The article also doesn’t dwell on different applications of ML. It assumes that the PM or a business leader has identified an application and wants to apply ML to it. I rather emphasize that finding a “real” problem is one of the most critical tasks for a PM in a Machine Learning project.

Lastly, I have aimed to be comprehensive yet consumable so that readers don’t save it on their digital memories and never get to it. Hence a few basic items will not be defined here; a quick google or a GPT search will suffice. :)

Contents:
1. What is Product Management
2. PM in ML/AI — Be Strategic
3. PM responsibilities in complete ML Life Cycle
Scoping → Data → Modeling → Validation → Deployment
4. Additional comments
5. Tools and techniques

What is Product Management?

Product Management is a vast role and significantly differs from one company to another. I recently read an excellent 600-page book on Product Management called, Sacred Seven: The skills required to be a world-class PM, and I cannot compress that in this document. However, the image below from McKinsey very well generalizes what PMs do on a daily basis.

PM in AL-ML

Companies rushed to the cloud; now, will they rush to AI?

Companies, big or small, rushed to cloud computing (or got oversold by the Tech consulting companies), but more than 50% of companies cannot reap benefits. The quote below is one of the reasons why?

“Many organizations made mistakes during the pandemic with overly complex infrastructure and without proper consideration of the operational impact.”
 — David Linthicum, chief cloud strategy officer, Deloitte Consulting LLP

The GPT wave means there will be a mad rush to implement AI, and a good PM will ensure that their company does not repeat these same mistakes.

PM role: Be Strategic

Companies, as well as individuals, are all trying to ride the AI wave. Companies include “AI” in their domain name and product description, and CEOs use the word “AI” or “ML” in investor calls to lure investors. Nothing is damaging to the company with that unless you have nothing to do with AI, and then FTC catches you ☹. However, when developing a product, a PM needs to be more strategic and ask themselves if s/he really needs AI. Investing in AI for the sake of it can lead to a loss in value and degraded customer experience.

A PM in AI must consider Return on Investment (ROI) and perform a Cost-Benefit analysis before undertaking the project. ML Product life cycle is not as deterministic as traditional software development; hence, it has higher risks than traditional software product development. So, a higher discount rate should be attached to AI projects. Performing cost-benefit analysis and calculating ROI is complicated because everything in an AI project is interweaved, and your initial estimates will certainly be quite off. Referencing industry case studies and competitor results could be helpful. Develop a hypothesis, make assumptions, and build a model. Define success metrics and define what success will look like.

I have slightly modified a framework from Google PM, James Smith, to determine if you should use AI/ML.

As a Product Manager, one should not be moved by how cool an ML model, experiment, or product is but rather by how significant the business impact the product is making; Positive business impact should be the number one goal. In many coffee chats and conference talks, I have heard many ML Product managers say they have made a bigger business impact by saying no to ML. ML adds complexity to the system, and complexity costs a lot.

Team capability (Who should be on the team?) — Team structure

Some companies try to hire Data Scientists or Machine Learning engineers who can identify patterns, develop models, deploy them, prepare dashboards, and keep them updated. This is unrealistic. However, what companies can do, at least initially, is to hire contractors. Contracting has the drawback of losing know-how when a worker leaves, but it could be effective before the business case is validated. Or start with one data scientist/ML Engineer, and then you will quickly see that need for Software Engineers and IT (Operations) people will grow much faster than for data scientists.

AI/ML Product Life cycle

The McKinsey product management tasks are quite general. Now, let’s see how they apply to different ML Project stages. We will go through each of the five stages, look at the PM’s task, and then make some additional comments.

Two key points I want to make from the above diagram:
1. ML is an end-to-end iterative process.
2. And hence there is a lot of overlap between all the five stages

Scoping

Defining the Problem and Goal

The Product Manager’s primary job is to define the problem that team is solving. Though on the surface it may seem straightforward, a PM should strive to define/explain the problem with such clarity that if that problem statement is given to a room of people, everyone will come up with a similar high-level solution. Essentially that problem definition becomes the driver of the solution(s). This is where a PM will spend much of their time. As with any other project, the problem should be customer focused (internal/external) and will be defined after extensive customer research and discovery.

Define what you are trying to beat. — Establish benchmarks against Human-level performance, current firm capabilities, and industry literature. Has the competition deployed AI/ML to deploy for a similar problem?

Why is TikTok’s recommendation engine better than Instagram’s and YouTube’s?

TikTok was designed to be an ML app from day 1, for IG ML features were added. So TikTok was designed to see only one video at a time, whereas in IG, a user could be seeing multiple images/videos at a time. It is easy in TikTok to measure user interest and engagement for video but not so for IG. The takeaway is to involve UX and ML teams in product definition so such parameters can be considered early on.

Define rough order deployment infrastructure.

Work with the Technical lead to translate the user requirements to product specifications in detail. Things like where the model will be deployed — Cloud, On-prem, or user device and batch or real-time inference, latency requirements, security and privacy constraints, Input data, outputs, metrics, etc. Such specifications will allow teams to work on cost vs. latency and throughput tradeoffs from the get-go.

Compliance with GDPR and CCPA should also be discussed from the prototype stages. Adding compliance at the deployment stage makes operations complicated. For example, GDPR has the right to be forgotten. That means the company must delete user data from ALL the databases upon a user’s request.

Provide Proof of Concept and Production goals:

POC Goal is to determine if the application is workable and worth taking to production. Refrain from spending time establishing data/model pipeline or automating the system. Encourage your team to be scrappy, work on notebooks (Jupyter notebook, etc.), and take notes so results can be reproduced. Use as many open-source tools and end-to-end managed services as possible to quickly deploy a model. For an average use case ML model, this can be achieved in 3 sprints (4~6 weeks).

The idea here is to experiment and fail fast and learn faster. If you are a small to medium size company, then you should not invest in creating your own ML model. Use transfer learning (re-use an established pre-trained model on your learning task).

Define Go vs. No-Go Metrics.
Early on, discuss the model and data dependencies and what-if scenarios.

Production phase goals: Encourage the team to utilize sophisticated tools to automate the system. A high automation CI/CD/CT/CM (Continuous Integration, Delivery, Testing, and Monitoring) level will keep the models fresh. PM’s job in production is to ensure that improvement in ML metrics indeed translates to improvements in business metrics.

Why do ML models get built but not deployed?

This question partly inspired this article. I recently read that Gartner estimates that 85% of the built ML models never get deployed. I have taken a few classes where I have learned how to build a recommendations system and image classification models within a few weeks, but can I deploy those models and make them a service for people to utilize? Not so easy. This led me to dig deeper and then take a Machine Learning Operations (MLOps) course.

Teams and companies underestimate the technical stack required to deploy and maintain an ML product. A google paper pointed out that Machine Learning resource management, server infrastructure, monitoring, and data pipelines are much bigger pieces of an ML system. The image below from the paper shows that.

This tweet from the year 2018 says it all. However, many tools have been developed to address the problem since then. The last section of the article has a drawing of tools to be used at different stages of the ML project. Additionally, in deployment, a system is never just a single model; multiple models interact with each other and with other parts of the system/microservices, increasing the complexity of the system. Additionally, in most companies, there is a big wall between Data-Scientist and Software Engineers, and PMs have to work to bridge that.

Setting the culture

PMs are responsible for setting the culture of teams. This culture will play a significant role in ensuring that the product developed respects privacy and is fair, unbiased, and useful.

Data

PM should be involved in the complete Data Life Cycle of an ML project and ask the following questions.

What to do when data is not available?

■ Can you collect Live data from your business?
■ Web scrape to get data? Forums where customers discuss your and competitors’ products?
■ Open data source Gov, Kaggle, Google, Amazon, etc.
■ Purchase data from someone like Neilson.
■ Synthesize good data by combing data from multiple sources.

Feature Engineering:

Through feature engineering, you can create new data that is valuable for the model. As the number of features for your data increases, the cost of training a model and serving inferences increases, so creating and selecting the right features for your problem is both an art and a science. To generate these features, you should involve a domain expert on the ML problem you are trying to solve and take their know-how.

Utilize a data-centric approach rather than a model-centric approach

“A reasonable algorithm with good data will often outperform a great model with not so good data” — Andrew Ng

“No other activity in the machine learning life cycle has a higher return on investment than improving the data a model has access to.” — Gojek

Focus on good data that
■ Covers all the important edge cases
■ Ensure it is not skewed or biased
■ Consistency in labels
■ Establish a system/process to provide timely feedback from production data on data and concept drifts.

Items to discuss with Engineering Manager or Technical lead:
■ Data tracking infrastructure: Data provenance and data lineage.
■ Compliance with GDPR, CCPA, and other privacy laws.
▬ Right to be forgotten — Request to delete user data from ALL data sets and copies.
▬ Right to ratification — Edits/corrections in all the data sets.
▬ Right to access user’s own data.
▬ Other GDPR rights.
■ Discussion about security measures to be taken.

Model development

After the goals and data source are well defined, the PM steps back and lets a team of engineers work on the model development to develop options. Bring collaboration between Data Scientists (or ML Engineers), Data Engineers, and IT/Software engineers from the initial phases of the project. This will ensure the input (data) and output (model) infrastructure comes up along with model development. I speak from my experience as an engineer, engineers like to build and deploy fancy stuff, and a PM must bring discipline and focus so that the system is not overbuilt. This is where the PM sets priorities.

One non-negotiable ask from the engineering team is the reproducibility of the results, and that will require a team effort. The PM will have to bridge the gap between Data-Scientist and Software Engineers. It will also require new processes. In software development, there is standard practice to track code, but for ML, along with code, the engineering team needs to keep track of experiments, which will require keeping track of combinations of code, model, data (used for that experiment), hyper-parameters, and associated results. There are tools such as Neptune.ai developed specifically for ML experiment tracking.

Code tracking → Model tracking → Data tracking → Hyper parameters tracking → Result from tracking

Items to discuss with the tech lead:
■ how do we measure the performance of two models besides just considering the accuracy
■ Continuous Integrations & Continuous delivery
■ Continuous Testing
■ Testing model over slices of data that represent all use cases
■ Stages of MLOps (Levels 0,1 &2)
■ Data & Experiment tracking infrastructure
■ Reusability of models and Modular design
■ Use of managed services

The last section of this document shows a diagram of tools and techniques for different ML Product Stages (Not exhaustive). These are technical items, and the decision to use them lies with the technical lead, but the PM should know them and understand them at a high level because each of these has cost and user experience implications that ultimately affect ROI.

Product Validation:

The Product Manager validates the results of different ML models through A/B testing. A/B testing is powerful, but more advanced testing methods such as Multi-Armed Bandit, and Contextual Bandit should be utilized for multiple model testing. Product Managers must work with IT managers to develop a system where multiple models can be tested, as model performance on test data will significantly differ from real-world live data. In this blog, Netflix describes a method they used to fast-track the testing of hundreds of recommendation models.

Models carry the bias present in the data. PM can and should build negative test cases to prevent poor outcomes by testing the system performance on slices of data. ML systems need to be fair, accountable, transparent, and explainable. A poorly modeled system can cause representational harm, opportunity denial, disproportionate product failure, and harm.

Translate model metrics to business metrics:

PM translates improvements in model metrics to business metrics and observes that relation continuously. For example, a 10% improvement in CTR (Click Through Rate) may translate to a 100% improvement in ROI for marketing activity. Similarly, a 10% improvement in product recommendation could mean driving more traffic to that one vendor. Because of the flywheel effect, other vendors eventually drop out of the system (an undesirable outcome). PM has to take a holistic view of the system and make decisions for positive business outcomes.

Key takeaway: ML Metrics are not business metrics. The relationship between these metrics is complex, and the PM has to hypothesize and validate the results through experimentation. As far as I know, only through experimentation can one develop a relationship between ML metrics and business outcomes. This is a stage where you have to tie your ML metrics to business metrics for successful outcomes.

A deep understanding of ML metrics like Precision, Recall, AUC, F1 score, etc., is a must. These metrics won’t be defined here, but I would encourage aspiring ML PMs to dive deep into metrics to a level that they can build an intuition to know what metric would apply to a problem. Running a regression (y on multiple independent variables x) takes a few mins, but the analysis (to understand the story that data is telling can take a few hours). Attention to detail is paramount; extra attention to the outliers will save you headaches later.

Suppose you have a few contending models with similar performance. In that case, you may want to go with one that is inherently explainable, even if it is slightly inferior to your best model. Internally and externally, users will be reluctant to use the ML model, the results of which cannot be explained. If your ML is like a black box with Billions of parameters, how do you place guardrails on it? There is no perfect answer to this, nonetheless should be discussed with the team.

Deployment

• Work with marketing to provide IT with an estimate of growth/traffic/user requests. ChatGPT grew to 1 million users in just five days. While this is great for the company, it comes with a cost. Providing estimates of request growth to the IT team and working with them will enable cost-effective and efficient scaling.
• PM must also work with Software Engineers to define automatic triggers in the system for data and concept drift.
• Define when and how frequency re-training of the model will be done. Automatic re-training of the model is not recommended; triggers should be set to alert when new data and concept definitions are available.
• PM must work with legal teams to ensure models adhere to the regulatory requirement of the jurisdiction.
• Protect user data against attacks: Work with the model development team to safeguard against adversarial attacks. Adversaries can fool the model to produce undesirable outcomes, steal the trained model, and, even worse, recreate the training data just by utilizing the models and their outcomes.

For a company as big as Airbnb develops its tools, most of the companies outside of the hyper-scale companies are better off using managed services or open-source tools. MLOps is an extension of DevOps, which is standardized, but the key difference is the unknowns in MLOps. I am repeating myself, but almost everything is iterative, which makes it challenging.

Messaging to your user

Models are not perfect; in some cases, even with just 60–70% accuracy, it still may make business sense for them to be deployed. But the challenge is to address your customers and set the right expectations with them. As human behavior goes, people remember failures. Your model may provide false or nonsensical results only one out of ten times, but people will discuss that failure more often. That is how humans work, so there is heavy lifting needed from PMs and PMMs to ensure users understand the limitations of the service.

Things that will help:
■ Think hard about the way you present results to win user confidence.
■ Provide easy-to-read disclaimers,
■ Provide confidence levels or range in the results,
■ Where possible, show the top 3 recommendations/outputs instead top 1
■ Provide explanations, a single line mentioning the key features used for the output.
These items will go a long way in winning users’ trust.

Models will fail, be ready for that moment

ML Products in production or testing will go wrong and create negative PR or a crisis moment; this is where PM must be ready to handle crisis moment gracefully. Soft skills are critical here. Being transparent with your customers from the beginning will make sailing through such episodes less painful.

Additional Comments

Two types of AI products
1. AI Tools
2. Developing products using AI capabilities.

If you’re a PM developing an AI tool, for example, a feature store for AWS’s Sage maker, knowing technical details is necessary. After thoroughly understanding the workflow, you can define gaps, identify customer pain-point, and propose solutions. However, suppose your product utilizes AI capabilities to enhance its value to customers; then, one must know the capabilities of different APIs available in the market and development.

Keep up with the advancements and limitations.

Recently, I underwent a rigorous class in Software Entrepreneurship, which three legendary professors concurrently teach (Greg Gottesman, MD and co-founder of PSL, Ed Lazowska — Professor, and the Bill & Melinda Gates Chair Emeritus and Oren Etzioni — serial entrepreneur, CEO of the Allen Institute for AI). In this class, business school students are teamed with CS and Design school students to develop an idea and a solution within ten weeks. Many teams developed mind-bending AI/ML products within 5–6 weeks (the first four weeks were spent on ideation). Products like on-demand podcast translation services, including the translation of emotions! Not taking any credit away from the amazing capabilities of UW CS students, but what I learned is that by using existing APIs, one can develop great AI/ML products in a short time. PMs can innovate ways to solve customer problems utilizing APIs readily available. Knowing the high-level technical feasibility can drive many ideas. Additionally, while the speed of development is important, modularity is equally important. Developing solutions such that one set of APIs can be easily replaced with a new, better-performing API is necessary. Within weeks after OpenAI released its GPT 3.5 API, Notion used them and deployed them into the platform. They need to be able to switch easily, let’s say, Google Bard API if that proves to be more efficient and accurate over time.

Repeating for the third time: Almost all aspects of Machine Learning Product Development are iterative — iterate on data, models, and features. It is dynamic and continuous, fast-paced, and highly rewarding. Keeping up with advances in other technologies such as cloud computing, semiconductors, sensors, and IOT will open up a myriad of possibilities for solving problems through machine learning.

Knowing that ChatGPT was trained on data until Sept 2021 data will tell you that it cannot be used in your news aggregator app.

I recently completed a specialization on Coursera: Machine Learning Engineering for Production (MLOps) by DeepLearning.AI (Famous Andrew Ng’s company). Below is a visualization of techniques, terminologies, and tools I learned through the course.

ML Life CycleL: Tools, Techniques, and Terminology — (Certainly not exhaustive)

Thank you for reading it. I hope you enjoyed it and found it worth your time. If you want to discuss any aspects with me, please contact me on Linkedin.

Thank you,
Ishan Shah, Engineer & MBA
Linkedin.

Tesla has ways to go before achieving its target costs and production capacity. To many outsiders showcasing robot development is yet another distraction. Any other CEO would be punished for undertaking such endeavors but not him.

Reasons Elon gave for Humanoid robot development:

1. “The foundation of Economy is labor” there is a labor shortage, and one day envisions that these humanoids will work in Tesla factories.
2. He thinks these robots will help people with day-to-day tasks.

The other reasons that I think justify robot development:

1. Elon is working hard to make Tesla “more than a car company.”
2. To promote R&D, knowledge can be shared with other product developments.
3. R&D people need some public consumer focus/attention to keep them motivated. That event was dedicated to engineers who worked on the robot. They took center stage.
4. Showcase company expertise. Honda, in the past, has developed humanoid. People knowing Tesla’s engineering prowess will trust Tesla cars.
5. Tesla, in the past, has raised small capital by selling toys to its fan base.

What would be my top priorities for Optimus Robot

1. Do absolutely no harm, Issac Asimov’s first law of robotics. :)
2. Robots are painstakingly slow — If you haven’t observed, humanoids, except the Boston Dynamic’s dancing bots, are too slow to be used for a practical application. Beat the benchmark by 10x.
3. Intelligence needs companionship. The robot should not just serve as a laborer but be a companion. The world is crowded yet getting lonely.
4. Specialize in the tasks of a caregiver before learning the skills for the factory. Cobots and industrial robots are great in factories; having humanoids are caregivers will make them accepted in other arenas.
5. Like millions of Teslas continuously learning, these robots need to be out there and learn.

Recently my friend made a statement, “There is a common consensus that stock trading adds little to no value to society.” Having followed stock markets since my childhood, I was only partially surprised to hear such a comment, the only surprise being that the friend is a fellow MBA student. I have heard similar statements many times before and know that many people hold the same view, but hearing it from a fellow business student intrigued me to write this post.

Harvard Professor Mihir Desai poses this question in his Book, How Finance works? He asks whether the people in the Financial markets, the traders, analysts, managers provide any value or are they leechers sucking value out of companies and ordinary people. He says that these people are solving the problem of asymmetric information. While there are many views to this question, I want to put my thought into words.

First of all, we have to understand that any market comprises people, its participants; Buyers or Sellers, Providers or receivers, and agents. So a statement that puts financial markets in good books but looks down upon the people of those markets is ironic.

So how does stock market/stock trading provide value to society? First and foremost, it provides a mechanism for companies to raise capital. Whether that company deserves to raise money and how much and at what cost is left to the people. Even though the process of raising the capital by issuing stock is initiated and assisted by Investment Bankers, once the stock starts trading in the open markets, it is assumed that Efficient Market Theory will work. The theory says that over a period of time, the market will fairly price a stock. I agree there are many subjective words used, but most of the time, collective intelligence works.

The financial market (includes insurance market, bond market, etc.) employs more than 7million people in the US. This represents significant livelihood. To the participants of the stock market, it provides a way to build wealth. Beat inflation, not only sustain the value of your money but increase it. The stock market has also destroyed people’s savings and retirement funds, but it has been the best vehicle to grow your money over a long period, sometimes the only vehicle.

Stock markets have rewarded the companies that have saved our society. Think Moderna, should not we reward the company for coming up with a vaccine against COVID? or should we also not reward the people who backed Moderna not only with confidence but with their money? The stock price of Moderna rose from $20 in Jan 2020 to $416 as of this writing. Whether this reward is fair or not is subjective and again left up to the Efficient Market Theory. Another thing is thousands of people played a role in vaccine development, but few were disproportionately rewarded. However, this problem is of equity and company structure, not of the stock market. What about the people who are in the middle of the stock market? Do they add any value? Absolutely they do. They bring together the people — the help in making a robust market. They help in bringing liquidity in the market; the stock market is the most liquid of all markets. How much would you reward them? How much would you reward a real estate agent that brings buy-side and sell-side together? These people in the middle help transfer the information and bring people together. If you value and respect facebook for bringing people together, then you value these facilitators of the stock market.

Another thing to point out is the stock market is a very small part of the wider financial market which comprises bonds, insurance, etc. The bond market is multiple times bigger than the stock market.

Like everything, Financial markets have their problems. The main problem is asymmetric information. Another significant issue is equity; Not all participants have equal influence. Another issue is accessibility, but with technology, far more people are able to access these markets than before. The stock market forces the managers to make short-term decisions that benefit just the stock of the company. This short-term behavior is not only damaging to the company’s health but, in many cases, to the environment, employees, and the general public.

This article is not meant to be an exhaustive list of how financial markets add value to society and the issues in it. I would like to invite my classmates at the Foster school of business to a healthy discussion about what Financial markets bring, their problems, and ways to solve them.

A Black Swan event has hit humanity and it is called COVID -19 or Coronavirus whichever you prefer. The disease is called COVID-19 and it is caused by the SARS-CoV-2 virus. The World Health Organization (WHO) declared COVID — 19 a pandemic on March 11th, 2020. More than 12000 people have died as of this writing. A recession is a sure shot, big hotel chains have already started to lay-off people. To know more about Covid-19 please check out this amazing video which talks about how the virus attacks the human body, some in-depth biology, more about the pandemic and about the actions that are taken around the world. The worst pandemic of the 20th century, the Spanish flu of 1918, killed about 50 million people worldwide. Few argue that the healthcare system now is 100 times better than it was in the past. The counter-argument is that the connectivity of the world is also 100 times better than the past.

To get the latest infection updates please refer to these websites.

1. Coronavirus Tracker from The John Hopkins University.

2. The message from the WHO is simple “Test — Test-Test”, so how good is testing in your country. Check out this website.

Why Some countries are affected more than the others

There are multiple reasons:

1. It depends on how far behind or how far ahead your country is on the exponential (or ‘S’) curve.
2. How soon travel restrictions were imposed by your country.
3. How early did testing and effective quarantine start in your country?
4. It also depends on the connectivity that a country has with China the presumed source country. For example, every day before travel restrictions were in place more than 27 flights landed in Boston from China every day, whereas less than 30 flights connect India with China. Of course, once the disease was global India saw tertiary effects. But countries such as the EU, the US, Korea, and Japan are more direct hit.

Is summer going to reduce the infection rate of the virus?

No one knows for sure. If there is one message that I would like to pass on via this article is this: We (world) know(s) very little about this SARS-CoV-2 virus. Even though we know about the family of this virus the characteristics that this virus has shown are unique, to say the least.

There is some research that suggests that the virus will slow down with high temperature and high humidity. Here are other articles(1) and (2), and videos(1) and (2). This research, however, is not peer-reviewed. Nonetheless, this may also partly explain why India has seen so few cases considering its population density and more than 90% of the cases in India are with travel history. The other reason is India is only doing limited testing. As testing ramps up we will likely see an increase in the numbers of infected people. India is still in-between stage 2 and stage 3 and it is hoped that all the quarantine measures will break the chain.

How long will it take for drug development?

Although I am not a medical expert, I can confidently say that it will take at a minimum of 1–2 years. Allow me to quote something

“On average, it takes at least ten years for a new medicine to complete the journey from initial discovery to the marketplace, with clinical trials alone taking six to seven years on average.”

So hoping to see a magic drug is nothing but naivety. Yes, there may be a symptom reliever like many for the common cold.

The young population that was presumed to be safe is not so safe after all.

The initial data from China showed that the young population may be much less affect by Covid-19 than the elders but data from Italy and the US is proving to be otherwise. The WHO gave its own warning. The initial chart of cases in the US looked like this.

This news is both bad and good because the young generation will take this virus scare more seriously. The parties on the beaches must stop. The people who knowingly escape quarantine measures are nothing less than mass murderers. These selfish people place the whole country in danger. Such cases show the limitations that democracy faces in fighting this novel disease. These times of emergency call for extraordinary measures and justify curtailing the right of people until the crisis is resolved.

Now Let's dive into the economic aspects.

The global economy is in shock. Suddenly the 87 Trillion US dollar global economy has come to a sudden stop. Financial markets all around the world have fallen by 30–35% from their peak. Most of the money managers were caught off guard by the speed with which the markets have fallen. The fall is unprecedented and something that has never happened before. Suddenly there is no supply and no demand, this is something which the world has rarely seen before. We are entering a global recession, how long will it last largely depends on how long the Coronavirus lasts.

Why did global financial markets fall so sharply?

The image below shows how fast the stock market became a bear market. Just within 16 days, the market fell by about 30% from its peak and as of this writing, the Dow Jones Index is (-34.8%) from its peak just a month back.

So why fall so sharply. Many reasons, firstly fear of unknown unknowns. This event is taking the world into uncharted waters. This crisis is not caused by a bank failure or a war but by something which humanity has not seen in a half a century. The last pandemic happened in 1968. To know more about past pandemics look here. This event is not a limited event.

Secondly, more and more traders are using advanced technologies for their trades. US hedge fund managers have in their disposal their own private satellites, with which they take pictures of retail parking lots to estimate the sales a retail chain might have during a festival, they even take pictures of the oil storage tankers of all the countries around the world and estimate the world reserves. Some even measure the amount of pollution in the atmosphere above an industrial area to estimate industrial production. With this advanced information, these money managers act fast. So the second reason is the speed of knowledge spread, which is undoubtedly faster than anytime before.

The third reason is that the past bull market was the longest bull market in history. For the past 2–3 years people were anticipating a correction, they just did not know when. Now that a crisis is here every investor is rushing to the doors.

2008 VS 2020

Naturally, everyone compares and everyone is comparing this downfall with 2008. 2008 crisis was caused by a structural flaw in the system and was a product of Financial Engineering. This crisis is an event-based crisis. The US economy was going very strong until February. The unemployment was at its lowest. The Global Economy, barring a few underperformers, was also doing okay. Nonetheless, the event we are facing is huge. World leaders are describing Coronavirus as a war-like situation and it is, except this time we are saving humans!

The 2008 crisis took the indices from their peak to bottom in about 15 months. It took another 15 months to regain the lost glory. Depending on how long it takes to defeat the virus, it will decide how long the recovery takes. It is unreasonable to expect a sudden recovery. However, it is important to note that US President Trump is up for reelection in Nov 2020 and he loves the stock market. He judges his own performance in the presidency by how good the stock market is doing. It is safe to assume that he will leave no stone unturned for the recovery of markets before the election. The total bailout packages and quantitative easing for the 2008 crisis were 800B to 1Trillion US dollars VS the 1.5–2 Trillion dollars of the package already in works. Steps like sending 1000$ checks to US citizens are in the plan something that was not done in 2008.

Why the interest rate cut and liquidity injection is not helping calm the markets?

The problem is not of a liquidity crisis, in fact, the heights that markets achieved is because of excess liquidity in the markets. The problem is of lack of confidence. Lack of confidence in the government’s ability to tackle the crisis and again the fear of unknown unknowns. No matter how much money you hand-over to people they are not going to fly, or go to a restaurant, in fact, the government wants people to stay indoors. The only way, at least for now, to kill the virus is to kill the economy.

As of this writing, the Federal Reserve has taken interest rates to 0% and promised to inject 1.5 Trillion Dollars into the system. Bailout packages are being prepared. These bailout packages are going to be non-partisan, a few years later there won't be any fingerpointing for these packages. Yet there are no signs of ease in the markets.

All these measures are not helping now but they will certainly help once investors see the end of the tunnel. It is very important to understand that financial markets are not the real economy. While the real economy can and will take a couple of years to recover, the markets can act faster like they are right now. All these quantitative easings will help markets recover later.

Why Gold, yields, and stocks all fell at the same time?

History has told us that people sell stocks, leave equity markets, and rush to safe heavens (Bonds, US Treasuries, and Gold). Surprisingly even gold fell sharply why? One can only speculate. It seems that people sold gold to raise cash. Cash is king during the time of crisis. The Greenback, US dollar, surged to a near all-time high vs all other global currencies. For the past 2–3 years we already have negative interest rates in Europe and it seems the US is also entering the same territory. This makes it harder to predict how the bond markets and the economy as a whole will respond with zero or negative interest rates.

Some numbers and random facts

1. The measures taken to stop the virus from spreading is bringing the economy to a standstill. Assuming a complete loss of 2 months' economic output the markets seem to already have taken that into account. Regardless there will some undershoot.
2. The US stock market provided gave a handsome 33 percent return in 2019 vs the Indian stock markets which gave a return of only 12 percent in 2019. The prices of Indian companies are already attractive and hence I believe the recovery in the Indian stock market will be as fast if not faster than the US stock market. The PE ratios of Indian blue chips are much much lower than the US counterpart.
3. The market cap to GPD ratio of India is currently at 47%, the historical mean has been 76%. This ratio reached a peak of 158% in 2007.
4. Bailout packages are being prepared with the airline industry first in the line, followed by hospitality.
5. Asian Central banks, including RBI, will defend their currencies.
6. AAA-rated companies have already started debt offerings.
7. ECB announces a 750B Euro coronavirus stimulus. Below is a summary of all fiscal and monetary policies that governments around the world have taken to fast track economic recovery once the virus is defeated. This data is true as of 3/15/2020, more packages are announced after.

So what investment strategy to implement?

When you board the journey of the stock market, rarely you know at what point you have boarded, Sometimes near the peak and rarely at the bottom, and rarely people have a strategy at what point to get off. If you are one of those who have entered the stock market in the past 2–3 year you will have to wait another 4–5 years before enjoying the fruit of your investment. The historical fact is that the stock market does indeed provide great returns provided that you have the courage to ride through volatility and provided that you have the wits to buy at low prices. One of my friends reminded me why one should always keep 20-30% of savings in cash or debt market. It is exactly for these times.

The fundamental principle of making money in the stock market is to buy low and sell high but rarely people are able to follow these wise words. The FOMO (Fear Of Missing Out) brings people into the market and the market volatility breaks their nerves and kicks them out.

One of the most simple functions in Finance is the one most powerful, the average function.

The best strategy is to let the volatility return to normal and then start nibbling at the quality stocks. If you are already enrolled in the SIP (Systematic Investment Plan) you should continue it. Stopping SIP would be a horrendous crime. I say this for a simple reason, for a given investment amount when the prices are high you get fewer units and when the prices are low you get more units and this averaging is far more powerful than people can imagine.

Prof. Robert Shiller, the Nobel laureate, said in one of his classes that the probability of Dow Jones falling by 27% in a single day is less than a person winning lottery yet the Dow did fall. Black swan events happen every now and then. In the past year, the crude oil prices have both increased by more than 25% in a single day, when there was a drone attack on Saudi oil facility and decreased by more than 25% in a single day, when the OPEC+ Russia communication broke down. World markets have been extremely volatile in the recent past.

The world’s first stock market is the Amsterdam Stock Exchange, established in 1602 and the world’s first IPO (Initial Public Offering) was The Dutch East India company. India was invaded on public money! The world financial markets are running for more than 400 years and will continue to run. The economic boom and bust cycles are only natural and both the boom and bust happen for various reasons. Many people see the recession as an opportune time. The markets have provided returns and livelihood and they will continue to do so. Mohamed El-Erian, Allianz Chief Economic Advisor, recently said this on CNBC “I am not afraid of the destination, I am good with the final destination, I am afraid of the treacherous journey.” and he said, “ Market volatility brings the worst out of people”.

All this said I am optimistic. The best course of action is to follow government advisory, beat the coronavirus and get back running. Again this crisis is not due to systematic flaws but due to an event that will surely pass.

Lastly, I am not a journalist and I did not find these facts on my own. Like many, I read these things online and watched them on TV but I am a firm practitioner of fact-checking, I rely only on authentic and credible source and strive my best to provide the source of my information.

Stay calm, Stay indoor!

http://www.carewealth.in/

The unconscious lavish spending is that gym membership that you signed up for two years back, then acted smart by enrolling for the auto-payment plan because it gave the first month free and an instant 10$ discount. Two years are now passed, you don't even recall the enthusiastic yourself who signed up for working out. Little did you know about human psychology and that you are no different. The gymnasium is now 1344 $ richer, thanks to you. The example here was still an analogy for all different kind of subscription products or services that you may have signed up for. Be it any kind of subscription, there is a very high chance that you are underutilizing it. Subscription business is exploding in growth and it is coming at the expense of you unable to make your dreams come true. The clever marketing techniques of these subscription-based services or products would lure you into signing up. Don’t fall for the initial free trial and for “for the cost of a single coffee you can now get Blah Blah Blah ….” kind of marketing trick. Think of your budget, it all has to add up. It would not be fair to single out any single subscription service or product but the fact is most people are signed up for more than they can utilize. One needs to seriously introspect, time to time, what subscription or memberships are absolutely needed.

Don’t get me wrong, buying things for yourself and entertainment is good for physical and mental health but it has to adhere to the law of moderation. Your signing up for 10 different kinds of entertainment subscriptions does not grant you a single second more in a day than anyone else in this world. Time budgeting is equally important if not more. A recent Harvard study says that we will be happier if we spend money on services that buy us time, but by no means, the study indicates to wreak your financial budget.

You must be still wondering, how can little savings here and there can get you that dream car. The answer here is through investments and through the power of compounding. Millennials and generation Z who are making 70,000$ or higher have significant disposable income, and they are in the best position to take on calculated risky investments. A very small part of savings can definitely go towards “a diversified set of calculated risky investments”. 50–100$ every month which otherwise lost in useless subscriptions can be systematically invested. Over a period these investments will return enough to make your dreams come true. Anyone who invested 100$ every 4 weeks in Tesla stock over the last 7–8 years would get enough return to buy themselves a Tesla. You probably have read and seen many articles saying that had you invested 100$ in bitcoin in 2011, by the end of 2017 your investment would have transformed into a whopping 6 million USD and even after a crash in 2018 that investment would still return 2 million USD today, but such events are extremely rare and channels of investment into bitcoin were not available until after the fact. The key here is depending on one's age and income flow, a very small portion of savings should be invested in very diversified calculated-risk investment vehicles defined by financial advisors. For a risk-averse person let me introduce to you the eighth wonder of the world, compounding. Most people fail to realize the power of compounding, and they realize it when it is too late for them to reap the benefits of compounding. It would be best for millennials and generation Z to realize it now. Everyone knows the textbook definition of compounding which is earning interest over earned interest. However, it is very difficult to visualize the gains over a long period of time. A 1000$ invested every month over a period of 20 years with an annualized return of 12% would return 1 Million $, that would allow you to retire a decade earlier. Indian equity markets have consistently given over 15% return over the last 3 decades, so we are not even talking about risky investments. The key here is consistency and patience.

Even though this article is written for US residents, in general, it applies to people living in other countries. The NRIs (Non-Resident Indians) should look into ways to invest in the equity market back home or even enjoy higher interest rates offered by banks. Naturally, any NRI would be worried about the cost associated with transfer fees and depreciation of Rupee but over a long period of time, the interests and returns offered by Indian banks and equity market comfortably beat the sins of lying money idle in US bank at 0% interest rate. NRIs who are investing in US stocks should continue to do so but also look into other emerging markets which are likely to beat US markets and provisions are already in place to allow trading in the US dollar.

Millennials and generation Z should seriously look into their monthly subscription bills. The easier it has gotten to pay monthly bills the lesser consumer is aware of their monthly spending. The saving resulted from spending cut back can be invested in equity markets for short and long term gains. In long run, these small steps make a big difference.

NRIs willing to Invest in Indian equity markets should consult Care Wealth Management Services at +91–9825064032 for guidance on your investment strategy taking into account your age and income flow. Let Care Wealth create a diversified investment portfolio for you.

http://www.carewealth.in/

If you were not able to catch my drift in the title, please read the whole article.

Crude oil has been both the maker and destroyer of nations. Five of the top ten fortune 500 companies deal in crude oil/petroleum. Having abundant of it is a blessing, but that does not necessarily translate to the prosperity of a nation. It is a double-edged sword, just like plastics, we would have never made it this far without crude oil. In fact, most of the plastics are derived from crude oil. Over 6000 items are made from petroleum, a partial list of those items include vitamin capsules, cold cream, food preservatives, movie film, deodorant, credit cards, toilet seats. Reading this list, it is clear that a religion can dare to ask its devout followers to renounce beef, pork, leather products but never petroleum products.

Average OPEC crude oil price in 2018 is 70.28 USD per barrel (1 barrel = 42 US gallon = 158.99 liter), this is a sharp rise from 52.51 USD in 2017. This sharp rise in crude prices caused Indian Rupee to be the worst currency performer in emerging market, losing its value by whopping 17% from the start of year, reaching the low of 74.33 VS USD by Oct. Now let’s zoom into the month of November, it’s the Black Friday after Thanksgiving and there is a discount on the crude prices as well (WTI -7.7% 50.42$). Crude prices have been dropping since late October and are expected to continue its downward trend. With the fall in Crude prices, the Indian rupee is strengthening, getting back to 70.44 VS USD. This makes INR one of the best currencies to recover from oil sell-off. This shows a very high correlation between INR strength and Crude oil price and it is no secret. India import about 85% of its crude oil requirements. On average for last two decades, India pays an annual bill of roughly 100 Billion USD. Crude is a single component, frustrating Indian governments to achieve its fiscal deficit target. Also, a single most important factor that can derail its burgeoning Economy. Rising fuel costs lead to heavy inflation, which places tremendous pressure on the incumbent government.

So, what happens next …?

Oil prices are driven by supply-demand and Geopolitics. Just like there were speculations of Oil reaching to 100$ per barrel now, there is a growing possibility that oil may return to its previous lows. Maybe even as low as 30$ a barrel. While geopolitics played a major role in its rise to 70$ mark, actual supply-demand may bring it back to low 30–40$. The trade war between US-China is causing a demand slump from the world’s second largest economy and the largest economy may pump so much oil into the market in 2019 and 2020 that it is set to become the single largest producer of oil. Importantly both favor lower oil prices. In the past decade, the US has drilled thousands of new sites in the Permian Basin of West Texas, these sites are profitable even at 30$ per barrel. They will keep the supply high enough to counter OPEC cuts or any geopolitics.

Low oil prices good for India …? Not Good for a long run.

The polluted air of New Delhi and other like cities is the reason low oil prices are not ‘all’ good for India in long run. If oil remains low for long period of time, India will achieve its target of fiscal deficit, the fuel prices will remain low, people will keep on being traditional and the government nor private sector will be motivated to embrace clean mobility. To clarify my position on this, low oil prices are good for the short run, while India tries to lift millions out of poverty and increase its GDP per Capita. During this transition, it needs to prudently invest and promote clean energy projects and build massive public transport infrastructure running on clean energy. We do not want to be wanting oil in the 2040s and 2050s.

Since 1990s India has truly walked the path of capitalism, and If oil prices remain low, capitalism will not be able to clean air of India’s cities. India does not need to divert from capitalism, the government just need to intervene at right time to steer the country towards clean air. India became a leader in solar energy not because solar energy was the right thing to do, but because the subsidies and the tax cut offered by the government made solar energy much lucrative compared to thermal power plants. The capitalist approach is by no means perfect and it never considers indirect socioeconomic consequences. All around the world, the far right has miserably failed to consider the health care cost, be it the tobacco industry, the big pharma or energy companies. It is assumed that problems will come along business opportunities and the demand gaps will be eventually filled. Sometimes these problems may get solved eventually, but end up costing far more than justified.

As per world bank’s report, air pollution cost India 8.5% of GDP in 2013, this translates to 165 Billion USD and the cost is only increasing since then. The government does not produce its own numbers and these numbers can be debated but there is a consensus that air pollution has a detrimental effect on the health of individuals as well as the economy. No government wants to be seen to be hampering growth for environmental reasons, worst yet people are not blaming the government for air pollution and political parties are usually not punished in elections because of air pollution. The current methods of counting economic success do not take into account this health care cost and, without taking those into account, chances of achieving a true economic success are slim.

In conclusion, if oil prices go higher this will inflict temporary pain on the Indian economy but this will drive India to go toward clean renewable energy and clean mobility since the reward of clean unlimited energy will be obvious. If, however, oil prices fall back, the government needs to take this bumper savings and build a platform for electric cars, electric high-speed rails, clean electric cooking, work towards cleaning the air of its major cities as well as rural areas. The motivations for doing all of this will be indirect and hidden but the cost of not acting proactively will be extremely high.

About Author — I am a Power Electronics Engineer and I chose to specialize in Power Electronics because I believe in climate change and we need green power and greener planet. My everyday work indirectly helps in saving this planet. My political views are center-right as per the Indian Political spectra. I am a staunch practitioner of practicing before preaching. I would like myself to be seen as a Global citizen originating from India.

World population is expected to be 9.7 billion in 2050, India’s population at the same time is expected to be 1.7 billion. As consumerism is growing in India at double digits, India is set to become the third largest consumer market by 2025. Along with every consumer goods and products, the usage of plastics will increase. Clearly, between now and 2050 the way Indians deal with plastic will have a huge impact on the environment at a global level.

Thankfully scientists all over the world are coming with new plastic alternatives or new low-cost plastic recycling methods. Meet Prof. Rajagopalan Vasudevan, aka The Plastic Man of India, Dean of The Thiagarajar College of Engineering in Madurai, India. In the early 2000s, he invented a method which uses plastic waste such as polythene bags, soda bottles, chip bags, other common plastics to make roads. His invention replaces about 10–15% of bitumen used in making roads with plastics, which results in huge cost savings. Bitumen is a product derived from crude oil, and is a glue that holds the components of the road together; just like cement is a glue in concrete. He then went on to revolutionize pavements. His second invention coated the waste plastic over a hot stone. This stone, called Plastone, is regularly used in pathways, footpath and in some cases internal pathways of a community. He patented the technology and gifted it to the Government of India. The roads and pavements made with his technology are proven to be stronger than the traditional 100% bitumen road. This is a win-win scenario. In 2018, Prof Vasudevan was awarded the Padma Shree by the Hon’ble President of India, Ramnath Kovind.

Using Prof. Vasudevan’s method 1 km of road, with a width of 3.75 meters can make use of 1 million polythene bags, that results in saving of 1-ton bitumen and 1 block of plastone can make use of 300 polythene bags and about half a dozen polyethylene terephthalate bottles. Realizing the potential of these methods, Government of India, in November 2015 mandated that all the construction of new roads should utilize plastic waste as a mixer for the bitumen. The plastic waste must be sourced from a periphery of 50 km of the road construction site. Until 2014 approximately 5000 km of roads were constructed using plastic waste. The Modi government has dedicated about 11 billion USD for the construction of 83,677 km of new roads, most of which will utilize waste plastic from its surroundings. Nowhere in the world are plastic roads made at this scale. I don’t want to dwell further into the stats, but this highlights how innovation with good policy can solve the toughest problems. This prudent policy of PM Modi will lead India to great heights and at the same time help save the world.