmented Reality product info to know product details better

The product details printed on product packages are very tough to read as brands have to print lots of info on limited spaces. Plus to compare items users have to manually check each item. In the digital era, the grocery shopping app can provide an innovative way for shoppers to digitally view the details of items in a physical store . It can provide an Augmented Reality interface to display major product details as a info panel. The AR interface can also provide a quick product comparison option to view relative details of competing products.

ii, Digital carts by scanning QR codes instead of physical shopping carts

Physical retail stores could do a major change in their layout. Currently most outlets with limited physical area have an overcrowding of items which makes the store very cramped. This is because most of the outlets store their majority stocks in the user shopping area as seen below

Retail outlets can be possibly reinvented with the aid of a digital shopping app in the following manner :

• Seperation of shopping area and storage area
• For pre-packaged item , the shopping area contains samples of each items for each SKU with adjacent QR codes that a customer can scan in her/his grocery shopping super app to add it to the digital cart.
• While adding an item to cart , a shopper can also click whether that item is self pickup now or to be home delivered later.
• At the checkout counter, the store staff will pick the items from adjacent storage area , pack and give to the customer on payment. This will also eliminate customer theft issues in reatail stores.
• Items like vegetables, fruits, etc may still be kept in the existing way of having majortiy stock in the shopping area itself based on whether pre-packed or not. Customers can carry these in physical shopping carts.

iii, Option to choose between self pickup or a delivery service provider for each item at payment counter

Some of the items purchased at a physical store are heavier ones that are difficult for customers to carry inside a store to the checkout counter and also to transport to their homes. Best examples are 5kg+ packs of cereals or flours as seen in the below pic. Physical stores can offer a hybrid model where customers can decide some of the items or even the entire items are delivered later by the store’s own delivery crew or by a third party servcie provider for a fee. Customers may prefer to take the lighter and urgent need items with themselves instead of getting delivered later.

In most of these instances, it is about brands or celebrities creating NFTs and fans buying them and then reselling them. This blog explores another possible approach to NFTs which will allow users or fans to generate their own NFTs in different contexts (location + event + time) and then keep those NFTs in their collections or offer them for sale to other users on a blockchain platform. Below are some of the the new use cases that can emerge from this approach:

1, Sports collectibles created by fans from match pics captured by them or by collecting geoAR tokens

Before going into this use case in detail, here is a sneak peak on the status of sports collectibles from the pre-digital era to the current NFTs era

Sports collectibles in the pre-digital era : physical sports cards

In the pre-mobile era, the official collectibles for fans were mostly in the form of physically collectible cards that are were launched by the sports federations or some authorised brands who had an official tieup and were running a promotion . Below are few such collectibles from the 90s. Some of the major challenges were : i, preserving the cards for a long time ii, finding the ideal market value for a card based on rarity, demand, etc iii, interacting with other users who may want to buy a card or exchange.

Currently available sports collectibles in the digital NFT era

In the current Web3 era, when it comes to official sports collectibles, the most popular example is that of NBA Topshots http://nbatopshot.com / launched by NBA, NBA Players Association and Dapper Labs . It is an NFTs marketplace which allows users to collect and trade popular video clips from various NBA and WNBA matches. As shown in the screen below they offer different match moments as collectibles at diffrent price ranges. Fans can not only buy new mints , but they can also resell later to other fans.

Recently OneFootball launched Aera https://aera.onefootball.com/ a similar sports collectibles , starting with Italy’s Serie A . It provides NFTs of top moments in a match and is hosted on the Flow Blockchain

New use case : Fan created digital sports collectibles

A possible next evolution for sports collectibles is to make sports fans the creators by allowing them to create NFTs from the match pics or videos they have captured or from the geo AR fan tokens they have collected.

A, Pics & Videos created by fans : As official match video clips and pictures become widely popular on social media, some of the pics and videos created by fans during a sports match also gain popularity and provide a different view from the official ones. A below example is of a fan pic captured during a qualifier match for 2018 FIFA Men’s Football worldcup between India and Guam . It was clicked by the famous West Block Blues fans from the West block A stand at Kanteerava stadium in Bengaluru.

If such historic pics captured by fans that become popular on social media are allowed to be uploaded as an NFT alongside the official match NFTs, then it will have two major benefits : i, Increase the variety of contents available for purchase as NFTs thus making them more interesting to fans. ii, Increase fan engagement by allowing them to convert their creative pics and videos to NFTs and even earn money from them.

B, AR tokens collected by fans : Another approach to sports collectibles is for federations and teams to launch AR tokens. These can be in the form of GeoSpatial AR tokens that are tied to different locations . Then fans can discover them by visiting certain places and scanning for AR tokens and then upload such collected tokens as NFTs. The below video shows an example of such a token that can be created using location based AR like the GeoSpatial API in Google ARCore. This demonstrative example is of a Virat Kohli fan token that can be placed at different locations by his IPL team and can then be collected by fans of Bengaluru team and Kohli from various popular locations across Bengaluru.

As a next step such a collected token can be uploaded as an NFT by the fan who collected it and can continue owning it or even offer it for sale. The fans can be given an option to upload the pic to a distributed storage like IPFS and converted to an NFT or even directly published to an NFTs marketplace. Below example show how the AR Geo token collected in above video example has been uploaded to Opensea to generate an NFT :

https://opensea.io/assets/matic/0x2953399124f0cbb46d2cbacd8a89cf0599974963/78264859506846578750605159776006226996410453808808605726948358104430474690561/

Major benefits of such an approach to sports NFTs are :

i, This provides a new variety of sports NFTs that are not just match pics but also other creative graphics of players and teams.

ii , This creates an option to engage fans on non match days and off season with quests to discover AR tokens that can then converted to NFTs . Fans with such AR token NFTs can be provided free tickets or merchandise.

2, Location based tokens that are unlocked by people who explore a city or a tourist spot

The second use case of user generated NFTs are the tokens they can collect for visiting various tourist places or cities. Before touching up on this use case, here is a brief history of location related collectibles so far.

Location based collectibles in the pre-digital era : postage stamps, coins and currencies

In the pre-internet era, stamp collection or Philately was a major hobby worldwide that allowed people to keep tokens of different countries. The value of the stamps used to depend on their age and rarity. Coins and currencies collections also had such a location or country element.

Currently available collectibles in the Augmented Reality and Non Fungible Tokens era

The most famous example of location based collectibles is Pokemon Go which is an Augmented Reality based game that was launched in 2016 and became a worldwide craze. It uses mobile devices with GPS to locate, capture, train, and battle virtual creatures, called Pokémon, which appear as if they are in the player’s real-world location.

A great example of Non Fungible token based game is Axie Infinity , which allow players to collect and mint NFTs which represent digital pets Axies.

New use case : User genereated NFTs from geolocation based AR tokens collected by visiting tourist spots or cities

In the current era of Augmented Reality and NFTs, there is possibility of having a globally standardised location based AR tokens concept that allows uers to discover AR tokens on visiting a location and then convert those tokens to NFTs and trade them with other users all around the globe.

This concept revolves around having a common global standard for embedding of AR tokens at various locations , a standardised user interface for apps or games that want to allow users to discover such AR tokens and an option for users to easily mint NFTs from such tokens. For example, below example shows such an AR token based at Indiranagar, which is a major tech startup and shopping hub in Bengaluru, India.

Benefits of having such location AR tokens that can be converted to Non Fungible Tokens are :

i, A possibility of reinventing stamp collections in the current AR and NFTs era by allowing users to trade these location based NFTs on a global blockchain plaform. A further design will allow the starting value of an NFT to be decided based on the rarity of the token. For example an NFT genereated by a user by collecting an AR token from a less travelled tourist spot like Iceland can have more starting value than a token based on a highly visited place like Paris as the latter will be available in more numbers. Other factor that can be keyed in to determine the price is how far the minting user has travelled to reach the particular location , the date of token collection like whether it is a major festival or event day, etc.

ii, Governments and tourism departments can do city wide tourism quests to encourage tourism by incentivising users to collect varied tokens by visiting different tourist spots. Users can be incentivised with discounted entry fees and other benefits based on NFTs from such AR tokens

iii, Great opportunity for local merchants to run highly targeted promotional campaigns for users who visit their locations for collecting the AR tokens embedded by the brands.

Addendum : Few code snippets relevant to building above sample

1, Google ARCore Geospatial anchor allows to place AR content to a geographical location using latituge, longitude and altitude

// Enable the Geospatial API.
session.configure(session.config.apply { geospatialMode = Config.GeospatialMode.ENABLED })

// Obtain a VpsAvailabilityFuture and store it somewhere.
val future = session.checkVpsAvailabilityAsync(latitude, longitude, null)

// Poll VpsAvailabilityFuture later, for example, in a render loop.
if (future.state == FutureState.DONE) {
  when (future.result) {
    VpsAvailability.AVAILABLE -> {
      // VPS is available at this location.
    }
}

if (earth.trackingState == TrackingState.TRACKING) {
  val cameraGeospatialPose = earth.cameraGeospatialPose
  // cameraGeospatialPose contains geodetic location, rotation, and confidences values.
}

if (earth.trackingState == TrackingState.TRACKING) {
  val anchor =
    earth.createAnchor(
      /* Locational values */
      latitude,
      longitude,
      altitude,
      /* Rotational pose values */
      qx,
      qy,
      qz,
      qw
    )
  // TODO: Attach content to the anchor specified by geodetic location and pose.
}

2, React fronted app for uploading an image to IPFS to use it for generating an NFT

instantiateContract() 
{    
const contract = require('truffle-contract')    
const ipfsuploader = contract(IPFSUploader)    ipfsuploader.setProvider(this.state.web3.currentProvider)     

// Get accounts   
this.state.web3.eth.getAccounts((error, accounts) => {        
  ipfsuploader.deployed().then((instance) => {        
    this.ipfsUploaderInstance = instance        
    this.setState({ account: accounts[0] })        
    // Get the value from the contract to prove it worked.        
    return this.ipfsUploaderInstance.get.call(accounts[0])      
 }).then((ipfsHash) => {        
    // Update state with the result.        
    return this.setState({ ipfsHash })      
  })    
 })  
}

captureFile(event) {    
event.preventDefault()    
const file = event.target.files[0]    
const reader = new window.FileReader()    reader.readAsArrayBuffer(file)    
reader.onloadend = () => {      
   this.setState({ buffer: Buffer(reader.result) })      
  } 
}

onSubmit(event) {    
event.preventDefault()    
ipfs.files.add(this.state.buffer, (error, result) => {          
   if(error) {        
         console.error(error)        
         return      }      this.simpleStorageInstance.set(result[0].hash, { from: this.state.account }).then((r) => {        
   return this.setState({ ipfsHash: result[0].hash })                
   console.log('ifpsHash', this.state.ipfsHash)      
   })   
 })  
}

render() {    
return (      
<div className="App">        
  <nav className="navbar pure-menu pure-menu-horizontal">           
       <a href="#" className="pure-menu-heading pure-menu-link">IPFS File Upload DApp</a>        
  </nav>         
 
  <main className="container">          
     <div className="pure-g">            
       <div className="pure-u-1-1">              
          <h1>Your Image</h1>              
          <p>This image is stored on IPFS & The Ethereum Blockchain!</p>              
          <img src={`https://ipfs.io/ipfs/${this.state.ipfsHash}`} alt=""/>              
          <h2>Upload Image</h2>              
          <form onSubmit={this.onSubmit} >               
              <input type='file' onChange={this.captureFile} /> 
              <input type='submit' />              
           </form>   
         </div>          
      </div>       
   </main>      
 </div>    
 );  
}

Buying crypto currencies in a centralised exchange

A centralised exchange like Binance or Coinbase is similar to traditional stock exchanges in the sense that they allow to have an account with a simple login instead of dealing with private key or seed phrase. They abstract and handle these details for you. Also they allow you to buy or sell crypto coins using fiat money via different options including credit and debit cards. For this tutorial I am using Binance exchange and INR as currency.

To get started in an exchange or even a wallet, users are typically required to do an online KYC to verify their identity to ensure safety in such platforms. This is the screen that asks for such details. In Binance & earlier in Metamask also the committed time to finish KYC was 2 days, but it got completed in less than 2 hours.

On choosing Indian Rupees (INR) as currency, Binance gives me an option to buy crypto coins via P2P transactions , ie I am directly paying to a seller who wants to sell the particular coin I am looking to buy . Limitation with P2P is that it is a manual process. The exchange will hold the coins in custody from the seller, buyer transfers money to seller directly, then seller has to give confirmation to exchange on recieving funds and then exchange will transfer the coins to the buyer.

In this example I bought around 19 MATIC tokens on Polygon chain using FIAT money which is INR. Now the coins will reflect in trading wallet section of my Binance account. As a next step, I want to move it out of the centralised exchange to my fully controlled ethereum address so that I can manage the coins using my own non custodial wallet in a purely decentralised manner without any centralied exhanged acting as an intermediary. Remember once you have a private key and account(s) for Ethereum , the same is valid for other EVM compatible blockchains like Polygon and Solana and even testnets like Rinkeby and Kovan. So here I am using the address of my account 1 in my metamask wallet for ethereum chain to send the MATIC tokens on Polygon chain.

Since the transaction is happening on Polygon, you can see how low the gas fee is . If I had ETH and was trying to send to the same address on ethereum chain, then the gas fee would have been much higher. It may take upto 10 minutes as shown below for the transacion to be complete and the coins to be visible against your address.

Now in my metamask wallet, on choosing Polygon mainnet as the network, I am able to see the MATIC tokens against my address

Experimenting with DeFi and DApps : Aave on Polygon Chain

Once you have coins on a blockchian, you can also experiment with DApps (Decentralised Applications) especially DeFi (Decentralised Finance) apps . I have used some of the MATIC coins on Polygon that were purchased earlier to get hands on with Aave. As per the polygon blogpost https://blog.polygon.technology/polygons-defi-series-aave/

The Aave protocol is an open-source non-custodial decentralised liquidity protocol built on the Ethereum blockchain. It is the largest DeFi protocol in the world (by total value locked) and has a distributed team of over 70 people all over the world. The Aave protocol allows users to supply and borrow cryptoassets and earn yield on cryptoassets supplied to the protocol.

Go to aave.com and connect Launch App and then connect your wallet

When Ethereum mainnet is the chosen network on my metamask wallet, Aaave will show me the coins and the APY I can earn for them and also gives me option to borrow

On clicking supply you can see the the next screen that will show the gas fee. Since I am on ethereum network, the fee will be higher.

Now I switch to polygon network on my wallet and in Aave DApp to see my coins listed with option to Supply

On clicking Supply for MATIC it will show me the gas fee, which is very low on Polygon network as against the Ethereum chain

I went through with the transaction and then after it is complete , I can see that I have supplied 1 MATIC or approx 0.66$ to Aave so that I can start earning interest. I have the option to withdraw the amount or swap for another coin which will provide a different APY

Swapping coins on a Decentralised Exchange (DEX)

DEX or a Decentralised Exchange allows a crypto user to trade on crypto coins by swapping one coin for another. The process is achieved using a concept called Liquidity Pool which exists different pairs of coins. You can easily find detailed writings on the topic in the internet. Below you can see how this can be done on both Polygon and Ethereum chains.

1, QuickSwap on Polygon chain : QuickSwap is a DEX on Polygon chain. In an earlier part I had shown how I bough MATIC on Binance exchange and then sent them to my address. Now I will be swapping some of those MATIC tokens for WETH (Wrapped Ether) and DAI tokens. For example in the screen, I choose to swap some MATIC for DAI and then I need to click confirm

The my metamask wallet will ask me for confirmation and show the gas fee , which is very low and has to be paid in MATIC token , since this entire transaction is on Polygon Chain.

Similarly swapped some MATIC for WETH on Polygon.

The updated balance of all 3 tokens will show up in the coins dropdown option in QuickSwap and also in metamask web extension

Now, if the tokens are not appearing in metamask wallet, one needs to import the tokens. Click on Import Tokens screen after choosing Polygon Mainnet as the network. For WETH following are the details that can be obtained by seeing the WETH contract page on Poylgon scan. On import, the WETH balance is updated.

Below you can seen how to check in Polygonscan to find the contract address for DAI and then use that to import the tokens in metamask mobile wallet

2, UniSwap on Ethereum Chain : UniSwap is a Decentralised Exchange (DEX) on Ethereum blockchain. In the below I am trying to swap some of the ETH in Ethereum chain to get USDC on ethereum chain.

On clicking swap and then on the Confirm Swap on the popup, metamask wallet will popup to show the expected gas fee

In this case I cancelled the transaction after seeing the details as I did not want to execute a transaction with such a high gas fee on ethereum network.

3, Swap within Metamask wallet : Metamask wallet provides a swap option. It is not a standalone DEX like Uniswap or Quickswap, but instead it instead it fetches quotes from multiple DEX aggregrators and suggests user to choose the one with lowest charges.

Here you can see the price quote and gas price estimate for swapping ETH to USDT and WBTC. I did not execute the transactions and cancelled after seeing the price details.

Since gas fee lower in Polygon network , I switched to Polygon mainnet and did a swap of MATIC to WETH at a very low gas fee as shown below

Buying an NFT on OpenSea marketplace

OpenSea is the largest marketplace to buy and sell NFTs. Non-Fungible Tokens can be based on any of the two different standards ERC721 or EC1155. The latter standard actually enables the creations of tokens that are a mix of both fungible and non-fungible. Further, OpenSea allow transactions of NFTs that are based on Ethereum, Polygon or Solana blockchains. In this example below, I am buying a part ownership in an ERC1155 NFT on Polygon. I am doing this so that I can experiment at a very low cost.

1, Add Wrapped Ether for transacting on OpenSea : This is needed because OpenSea allows transactions using WETH which is an ERC20 token. For buying Ethereum NFTs, it must exist on Ethereum chain as WETH on Ethereum and for buying Polygon NFTs, it must exist on Polygon chain as ETH on Polygon .

If there is not enough WETH, more can be added by wrapping ETH balance or MATIC balance based on the chain in which balance is needed . See below example of wrapping ETH to WETH on ethereum

2, Then I searched for Polygon sports NFTs using the search filters and then sorted them by Price: Low to High

3, Checked the details section to confirm it is on Polygon and can also see type as ERC1155

4, To buy the chosen NFT click checkout, complete the Unlock and Sign parts of the transaction by clicking on the wallet , thus completing the purchase.

Importing NFTs to non custodial wallet like Metamask

To view any NFT bought on OpenSea in your metamask wallet , you need to import the NFT with the required details like contract ID and token ID . Go to your wallet and choose the network based on where the NFTs exist . In this example it is Polygon. On the NFTs tab, click import NFT and then enter the contract address and token ID that can be obtained form OpenSea. On completing the import, the NFT will be displayed in your metamask wallet.

Multiple non custodial or self custodial wallets using the same seed phrase or recovery phrase

For a non-custodial wallet , the 12-words seed phrase or recovery phrase is everything. Hence the major warning signs that you keep seeing on “never to lose this phrase” and to “create safe backups” . Ethereum seed phrases are based on BIP39 standard. It is a code generated when you create a new wallet for ethereum blockhain for the first time using a non-custodial wallet. You can play around on this site https://iancoleman.io/bip39 / to get an idea on the link between the 12-words seed phrase or recovery phrase, private key and public key.

This means that a user can reuse the same phrase in a new non custodial wallet to access the exisinting wallet without having to get a new 12-words mnemonic. I have used the seed phrase that I got while setting up my metamask wallet to import the wallet account to coinbase wallet app as shown below. This way I am able to see all my crypto coins that I hold on Ethereuem and Polygon chains in my coinbase wallet app . This is just by using the same recovery phrase or seed phrase that was provided to me when I had set up my metamask wallet.

Viewing transaction and coins holding details of an address using Etherscan and Polygonscan

Etherscan is the website where you can view details of a transactions and addresses in a more easy to read UI/UX.

You can view the details of the transactions against an address and also the tokens owned by that address by visiting Etherscan. Example : https://etherscan.io/address/0xad085736a7a27009dd346ecd1ce51cfab80c5e85 . If a particular address is also linked to an ENS (Ethereum Name Service) .eth domain, then that can also be used instead of the hexadecimal address. Example : https://etherscan.io/address/tonymathew.eth

As mentioned before, a user will have the same private key, public key and addresses in any other EVM compatible chain. So transactions and coin holdings of a given address on thePolygon network can be viewed by visiting PolygonScan and going to the page for that address . ENS name will work only for Ethereum chain and not for Polygon. Example : https://polygonscan.com/address/0xAd085736A7a27009Dd346ECd1cE51CfAb80c5E85

Transferring between Ethereum and Polygon blockchains

To view any NFT bought on OpenSea in your metamask wallet , you need to import the NFT with the required details like contract ID and token ID . Go to your wallet

1, Visit the website https://wallet.polygon.technology/bridge to do such inter chain transfers via the Polygon Bridge and connect your wallet

2, Ethereum to Polygon transfer needs just one transaction and cost gas fee on the Ethereum network

3, Polygon to Ethereum transfer has two stages with one on Polygon chain and other on Ethereum chain and will take upto 3 hours to be completed

Experimenting with wallets for other blockchains : Phantom wallet for Solana blockchain

Phanton Wallet is a famous wallet for Solana blockchain. If you want you can experiment with it. It provides an option to buy SOL using cards. Since minimum transaction is USD 300, i did not try any transactions. Here are the screens.

Push Cart vendors form a significant part of the people who are employed in the unorganised sector in India. They may vary from street vendors who are stationed at permanent locations on a street to those who move around in a neighbourhood, like vegetable vendors.

Two signficant factors about pushcart vendors are :

i, Big trust among customers , especially for the long serving vendors

ii, Great knowledge of all details of the areas they cover in their routes

Currently most of the push cart vendors in major Indian cities use phones with mobile interent connections and accept UPI payments from their customers . This blog suggests how to add a further digitisation layer to them to help them gain better customer reach and thus more income and also how they can be utilised for last mile deliveries.

1, Publishing of photos of items available at the start of a day

For many vendors, their regular customers would love to see the items available on a day, once the vendors have stocked their carts at the start of a day. There should be a common portal where vendors can click pictures of their carts and upload them at the start of the day and maybe even few times during the day to show current stock status. This can also attract new customers who looking to purcahse a particular item and want to see the nearby vendors who sell them.

2, Video call option to do shopping from stationary vendors

Regular customers may want to shop for perishable items like vegetables , fruits, etc during the start of the day to get the best available stock. In such cases , if the customer is not able to visit a vendor in the morning, a video call feature should allow them to shop and pay for the required items. The customers can then pickup the packed bags later as per convenience or get the bags picked up by a delivery service provider like Dunzo or Swiggy Genie.

A video call option maybe a good feature to enable vendors selling a large variety of items like books, utensils so that the customers who do not have a particular item in mind, can see all the available options and choose the ones that interest them.

3, Option for customers to indicate purchase interest to moving vendors

Push cart vendors who keep moving throughout the day have certain fixed routes which they cover everyday. By enabling a digital list of the items sold by such moving vendors and their approximate route information, customers can know beforehand the vendors that cover their neighourhoods. There should be an option for customers to notify the vendors about their interest in buying certain items on a given day . This will ensure that a vendor who sells a particular item will cover the interested customer’s route.

Such an option is very useful for items for which a customer’s demand may not be daily and is more random, like few days in a week.

4, Map based tracking of moving push carts vendors

Currently almost all pushcart vendors who are moving use two options to inform their customers when they are passing through an area :

i, Conventional mode of shouting out the main items they are selling

ii, Tech mode of using battery powered loudspeakers with pre-recorded messages

Pushcart loudspeaker

Digitisation of the route plans of such vendors with approximate map based tracking of their current positions can be used to enhance this process. With this sort of digitisation, customers can know in advance which vendors selling what items are covering their area on a given day, can give purchase interest if needed and then can get alerts when the particular vendors are currently in their areas.

5, Use of stationary street vendors as pickup locations for ecom

Many stationary vendors have fixed locations on the streets for many years and the customers in a locality have a great amount of trust on them. They are an untapped option for serving as a pickup location for eccommerce deliveries. For situations where the ecom customers are not at home during delivery, customers can be given an option to choose from list of nearby known street vendors as pickup locations for the items.

6, Use of moving street vendors to complete last mile delivery for ecom

Moving pushcart vendors who have been covering the same areas for a few years or more typically have great knowledge of details of the localities . They can easily identify customer addresses and landmarks. They are an untapped option for completing last mile ecommerce deliveries of easy to handle and non urgent items like clothes, books, etc. They can help ecom companies with asynch delivery problem . The companies can deliver items for a customer to a nearby hub or pickup point. Later when the customer is at home, she/he can either pick it up or can use delivery by a pushcart vendor. Based on the customer’s indication of the time they are available at home, a particular vendor who is covering that route can be assigned to pickup the item and deliver to the customer. This create additional income source for vendors.

Please check the report here for more insights : https://image-src.bcg.com/Images/Demystifying-the-Online-Food-Consumer_tcm21-238295.pdf

Indian food tech space is characterised by a duopoly of Zomato and Swiggy with market shares of roughly 50% and 40% respectively. The former acquired Indian operations of Uber Eats in start of 2020 , thus gaining a lead over its close competitor.

Following are some suggestions for next possible basic innovations in the Indian online food delivery space that could enhance the user experience

1 , Ordering System innovations at restaurant

Currently restaurants log into multiple apps , one for each delivery partner and also have a seperate billing system.

The following process improvements can be achieved if restaurants have a single software to manage their billing (for both online and at-restaurant orders ) and for tracking online orders from various food ordering platforms :

1.a , Billing software similar to ERP system to accurately track the available quantity of each dish at a restaurant :

• Most restaurants plan the quantity of each dish to be made on a day while purchasing the ingredients at the start of the day or a few days before.
• Restaurants should have an option to input these plannned quantities of each dish at the start of the day into their billing sytem.
• Each dish should have an available quantities indicator that can be seen by ordering users. With each new order, this quantity is auto updated.
• Such a system will also enable auto confirmation of online orders based on the available quanity instead of the existing manual verification.

1.b , Option to pre-book food much ahead of the desired delivery time , with assured availability of confirmed orders

• Given an option, users would like to make their lunch or dinner plans a few hours earlier or even at start of a day instead of just before eating time.
• Current advance ordering systems merely save orders and then release them to the restaurants close to delivery time and so lack assurance.
• For restaurants with ERP like billing sytem, customers should be allowed to order food at any time of their convenience and choose when they want it delivered, thus giving customers an option to order food in advance
• The billing system should reduce the total available quantity of a dish based on such pre-bookings and allow sales of only the remaining quantity for any orders received after that. This gives assured availability of pre-ordered items at the time of desired delivery, even if ordered much ahead

1.c , Accurate estimation of preperation time based on standard dish preperation methods and other simultaneously active orders :

• Based on preperation style and time, dishes can be roughly categorised into following types :

i, Pre-cooked in bulk and packed on order . Eg: idli, poha, thali meals, etc

ii, Cooked on order, with short cooking time. Eg : Paratha, Dosa, Pizza at QSRs, Shakes . For such items, more than one order can be cooked parallelly

iii, Cooked on order, with long cooking time. Eg : Dishes in premium restaurants. Each item is individually made and may have around 30 mins cooking time.

• The billing and order management system should give an option for a restaurant to input what is the preparation style (bulk or individual) and preperation time for each dish and how many quantities of an individual dish can be parallelly prepared.
• When an order arrives, an accurate estimate of preperation time should be generated based on the dish category and other parallel orders, etc .

1.d , System generated auto orders to create express checkout option for most in demand items during the heaviest demand slots:

• Based on past ordered items and the time slots in which they were ordered, the billing sytems should be able to offer a rough prediction of potential demand for some of the main items during the busiest slots
• The system can create auto orders for a small percentage of the predicted orders so that the restaurant can keep them ready for express checkout.
• When a customer order arrives for such a high demand item, the system can assign a customer order to an already pre-packed unit , thus creating shorter delivery time and better customer experience
• Food delivery apps can optionally pre-place a pool of delivery staffs near a set of restaurants with express checkout items during the busiest slots

2 , Clubbing of multiple orders from the same user or from different users in nearby areas

Order clubbing can increase the profitability of the food ordering apps and the margin can also be passed on to the restaurant partners and users :

2.a , One user can order from multiple restaurants in a single order

• Restaurants should be divided into approximate clusters by considering restaurants in nearby locations as a single cluster.
• When a customer places an order from one restaurant, the user can also order from other restaurants in the same cluster via the same order
• The delivery person can be suggested a suitable order of picking from the multiple restaurants, based on the preperation time of the dishes
• User will get all orders from all restaurants delivered by one person
• Users need to pay only a single fee instead of multiple delivery fees

2.b , Clubbing of orders from different adjacent restaurants by multiple users in the same apartment complex or nearby areas

• Curently orders from different users are only clubbed together in a single delivery if they are from the same restaurant at around the same time.
• As a next step, the above restaurant cluster concept can be used for clubbing of orders from nearby restaurants from different users
• The food ordering apps backend should club together users who are in nearby buildings or streets. All orders from such users in a certain time window from restaurants in the same cluster should be clubbed together.

2.c , Custom lunch menu for office parks, where advance orders from different users are delivered together in the cafeteria

• Food ordering apps should have a different interface and approach for taking orders from users at office locations like large corporate parks
• Users in each location should see a curated menu from select restaurants that can be ordered upto 1 hour before the delivery time.
• Few restaurants can be permanently chosen and other restaurants can be onboarded to this curated list on a rolling basis.
• Orders from hundreds of users will be delivered together in mini vans
• When orders are at the office location, users will be alerted via the app and can pick up from the cafeteria. A staff can be assigned for confirmation.

2.d , Tea runs that can accept pre-orders and instant orders

• Another possible approach for delivering lower priced items like teas, coffees, snacks, etc in a profitable manner is to use a tea run approach
• Multiple flasks of hot beverages and some quantities of popular snacks are carried by a delivery person in a few designated routes that have many office buildings or residential complexes
• The tea runs will be during desginated 30 mins slots like 9:30am, 10am, 4pm, 5pm, etc during the typical tea break times
• Users can pre-order for an upcoming tea run or see if any unbooked flasks are available in any of the active tea runs in their areas and then book.
• Delivery persons will be placing the items at front desks or security gates for self pickup by users so that they can service multiple orders quickly

2.e , Area based groupon type large discounts on select dishes for group buying by users

• Food ordering apps can provide a single dish or few dishes in a groupon like discounted manner. Users who opt for any of these dishes via clubbed delivery should get major discounts.
• This can be typically run for dishes matching a day’s special event. For example, Bengali dishes during Durga Puja festival, etc
• Restaurants can bid for the opportunity to offer the featured dishes

2.f , Ideal format of clubbed order delivery for max profitability

• This is a suggestion of an ideal max target for clubbing of orders and the resultant increased delivery person payments, restaurant profits, etc

• Ideal max clubbing for pre-booked items : 50 packs of price INR 100 each .
• All these should be delivered by one delivery person in a 2 hours window
• Restaurants pay 10% (INR 10) to the food delivery platform & get INR 90
• Each user should pay INR 10 as delivery fee on top the price of INR 100. So total price a user for such an order box is 100(price) + 10(fee) = 110
• Margin per trip for ordering app = 50 boxes * 20 rupees per box = 1000
• A full time delivery person will work for 10 hours = 5 * 2 hours = 5 trips
• Margin per person per day = 5 trips* 50 boxes* 20 rupees per box = 5000
• Margin in 30 days per full time delivery person = 30 * 5000 = 150,000
• Salary to delivery person + bike costs + insurance = 50k per month
• Allocation for part-time crew fees + delivery hub rents, etc = 50k pm
• Revenue to food ordering app after all payments = 150k -50k -50k = 50k
• Approximate delivery staff per major food ordering app = 100k
• Monthly revenue potential = INR 50k * 100k = INR 5B ~ USD 60M
• Annual revenue potential = INR 60B ~ USD 720M

3 , Enhancements in user’s ordering screen

Food ordering apps can upgrade the user experience from the current format of showing a list of restaurants and dish suggestions as follows :

3.a , Comparison of mulitple food carts from different restaurants

• Currently users can build only 1 cart with items from 1 restaurant.
• Give a user an option to create multiple baskets with items from different restaurants so that she/he can compare based on price, expected delivery time, etc and take a final decision from the created options
• UI/UX wise the baskets should float around on the screen as small bubbles so that the user can click on any one to expand it and see full details
• If some of the baskets built by a user are from restaurants in the same cluster, then the user can also be given option to order from those multiple baskets from different restaurants.

3.b , Nutrition planner

• An overall list that helps users to decide based on variety and nutritional recommendations, festival types, etc
• The food ordering app can track the items a user has consumed in the past days and suggest variety based on recommended nutrtional inputs

3.c , Multiple modes for browsing

• Basic mode : Standard option currently provided by food delivery apps
• Repeat mode or Quick order mode : Suggestions based on past purchase history . The app can derive the user’s weekly or montly purchase pattern and suggest quick order options accordingly.
• Wow mode : Dish based browsing with photos and short 5 sec videos that are posted by the restaurants as well as by other users . This is aimed more at incentivising users to try dishes they have not tried before.

3.d , Sweet box ,Cooler Section, Hot kettle, etc on checkout screen

• Many restaurants that cater to a particular genre of food items do not have a wide vareity of sweets and beverages in their menu
• While having lunch or dinner, users may prefer to have sweets and beverages to accompany the food , but they are restricted to ordering only those that are made or stocked at the restaurant they are ordering from
• Instead users should be shown a wide selection of sweets and beverages at the checkout screen so that they can order what they want
• Food ordering apps can create tie-ups with bakeries, grocery stores, dessert and beverages outlets to provide the menu options
• The delivery person should pick up the ordered item from the partner store that is nearest to the user’s location

3.e , Video call based ordering or customisations

• Many QSRs offer customisable combos from already cooked item. Eg: Eat.Fit. Currently users can customise from the app screen
• User experience can be improved if they can in person instructions regarding quantity of each items they want and also option to see the dishes live and then decide what options to choose.
• Allow users who pay an extra fee to have video call with the restaurant , via the food ordering app, while there order is being packed

3.f , Wishlist of items that are currently not available

• Many speciality items in low demand may not be available frequently
• Give an option for user to add non available special items to a wishlist
• Restaurant can see such wishlists and then confirm the order if they can

3.g , Option to choose reusable containers like Milton, Cello

• Tie ups with container makers like Cello and Milton to introduce reusable containers for delivery to ensure more freshness of delivered food when it reaches the customers hands, be it hot or cold type food.
• Customers need to pay an additional container fee for such orders
• An area based reverse supply chain system to collect back the used containers at the end of a day from multiple users in area who have ordered food in such containers
• Containers to be cleaned and reused again and kept at busy restaurants

4, A single user ordering on behalf of multiple users

A better UI/UX experience is needed for frequently ordering for a group like the members of a household, office colleagues, room mates, etc

4.a , User profiles and user groups

• An ordering user should have an option to create a group of user profiles
• Each member of an ordering group should have a profile with food preferences , allergies, etc. Other users of the app should be able to share their profiles with another user who wants to order for them.
• A user can create multiple groups for different purposes like family members, roommates, office colleagues, etc

4.b , Chat and select option

• Make a shareable template for major chat apps like whatsapp, messenger
• The ordering user can then share a set of options with others via any of the chat app
• Other users should be able to +1, add customisation comment , etc for the options and the curator should be able to directly submit this to the food ordering app instead of having to manually enter the same in the food app

4.c , Curated orders to a venue for events, etc

• An in app option for organiser or admin to create a curated list of available options for people attending an event
• Ideally the options should be shareable in other apps through which the attendees RSVP for the event
• Option for the curator get choices of the users and the submit the order in bulk to the food ordering app
• For large number of attendees, the curator should be able to provide menu options from multiple restaurants and not just one restaurant.

4.d , Multi location ordering option (like people in a zoom party)

• Multiple people in different location can chat and see the same screen and order together
• One or few admins can plan the menu and choose the delivery locations
• Assured delivery of items in set time bucket. Use of buffers to ensure similar time slot delivery
• A polling option to see and choose menu. Single payment and confirmation

4.e , Custom offers for each user group instead of same general offers

• Dynamically generate combos and offers that will incentivise the user instead of one size fits all offers

5 , Delivery hub and spoke model innovation

Hub and Spoke model can be employed in Indian food delivery space :

5.a , Local hubs for storage of food for customer pick up

• Hub is a place for customers to pickup their orders instead of being delivered directly to their houses. This avoids the customer having to coordinate with multiple delivery persons for multiple orders.
• In large residential and commercial buildings, electronic lockers like Qikpod can be setup near the main entrance gate.
• Another option is to sign up local shops with hot and cold store options
• Ideally there should be a hub within 5 mins walking distance of a any user

5.b , Part-time delivery crew of local residents , college students

• Delivery spokes can be added to the hubs by using local residents and college students who are interested in part-time jobs for picking up orders from a hub and delivering to the houses of customers
• Such deliveries are of short distance and so done by cycle or by walk
• Ideally employed during peak lunch hours and peak dinner hours
• Roughly a part-time delivery person who works 2 hours per day for 22 days in a month should be able to earn around INR 5k per month

5.c , Express food packs at hubs for pick up without ordering

• Few quantities of the most popular dishes in an densely populated area can be kept ready at a hub , with appropriate hot or cold storage
• While opening the food ordering app, a user can see such pickup options available , if any , at the nearest hub and book
• A user can also walk in to a hub, see available options and then buy.

5.d , Assured basic delivery mode during bad weather by using hubs

• Currently during rainy weather, the delivery process breaks down with many restaurants shown as not serviceable. Also bike delivery will take more time than usual and so all customer orders are delayed.
• Clubbed order delivery using Auto Rickshaws : Users can be provided an option to order via a clubbed delivery and several user orders will be delivered together by an auto rickshaw. The food delivery company should have a set of auto rickshaw drivers who are part of delivery during rains.
• If the order is deliverable only to a hub, then the user has the option to pay a bad weather convenience fee to get it delivered to home by a full time delivery person or a part-time crew member

6, Introduction of home chefs, reviewers, curators , etc in the platform

6.a , Home chefs who can prepare user odered dishes

• Home chefs can sign up on the platform by indicating the kind of dishes they can make on order. For a user, this experience is like having home cooked meals , but without having to hire an onsite cook at her/his home.
• When a user wants a home cooked meal or a custom dish that is not in standard restaurant menus, she/ he can give the order in the home chefs section of the platform a few hours in advance and get it delivered.

6.b , Home chefs with own menu items

• Home Chefs can also sign up for selling own menu items, like a restaurant
• The maximum orders that they can take in a slot must be restricted to ensure adequate service delivery to the interested users

6.c , Food enthusiasts to post reviews for new restaurants

• Onboard the most active users of the platform who give both positive and negative reviews in a balanced manner, as amateur food critics
• Invite them to try out new outlets or less reviewed restaurants on the platform for free and to post pics and reviews.

6.d , Curators who could help users to try new premimum dishes

• Platform could offer the service of food enthusiasts who can help users to try out new experiences by suggesting restaurants and dishes.

6.e , Gourmet experience at home

• Option to bring a premium restaurant experience at home by offering the assistance of a restauarant crew for serving the ordered food.
• Unlike a catering service, this is meant for a small number of people like a small family or a mini gathering.

7, Kitchenless restaurants and eating spaces

7.a , Restaurants with no kitchen

• Eating areas can be set up in major commercial or residential spaces where people can get tables and chairs like in a restaurant and they can order food via a food ordering app from a set of nearby quick service restaurants
• This is to cater to the user need for having the experience of going out instead of eating at home or office, but without being restricted by the limited menu items of a standard restaurant or food court.
• For busy timings, users can pre book seats like a restaurant.

7.b , Kitchenless hotels and lodges like oyo rooms, etc

• Food ordering apps can tie-up with lodges and budget hotels to provide a a limited menu options that will be delivered in a quick manner
• This helps such hotels to eliminated the cost of maintaining a kitchen
• The customers of such places get the advantage of having a variety menu instead of the limited items available in budget hotels with own kitchens

8, Innovations to help restaurant partners

8.a , Temporary kitchen and restaurant staff on hire

• The food ordering plaforms can train and onobard a few kitchen staffs for standard restaurant activities. These could be part-timers or full-timers
• The restuarants can then hire such staff from the plaform for some peak days or weeks or even just for the busy hours during a particular day
• Restaurants contribute to their salaries based on the duration of hiring

8.b , Option to accept orders for non menu items from users

• Customers can create a wishlist of items that are not available for ordering from any restaurants, but they would like to have
• Restaurants or chefs can see such orders and place their interests
• Example : Traditional foods that are eaten during certain festivals

8.c , Advanced ordering from far away restaurants for weekends and special occasions

• Food ordering platforms typically restrict a user’s ordering option to restaurants within a certain distance limit.
• Some premimum restaurants or renowned traditional outlets in a city may have only one or very few physical locations. Customers in far off location should be able to order from such places a few hours or days in advance
• Deliveries in such long distance cases be done via three wheelers

8.d , User preferences map for new restaurant opportunities

• Food ordering outlets can publish a map of customer orders based on genre of foods and the major locations from which demand arises
• They can create an addtional layer to indicate potential new restaurant opportunities and the food type missing in any particular area.
• Restaurants can access these details for a price to know new opportunities
• As a next step, food ordering players can provide ready to use real estate and equipments on lease to speed up the creation of new restaurants that could serve some urgent opportunities. Can give preference to women Self Help Groups

9, Improvement of customer hand over time

9.a , Option to give easily understandable delivery instructions

• Food ordering platforms should identify a set of easy-to-locate universal landmarks in each pincode area and list them in the add address section. These are the places that even a new delivery person can easily locate.
• When a user adds a new address , in addition to creating a pin on the map, she/he should also add instructions to reach the location from the nearest universal landmark from the above list
• The instructions should be added by choosing from a set of standard micro instructions like turn left, take stairs, etc that can be translated accurately to any language for aiding the delivery person

9.b , Incentivise users for speeding up delivery handover time

• Users should be encouraged to speed up the handover time from when a delivery person reaches thier location to the customer taking the delivery.
• For users who enable very fast handovers by meeting the delivery person at the map point or by using delivery hubs, give them micro discounts for their next orders or free premium samples from FMCGs, etc
• For user’s who cause excessive handover delay for multiple times, introduce small penalties like disabling of discounts for next few orders

10, Enhancement of earning potential

10.a , Sharing of delivery staff with other food delivery players during peak hours to work in a hub and spoke model

• During rush hours and option to share delivery persons between different delivery players to improve user experience in different plaforms
• Ideally use it in a hub and spoke model , wherein the final delivery touchpoint to the customer will be a person from the relevant ordering app

10.b , Sharing of delivery staff for ecommerce delivery , grocery delivery during non rush hours

• Allow delivery person to earn additional income by taking workloads for ecommerce delivery, grocery delivery during non rush hours

10.c , Tie ups with social commerce platforms to replace 2 hours of meal preparation time of stay-at-home women with part-time jobs

• To increase employment rate and income among stay-at-home women, food delivery platforms can tie-up with social commerce players
• Women should be incentivised to replace the approximate 2 hours of meal preperation , that includes vegetable shopping, actual cooking and cleaning of utensils, with 2 hours of part-time phone based jobs.
• They should have economical options to outsource their cooking to nearby restaurants that make similar types of dishes with some monthly packages.

AppMap (https://appmap.xyz)

Ideal for viewing on laptops or tablets . This page offers a summary of the apps used by consumers in different countries around the world . It lists the main food ordering apps used by users in some of the largest economies. The list is not accurate and is meant to give a rough idea. Uber Eats , Just Eat Takeaway.com and Delivery Hero have presence in a large number of countries around the world either directly or by acquiring major players in different countries. Doordash, the top player in USA, also joined the global expansion spree by acquiring Finland based Wolt in Nov 2021.

In 2019, Google launched a Stay Safer feature on Maps with these features. Recently Truecaller launched Guardians app which helps a woman to choose a set of guardians from her contacts and share her location with them or quickly send a help message.

In certain situations like when a woman is travelling alone late night and the journey is long, to make her feel absolutely safe, monitoring by relatives or friends alone may not be enough. A next layer can be built on top of the amazing safety apps to give an option for the travellers to share their location with the police monitoring room and also send auto alerts to them in case of an emergency . Following are few suggestions on what could be a potential design of such a system.

1, Live monitoring by police or civic authorities

> Women voluntarily share live info to the police control room: Women travellers get an option on their app to voluntarily alert the police monitoring team about their journey with start and end points and the number of the vehicle. Give them option to share their live location details automatically with the official verified police control room.

> Maps based dashboard at each police control room: Each police control team will be responsible for certain routes and see all the live locations of all women who have shared their locations on a single map, with stats and alerts. The system will auto alert if the location of any traveller deviates from the ideal route from start to end points of their reported journey.

> Mandatory location tracking for the vehicles: For cars or autos that are having women passengers who have requested police monitoring, the drivers should mandatorily share their tracking signal to the police control room. This ensures dual tracking for each trip, the passenger’s signal and the driver’s signal.

> Alerts for any missing signals and deviations: In case of any suspected deviation, the police team should see immediate alerts. To make it foolproof, the team should also be alerted if there is a loss of tracking signal due to network congestion, signal jamming, loss of mobile battery power, etc so that they can use their patrol vehicles or traffic cams to track the vehicle and ensure safety. In case of a missing signal, the nearest patrol vehicle should be alerted so that they can go and track down the vehicle.

> Integrated inside existing apps : Incorporate this functionality in the user app for cab or auto booking . Similarly the design of driver app should be such that vehicle location , driver details, vehicle number , etc will be automatically shared from the driver app. The concept called also apply to dating apps.

2, Co-traveller pooling

> This is more relevant for long distance travel from places like airports and railway stations which will have a large number of travellers at any point of time and hence higher probability of finding others who are heading in the same direction. Best example is Bengaluru airport which is located almost 40kms away from the city. The basic assumption is that women will feel safer when travelling in the company of others(trustable) rather than alone.

>Ticketing booking apps can have a mechanism where women travellers can share their travel plans, as soon as they book tickets, to a trusted pool and then the system finds an ideal reliable match based on time of travel and other factors. This is similar to how pooling or sharing works for Uber, Ola, etc, but the difference is that the matching happens hours or days before the actual trip and not at the last minute. The system can even give option for the matched co-travellers to contact each other before hand to build a mutual trust. To build trust, there could a combination of a one time government id verification + linking of social media profiles .

Linkedin is the world’s largest professional network with more than 600million users worldwide. Almost every professional and college student in the world has a Linkedin profile and use it for various purposes like networking, online learning, job hunting, profile building, etc. The following analysis presents a set of possible new features in Linkedin that could enhance the user experience and make it an even more powerful platform.

Linkedin Cloud CVs

• A new section within Linkedin for creating and saving CVs, instead of using external word document tools.
• Import Linkedin profile as starter template and add further details
• Create multiple variants for different profile types
• Job Application methods:

i, Authorise the job-posting company and recruiter to view the online CV

ii, Auto-generate a PDF or word version and email or upload as attachment

• Auto Job search mode : When enabled by a candidate, Linkedin auto scans the hidden CVs for details and matches the candidate to suitable job openings and provides suggestions

Revenue potential : Users can be charged for professional CV templates, more number of CV variants, CV review services and enabling of AI powered job search mode that auto suggest suitable jobs

Digital Certificates

• A solution to fully digitise academic and professional certificates. Replace paper format.
• Institutes and Corporations can publish digitally signed certificates to users’ Linkedin IDs
• Linkedin Junior : A minimal version of Linkedin to onboard school students with a unique Linkedin ID for receiving certificates. Handled by parents or school authorities.
• Verified entities : Organisation and institute profiles can be verified to increase authenticity of the certificates

Revenue potential : A monthly fee can be charged to the publishers of certificates, for providing them a permanent digital repository

New age hiring tools

• Shared Talent Pool : A feature that allows companies recruiting for similar roles within same timeframe to share candidate interview feedback , thus speeding up hiring for all
• Warm Pool Creator : A solution to the problem of talent shortage for niche skills. Companies can post tentative openings several months ahead with suggestions for required skills and preperatory courses, thus giving time for interested candidates to upskill and apply for the roles later
• Digital offer letters : Companies can send digitally signed offers to candidates Linkedin ID. Candidates can digitally sign and accept offers received via Linkedin
• Organisation Chart : A feature that allows the management and HR to draw organisation charts of current employees, map potential hiring needs in the chart and auto-convert them to job postings in Linkedin

Revenue potential : The above tools could be part of the premium tiers

Quick Company Launcher

• A platform that enables to setup the digital presence of a company in a few minutes
• Provides the best info and service provider details for legal registration, licenses, etc
• Coworking Space aggregator : Singe place to book seats across any coworking space provider

Revenue potential : Monthly fees for website and email hosting , as per standard cloud charges. Commission per seat of coworking space booked through Linkedin. Referral fee for service providers listed in the platform

Linkedin University

• A platform that catalogs and compares the online courses from all major edtech companies
• Recommendation of courses based on topics of interest , career plans , ratings by others, etc
• Option to sign up for courses with Linkedin ID and auto publishing of certificates on completion
• Scholarships for enrolment and premium job search features on graduation
• Tie-ups with corporates to evolve this platform as the ideal destination for employee upskilling
• Digital textbooks : Option to search, purchase and read relevant e-books within Linkedin

Revenue potential : Referral fee can be charged to the different course providers for signups received through Linkedin

Linkedin Surge

• A vernacular app that empowers blue collar and skill based professionals
• Notifications on all Public Sector jobs with details of application
• Upskilling Tools : Latest tutorials on skills needed for different industries
• Digital credentials : A digital repository for experience certificates and recommendation. Option for employers to create digital certificates within Linkedin even for very short work duration.
• Trust Score : Linkedin as a common platform for employees to build a trust score based on feedback, user rating, etc from across different jobs
• Mobile School : An option for semi skilled workers to improve their educational qualification part-time by learning through the app and take exams later to earn school or college degrees
• Continued Benefits Hub : For sectors where hiring is for part-time or short term contract openings, option for workers to keep benefits like family health insurance always active via a minimal payment, till next gig
• Assisted job search : Option for assistance by professional volunteers for digital job search

Revenue potential : Basic fee for courses and other services provided

Career Lamp

• A chat buddy that answers any queries on education and jobs
• Converts user’s conversations to queries on Linkedin’s massive database and returns a customised and context aware response

Revenue potential : All basic conversations must be free. Advanced insights maybe restricted to paid users of the platform. Also course providers and institutes can be possibly charged for referrals

Events and Conferences Center

• Event Page : A page for organisers to publish and promote events
• Speaker Invites : Organisers can do profile search and send formal invites to potential speakers
• Event Registration : Linkedin users can find all education & professional events and register / RSVP in one click using their profile
• Option to attend virtually through Linkedin Live
• Enhanced event interaction section that makes Linkedin the preferred destination for event day conversations over other social media platforms.

Revenue potential : A small percentage of the ticket fee for bookings done through Linkedin

Linkedin for Institutes

Ready to use solutions for admissions, alumni connect & placements

Revenue potential : Monthly or annual fee for portals built using Linkedin

Linkedin for Companies

• PR Hub : Linkedin company page as the center for official Press Releases by companies.
• CSR and Sponsorship : Linkedin as a platform for companies to collaborate with external stakeholders for their CSR initiatives and to receive sponsorship requests
• Widgets and APIs to power team, blog and career sections in websites : A new set of tools that enable teams to import jobs section, team profiles and blogs from Linkedin and display them in relevant pages in the official website by embedding as widgets or using Data APIs

Revenue potential : Monthly or annual fee to companies for using these features

Linkedin Contact Cards

• A new section within Linkedin that collects the QR contact cards. For each card, send a connect request in Linkedin or import into CRM tools
• Multiple card versions : A person can create multiple versions and share chosen one.

i, Casual contacts : This card version shares only work email and landline number

ii, Trusted contacts : This card version shares personal email and mobile number

• Dynamic updates : When a person updates his/her contact details, the same will update in their contact cards collected by others
• Option to scribble meeting details for each card

Revenue potential : A small fee for collecting more than 1k+ cards, etc. A fee to enable option to import to CRM tools

A, Elimination of typo errors

A1, Edit option

• Important tweets should not be ruined by typos or simple mistakes
• Allow tweets to be edited upto a certain number of words or characters to eliminate typos. Text parsing logic analyses if edit is for typo and is permissible and does not alter the full context of the tweet
• Provide a max editable window of 24 hours.
• A version history button to view the full edit history of a tweet, thus avoiding misuse of the feature or tampering with original tweet
• Indicate edited parts of a tweet with subtle coloring or underlining

A2, Spelling and grammar check

• A service, similar to grammarly, that is inbuilt in the compose screen
• User still has option to ignore the suggestions and tweet as they like
• Can possibly charge users for enabling this feature for the twitter account

B, Contextually apt reactions

Primary reaction other than like

• Twitter’s current single option of user reaction ( Like) keeps the interface clean
• Different contexts may have different more appropriate primary reaction, like condolence, anger
• Allow author of the tweet to set the primary reaction type for her/his tweet

C, Better organisation of tweets

C1, Enhanced search of DMs, bookmarks and interacted tweets

• DM Search : Option for users to search using keywords across all chats and within a single chat
• Bookmarks search : Option for users to search bookmarked tweets by keywords , tweet author, date, etc
• Own & interacted tweets search : Option for users to search their own tweets & replies and also the tweets they have liked or retweeted, by using keywords

C2, Multiple categories for bookmarking

• Provide users with min 3 categories of bookmarks to organise their saved tweets better
• Color code the categories and indicate a bookmarked tweet with a colored line for that category
• Can include default categories like books recommendations or user can create custom

C3, Notes and reminders for a tweet

• User can write a note behind the tweet for their own future reference
• Option to set a time reminder for taking any further action
• Provides a convenient option for user to organise the tweets they like or bookmark

C4, Channels for different tweet categories

• Highly followed users tweet about different categories like : career, personal life, hobbies
• A user can categorise her/his tweet by choosing a channel. Max 5 channels per user
• Followers can follow a user’s full profile and view all tweets or choose specific channels only and see only tweets in those channels

C5, Soft copy of text in screenshots and pics

• Many posts contain screenshots of long posts from other sites like Facebook and Linkedin or pics of books, newspapers, etc
• An OCR service enables users to get a soft copy of the text in such pics that are included in a tweet
• This will enhance the amount of reusable info that users get out of twitter

D, Enhanced threads experience

D1, Easier thread composer

• A composer where the user can type the entire content of a thread in one single window continuously instead of multiple tweets using + button
• Twitter auto suggests breakpoints at each 280 chars to help user to plan complete sentences as one tweet.
• Option to enable auto tweets counter instead of user typing the count
• A toggle switch to enable this threads composer or existing format
• Option to publish as full article to Linkedin, Medium, Facebook, etc

D2, Enhanced thread reader

• Option to read an entire thread in a single continuous text format
• Readers should be able to retweet, like and reply to the entire thread instead of current practice of reacting to first or last tweet in the thread
• Subtle indication of breakpoint for each tweet and option for users to tap to go into an individual tweet and see the replies , likes and retweets

D3, Threads for long posts and commentary

• Official press releases will display all tweets together in a continuous manner with subtle indicator for individual tweet breakpoints
• Sports commentary will display tweets with timestamp on the left and text on the right to make it more similar to conventional commentary format

D4, Audio listening option for threads or tweets using TTS

• A Text-to-Speech based engine that allows users to consume tweets, especially long threads in audio format
• Users can ideally create a playlist of threads or tweets for later listening and play all of them sequentially when needed

E, Improved timeline experience

E1, Multiple read modes

• Different mode options that allow the user to select the category of tweets to be seen in their timeline in the current session based on the amount of time they want to spend
• Short session mode — User sees only tweets and retweets of whom they follow
• Medium session mode — User also sees the tweets liked or replied by whom they follow
• Full session mode : User also sees the all other suggested tweets from full feeds algorithm

E2, Read tweets tracker mode

• Roughly estimates whether a tweet was read by a user or not based on amount of time spent on a tweet while scrolling
• Tweets that are already read by the user will not be repeated in her / his timeline
• This enables user to see maximum newer tweets while scrolling through twitter timeline
• Reduces the chances of the user missing important tweets
• User can disable this mode to experience twitter timeline in the existing normal mode

E3, Improved character counts & mentions display

• Make tweets with large number of mentions more readable by displaying the mentions separately from the tweet text and without counting towards 280 character limit
• Easier to understand mentioned profiles by displaying full names along with twitter ids

E4, Custom tweet viewer in Whatsapp

• Tweets shared in whatsapp appear like any other web URL and offer no custom interactions
• Increase engagement of tweets with options to view, like or retweet from within Whatsapp
• User can link her/his twitter id to whatsapp and complete various actions using that id
• Auto create ids for for users without ids using their phone numbers

E5, Summary of replies to a tweet

• Many popular tweets with open questions to the audience may have large number of replies
• Answers get very repetitive when a large number of users reply
• Text parsing based logic that allows the viewers of the tweet to see a rough summary of the replies in terms of the most popular answers , the number of users who gave an answer, etc

E6, More informative user profiles

• More information on a user profile page like the major topics that the user tweets about , level of recent activity, most liked and retweeted tweets
• Option for the user to showcase multiple pinned tweets in carousel format
• Common tweets liked, retweeted or replied which shows similarity
• Easier search of user profiles based on multiple input parameters. Example : i, Doctors in Poland, ii, Techies who attended Arizona State University
• Option for users to get verified indicators for their profile credentials like education, employment, etc to increase profile authenticity

F, Better discoverability of popular older tweets

F1, Twitter magazine

• A search window that surfaces highly liked interesting content
• Based on a user’s search keyword suggest amazing and varied contents that were liked by others or highly appreciated
• Auto build a curated set of posts based on topics and suggest to users in a magazine format
• Similar suggestions of popular, interesting posts based on tweet locations

F2, Older tweets as an article

• Some of the best contents in twitter were already tweeted and not recent
• A mechanism to rediscover some of the best contents that were already tweeted in the past
• Authentic contents that are very informative can be tagged as textbooks for particular keywords or topics and can be resurfaced again to users who have not seen it before
• Ideally such tweets or threads can be shown to new users of twitter based on their topics of interest

G, Reduce online abuses and enhance safety

G1, Self verification of profile pic & name and green ticks

• Users with original profile picture and original name are more likely to be at a better behaviour in social media
• Most common abusive profiles have fake names and random pics
• Eventually all twitter profiles will have either blue ticks (twitter verified) or green ticks (self verified)
• Non-verified profiles can be restricted in their features and given low priority in feeds & replies algorithms

Profile pic verification tick :

• An app section stores reference pics of the user by asking to click selfies with certain movement patterns to ensure that it is a live authentic person
• User can then input the intended profile pic and the app will run a Deep Learning model to compare with the reference pics and generate a verified tick for the DP if the pics are similar

Name verification tick :

• User can use an app section to send a name verification request to a service provider with whom the user had already done a verification using government id, like banks, insurance
• The partner service provider returns a name verification token to twitter for user’s full name

G2, DM Filtering

• Text parsing + image detection to fully block any abusive content in DM, before reaching receiver
• In case of suspicious content that cannot be fully classified as abusive or not, the receiver gets a choice screen whether they want to see the suspicious DM from the particular user
• Warning strikes and eventual auto-blocking of profiles of users who send abusive DMs or tweets

G3, Equality mode

• Many early and less ferequent users may find twitter more welcoming if they don’t feel intimidated by the various counts
• If a user turns on equality mode :
  i, Followers & Following counts of the user will be hidden. Other users can only see some common followers stats
  ii, Likes & Retweets counts will be hidden for the user’s tweets. Other users can only see an indication of the likes and RTs range for popular tweets of the user, like >1k

G4, Safety mode to avoid accidental likes or tweets

• Provides a precaution against a user accidentally liking a tweet while browsing in mobile. With safety mode ON, a user has to double tap on like button to like a tweet
• Adds protection against an individual posting unnecessary tweets from an official corporate id by requiring two or more individual ids to approve a tweet before posting

H, Freemium tiers

Twitter could introduce the new features as paid tiers

• Free to use : existing twitter features
• $1 per month — Above + Grammarly spellcheck + 3 bookmark types + Enhanced search of DMs, bookmarks and interacted tweets
• $5 per month — Above+ 10 bookmark types + twitter magazine + notes behind tweets+ more informative user profiles
• $10 per month — Above + Read a certain number of articles across multiple paywalled sites without any extra payments or logins. Example :The Economist, The Wall Street Journal