<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by javier godoy on Medium]]></title>
        <description><![CDATA[Stories by javier godoy on Medium]]></description>
        <link>https://medium.com/@javiergodoy?source=rss-4c79ef981eae------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*MzEtcqPkhg3QmaoTg7mG0g.jpeg</url>
            <title>Stories by javier godoy on Medium</title>
            <link>https://medium.com/@javiergodoy?source=rss-4c79ef981eae------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 14:41:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@javiergodoy/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Never start with the data]]></title>
            <link>https://javiergodoy.medium.com/never-start-with-the-data-aae15e86f4b1?source=rss-4c79ef981eae------2</link>
            <guid isPermaLink="false">https://medium.com/p/aae15e86f4b1</guid>
            <category><![CDATA[business-intelligence]]></category>
            <category><![CDATA[okr]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[excel]]></category>
            <dc:creator><![CDATA[javier godoy]]></dc:creator>
            <pubDate>Sun, 30 Apr 2023 15:20:56 GMT</pubDate>
            <atom:updated>2023-04-30T15:20:56.437Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gwV4eYIRceb8f3Rt23HCGQ.png" /><figcaption>Photo by Tobias Fischer on Unsplash</figcaption></figure><p>Understanding the business context before dealing with data is crucial for a data analyst for several reasons:</p><p><strong>1. Relevant insights:</strong> With a clear understanding of the business context, a data analyst can extract meaningful and relevant insights from the data. This ensures that the analysis aligns with the organization&#39;s goals and objectives, ultimately helping decision-makers make informed choices.</p><p><strong>2. Data prioritization:</strong> Data analysts often work with large datasets containing multiple variables. Knowing the business context helps analysts prioritize which data to focus on, ensuring that the most valuable and relevant information is analyzed first.</p><p><strong>3. Effective communication:</strong> Understanding the business context enables data analysts to communicate their findings in a language that stakeholders can easily understand. This helps bridge the gap between technical and non-technical teams, facilitating collaboration and informed decision-making.</p><p><strong>4. Identify data quality issues:</strong> Familiarity with the business context allows analysts to spot inconsistencies, errors, or missing data that could affect the analysis&#39; reliability. Early identification of data quality issues can prevent wasted time and resources on faulty analyses.</p><p><strong>5. Tailored solutions:</strong> Different industries and businesses have unique challenges and requirements. By understanding the context, a data analyst can provide solutions that are tailored to the organization&#39;s specific needs, ensuring a higher level of effectiveness and practicality.</p><p><strong>6. Regulatory compliance:</strong> Businesses often operate within regulatory constraints that dictate how data must be stored, processed, and reported. Understanding the business context enables data analysts to ensure their work complies with these regulations, minimizing the risk of fines or legal issues.</p><p><strong>7. Foster trust:</strong> Demonstrating an understanding of the business context can help data analysts build trust with stakeholders. When analysts can connect their analysis to the organization&#39;s objectives, it is more likely that stakeholders will trust and act on their recommendations.</p><h3>Ask for the objectives</h3><p>Therfore, instead of diving into the data as soon as posible, it is crucial for you a an analyst to first understand the business objectives, as this knowledge serves as the foundation for all subsequent analytical efforts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OCHXRKZZP0XKgj8yzAHYsw.png" /><figcaption>Photo by Austin Distel on Unsplash</figcaption></figure><p>By focusing on the organization’s goals and KPIs, analysts can ensure that their work is aligned with the company’s interests and generates actionable insights. To achieve this, the analyst should ask a range of questions related to stakeholders, data sources and availability, data understanding, analytical approach, timeline and resources, and reporting and communication.</p><p>Addressing these questions enables the analyst to navigate potential challenges, prioritize tasks, and communicate effectively with both technical and non-technical stakeholders.</p><h3>Objectives and Key Results</h3><p>Various goal-setting frameworks are utilized by businesses to align their objectives, measure progress, and drive success. Some of these frameworks include SMART goals, Balanced Scorecard, Management by Objectives (MBO), and Objectives and Key Results (OKRs). Among these, OKRs have gained significant relevance in recent years due to their simplicity, flexibility, and effectiveness in promoting transparency and accountability.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OCZAshlDbPmYD0eNId2lGA.png" /><figcaption>Photo by Aline de Nadai on Unsplash</figcaption></figure><p>OKRs, or Objectives and Key Results, is a goal-setting framework that encourages organizations to set ambitious objectives (the &quot;O&quot;) and define measurable outcomes (the &quot;KR&quot;) to track progress. This methodology was popularized by companies like Google, LinkedIn, and Twitter, and has since been adopted by organizations of varying sizes and industries. OKRs promote alignment between individual, team, and organizational goals, fostering a culture of collaboration and continuous improvement.</p><p>This is key as many companies struggle with setting effective objectives because they focus solely on defining the goal without outlining a clear path to achieve it. This can lead to confusion, misalignment, and a lack of progress towards the intended outcome. Let’s review some of these unintended consequences:</p><p><strong>1. Misaligned priorities:</strong> Without clear objectives, team members may prioritize tasks and projects that do not align with the organization&#39;s overall goals, leading to wasted time and resources.</p><p><strong>2. Lack of focus:</strong> Vague objectives can result in employees working on multiple tasks without a clear understanding of what is most important, reducing overall productivity and effectiveness.</p><p><strong>3. Inefficient resource allocation:</strong> When objectives are unclear, it becomes difficult to allocate resources such as budget, personnel, and tools optimally, potentially leading to inefficiencies and increased costs.</p><p><strong>4. Low accountability</strong>: Vague objectives make it challenging to hold individuals or teams responsible for specific outcomes, resulting in reduced ownership and accountability for achieving goals.</p><p><strong>5. Poor performance measurement</strong>: The absence of specific, measurable targets makes it difficult to assess progress and determine whether the objectives have been met or not.</p><p><strong>6. Reduced employee engagement:</strong> When employees don&#39;t understand the organization&#39;s objectives or how their work contributes to achieving them, it can lead to decreased motivation, commitment, and job satisfaction.</p><p><strong>7. Ineffective decision-making:</strong> Without clear objectives, decision-makers may struggle to prioritize initiatives and make informed choices, potentially hindering the organization&#39;s overall performance and growth.</p><p><strong>8. Increased risk of failure:</strong> Vague objectives increase the likelihood of miscommunication, misaligned priorities, and poor decision-making, all of which can contribute to project failures and unmet goals.</p><p>To mitigate these consequences, organizations should focus on setting clear, specific, measurable, and actionable objectives that align with their overarching goals and can be effectively communicated to all team members.</p><h3>Objectives, models and Excel</h3><p>Modeling in Excel is a recommended solution to confront the consequences of vague objectives, as it offers a structured framework for organizing, analyzing, and presenting data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9QFVh1yp0Cucs6ssBCHwHQ.png" /><figcaption>Photo by Wes Hicks on Unsplash</figcaption></figure><p>By utilizing Excel’s versatile features and functions, organizations can create various types of models, such as financial, statistical, operational, project management, and decision support models. These models help clarify goals, align priorities, optimize resource allocation, enhance accountability, and facilitate performance measurement. Moreover, Excel models support informed decision-making and forecasting, ultimately reducing the risk of project failures and unmet goals. By leveraging the power of Excel modeling, organizations can transform vague objectives into actionable plans, driving better business outcomes and overall success.</p><p>But before building an Excel model, it’s crucial to &quot;speak the language of business&quot; That is: understanding the organization’s goals, industry-specific terminology, business processes, stakeholders&#39; needs, and relevant metrics, ensuring the model effectively aligns with the business context and drives meaningful insights.</p><h3>Step 1 Speak the language of business</h3><p>How could you incorporate the idea of starting from a business perspective when creating an Excel model?</p><p>Let’s use a simple example: a retailer with the overall objective of increasing the efficiency of its advertising spend. Below are my step-by-step recommendations to follow when starting your modeling process.</p><p><strong>1. Identify the desired business outcomes:</strong> Focus on delivering results such as optimizing return on ad spend (ROAS), increasing sales, and improving customer acquisition and retention.</p><p><strong>2. Use domain-specific terminology:</strong> Label sheets, tables, and charts with retail and advertising terms, such as &quot;ROAS,&quot; &quot;Customer Lifetime Value,&quot; &quot;Cost per Acquisition,&quot; and &quot;Conversion Rate.&quot;</p><p><strong>3. Align with key performance indicators (KPIs):</strong> Incorporate KPIs relevant to advertising efficiency, such as ROAS, Click-Through Rate (CTR), Cost per Click (CPC), and Conversion Rate.</p><p><strong>4. Involve stakeholders in the process:</strong> Engage with marketing and finance teams to gather their input, feedback, and validation to ensure the Excel model meets their needs and expectations.</p><p><strong>5. Customize the model for the target audience:</strong> Create separate views or dashboards tailored to the marketing team, finance team, and management, focusing on the metrics and insights most relevant to each group.</p><p><strong>6. Integrate with existing business processes:</strong> Design the Excel model to work with existing marketing and financial systems, enabling seamless data import and export for analysis and reporting.</p><p><strong>7. Leverage industry-specific best practices:</strong> Incorporate proven methodologies and benchmarks for retail advertising efficiency to enhance the model&#39;s credibility and effectiveness.</p><p><strong>8. Create clear, concise documentation:</strong> Provide documentation that explains the model&#39;s purpose, structure, assumptions, data sources, and instructions, using retail and advertising terminology easily understood by the business team.</p><p><strong>9. Validate the model with real-world data:</strong> Test the Excel model using actual ad spend and sales data to ensure its accuracy, reliability, and relevance in optimizing the store&#39;s advertising efficiency.</p><p><strong>10. Continuously refine the model based on feedback:</strong> Collaborate with stakeholders to identify areas for improvement and regularly update the model to reflect evolving business requirements, market trends, and advertising strategies.</p><h3>Step 2. Translate the business problem into Excel functionalities</h3><p>The What-If Analysis tools in Excel, including Data Tables, Goal Seek, and Scenario Manager, can help the retail store in our example by allowing the analyst to explore various advertising scenarios and strategies to optimize ad spend efficiency.</p><p>By using the What-If Analysis tools, the analyst can:</p><p><strong>1. Test different advertising budgets, channels, and targeting strategies</strong> to see their impact on key performance indicators (KPIs) like ROAS, CTR, and Conversion Rate.<br><strong>2. Determine the optimal ad spend allocation across channels</strong> to maximize sales or other desired outcomes.<br><strong>3. Identify the advertising variables that have the most significant impact on efficiency</strong> and prioritize efforts accordingly.<br><strong>4. Assess the potential risks</strong> and rewards associated with different advertising strategies and budgets.</p><h3>Step 3 Transform the insights into OKRs</h3><p>Based on the insights gathered from the What-If Analysis, the analyst can then formulate effective OKRs (Objectives and Key Results) for the retail store&#39;s advertising efforts. For example:</p><p>Objective: Optimize ad spend efficiency to increase sales.</p><p>Key Results:<br>1. Achieve a ROAS of X% across all advertising channels by the end of the quarter.<br>2. Increase Conversion Rate by Y% by optimizing ad targeting and messaging.<br>3. Maintain a Cost per Acquisition below $Z for new customers acquired through advertising.</p><p>By aligning the OKRs with the insights generated from the What-If Analysis, the retail store can create a focused and data-driven approach to optimizing their advertising efforts, ultimately driving better business outcomes.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aae15e86f4b1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Creating Effective Roadmaps for Data Products in a Large Organization]]></title>
            <link>https://javiergodoy.medium.com/creating-effective-roadmaps-for-data-products-in-a-large-organization-b4a2513d0a72?source=rss-4c79ef981eae------2</link>
            <guid isPermaLink="false">https://medium.com/p/b4a2513d0a72</guid>
            <category><![CDATA[data]]></category>
            <category><![CDATA[roadmaps]]></category>
            <category><![CDATA[enterprise-technology]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[data-product-creation]]></category>
            <dc:creator><![CDATA[javier godoy]]></dc:creator>
            <pubDate>Mon, 30 Jan 2023 22:26:52 GMT</pubDate>
            <atom:updated>2023-01-31T21:44:36.463Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cx_0dXwwhTjXiIl9VfDgzg.png" /><figcaption>Photo by NisonCo PR and SEO on Unsplash</figcaption></figure><p>As the use of data becomes increasingly integral to decision-making within large organizations, the development and implementation of effective data products becomes crucial.</p><p>However, creating a roadmap for these products can be a complex and challenging process, particularly in a large organization where multiple teams are involved.</p><p>With data engineers, data scientists, stakeholders, and other groups all playing a role in the development and deployment of data products, it can be difficult to ensure that everyone is working towards the same goals and that the roadmap is comprehensive and achievable.</p><p>In this article, we will explore the key elements of creating an effective roadmap for data products in a large organization, and provide practical tips and guidance to help data product owners succeed in this challenging environment.</p><h3>Why would you need a roadmap</h3><p>The roadmap serves as a guide for the product team and stakeholders and helps to align efforts and resources towards a common goal. It also helps to anticipate and manage uncertainty by outlining a strategic vision and a series of milestones to achieve that vision.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QSL7C6fHwRrjKiuQS4ncKQ.png" /><figcaption>Photo by airfocus on Unsplash</figcaption></figure><h3>Data products in large organisations</h3><p>When it comes to data products in large organisations, the objective of creating a roadmap remains the same, but there are some unique challenges and considerations that must be taken into account:</p><p>These products often have to comply with strict data governance, security, and compliance regulations.</p><p>Because these products are intended for internal use, there is typically a complex ecosystem of stakeholders, departments, and systems that must be integrated and aligned.</p><p>Therefore, when creating a product roadmap for these products, it is essential to have a deep understanding of the enterprise&#39;s ecosystem and to align the roadmap with the overall goals and objectives of the enterprise.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7y_nKSsvazE7UyX9xAvslg.png" /><figcaption>Photo by Mark König on Unsplash</figcaption></figure><h3>Roadmaps to the rescue</h3><p>A product roadmap can help solve the challenges that large organisations present in several ways:</p><ul><li><strong>Alignment</strong>: A product roadmap helps align the efforts of the product team and stakeholders towards a common goal, which is essential when working within a complex enterprise ecosystem.</li><li><strong>Prioritization</strong>: A roadmap helps prioritize the development of features and functionality based on the needs of the enterprise and its stakeholders, which is critical when dealing with limited resources and competing demands.</li><li><strong>Anticipation</strong>: A roadmap helps anticipate and plan for the unique challenges and considerations associated with developing and launching products for internal use in big enterprises. It can be used to identify and plan for data governance, security, and compliance requirements upfront.</li><li><strong>Transparency</strong>: A roadmap provides a clear and actionable plan for the development and launch of a product, which helps increase transparency and buy-in from stakeholders.</li><li><strong>Flexibility</strong>: A roadmap is a living document that can be adapted and updated as the product and the enterprise evolves, which is essential when dealing with uncertainty and change.</li><li><strong>Communication</strong>: A roadmap can serve as a communication tool to keep stakeholders informed and updated on the progress of the product development.</li></ul><h3>How to create a roadmap for your data products</h3><p>When developing a data product roadmap for enterprise use, follow these steps to ensure all essential components are covered:</p><ul><li><strong>Identify the Problem:</strong> The first step in creating a data product roadmap is to clearly identify the problem that the data product is intended to solve. This problem should be specific and well-defined, and it should align with the overall goals and objectives of the enterprise.</li><li><strong>Define the Target Audience</strong>: Once the problem has been identified, it&#39;s important to define the target audience for the data product. This includes understanding the characteristics, needs, and pain points of the users and stakeholders who will be using the data product.</li><li><strong>Set Objectives and Goals:</strong> Once the problem and target audience have been defined, it&#39;s important to set specific and measurable objectives and goals for the data product. These objectives and goals should align with the overall goals and objectives of the enterprise and should be used to guide the development of the data product.</li><li><strong>Define Features and Capabilities:</strong> The next step is to define the features and capabilities that the data product will need to have in order to meet the objectives and goals. This includes understanding the data sources, data pipelines, and data architecture that will be required.</li><li><strong>Establish a Timeline:</strong> Once the problem, target audience, objectives and goals, and features and<em> c</em>apabilities have been defined, it&#39;s important to establish a timeline for the development and launch of the data product. This timeline should take into account the complexity and scope of the data product, and it should be realistic and achievable.</li><li><strong>Identify Risks and Mitigation Strategies:</strong> It&#39;s also important to identify potential risks and mitigation strategies for the data product development process. This includes considering the potential for delays or setbacks, as well as identifying ways to minimize the impact of these risks.</li></ul><p>In conclusion, creating effective roadmaps for data products in a large organization requires a thorough understanding of the data landscape, the goals and needs of stakeholders, and the technical capabilities and limitations of the data team. By using frameworks and tools that prioritize clear communication, alignment, and adaptability, data product owners can lead their teams towards successful outcomes and drive value for the organization.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b4a2513d0a72" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Do you want to pay less for your mobile plan? use it a lot (and ask some friends to do so)]]></title>
            <link>https://javiergodoy.medium.com/do-you-want-to-pay-less-for-your-mobile-plan-use-it-a-lot-and-ask-some-friend-to-do-so-66cad34a2897?source=rss-4c79ef981eae------2</link>
            <guid isPermaLink="false">https://medium.com/p/66cad34a2897</guid>
            <dc:creator><![CDATA[javier godoy]]></dc:creator>
            <pubDate>Sat, 02 Oct 2021 18:07:48 GMT</pubDate>
            <atom:updated>2021-10-18T21:50:48.693Z</atom:updated>
            <content:encoded><![CDATA[<p>Ok, so you are reading this post because you want to pay less for your mobile plan. Great! But first of all, let me explain that this is an article about how poor visualizations lead to poor decision making (or no decision at all) But… if you stay with me till the end, I promise you will learn something about paying less for your plan ( and making better decisions with visualizations, of course!)</p><p>Recently I discovered this visualization published in 2020 by <a href="https://www.visualcapitalist.com/cost-of-mobile-data-worldwide/">Visualcapitalist.com</a> , a website focused on using data visualizations and infographics to explain economic issues. Great idea, by the way.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*TRGrRgL_5uQTVp4F" /></figure><p>The problem was that I think this is a very poor application of visualization techniques. For me, it seems like an interesting theme, well researched and explained in the article but accompanied by a visualization not very useful. Something like: “eh guys, where is the viz we asked you for my article yesterday? We are publishing in an hour!”</p><p>Let’s see more closely to understand why I considered there must be a better use of this data, and how I spend some time trying to make it better.</p><h3><strong>Biases and limitations in the data:</strong></h3><p>The article ins featuring the research made by<a href="https://www.cable.co.uk/mobiles/worldwide-data-pricing/"> Cable.co.uk</a> so I have used the original research to evaluate the data at hand (Visualcapitalist.com does not provide too many details about this research)</p><h3><strong>Collection issues:</strong></h3><p>The data does not consider all the countries due to some technical issues explained by the researchers. Not too bad as it’s a small number and not very representative. (Please note that the link redirects to the 2021 results and not 2020 (the number of countries included this year is 230 and not 155 as mentioned by Visual Capitalist) I really appreciate the authors provide a detailed list in the data available for download.</p><h3><strong>Processing issues:</strong></h3><p>I haven’t found any problem in particular as the original authors take some time to document how the average prices have been calculated and reference the criteria followed to convert between local currencies and US dollars.</p><h3><strong>Insights issues:</strong></h3><p>Here is where I found it more problematic. There is no record in the data, nor a reference in the visualizations or tables used in the article, where I can find the facts that drive the author to sentence: “Researchers have identified several key elements that help explain the cost variation for mobile data between countries”</p><p>For me that’s the key of the article so why is there no data supporting this approach? And without data, the visualization becomes a beautiful collection of international flags ( most of them arranged in a way that makes it difficult to learn something)</p><h3><strong>What explains the cost variation for mobile data between countries?</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xMxh_dzWgbgAd8JZ" /><figcaption>Issue Tree for the new visualization</figcaption></figure><p>So, I decided to try a different approach. As learned in my Udacity Data Visualization Nanodegree I focused on defining a good problem statement using the issue tree technique (resulting in what you can see in the illustration)</p><p>Why on earth would you spend some time watching a data visualization if not because there is something in it for you? Knowing what countries have cheaper costs is interesting. Which ones are more expensive? (mmm, it depends. is it the country I’m living in now?) Knowing why some countries are cheaper than others.. that’s it!</p><p>Not knowing too much about mobile networks but with the clear way of thinking I have gatherer from muy teachers at Udacity I found three possible reasons some countries have cheaper/more expensive costs than others:</p><ol><li>Countries with better infrastructure can offer better prices to the customers than countries with obsolete networks more expensive to maintain.</li><li>Countries where more people use mobile services every day generate economies of scale so individual prices decrease.</li><li>If more companies compete for a certain number of customers prices would be cheaper as a result of companies trying to get a bigger share of the market.</li></ol><h3><strong>Why the original visualization could do it better:</strong></h3><p>I think there are at least three big issues in the original visualization</p><h3><strong>1. There is not enough data.</strong></h3><p>With just one metric (average price for 1GB) it is almost impossible to facilitate enough context so we can learn what is driving that price. Seems like the idea of the website was to easily reproduce the article published by the original source, just copying the insights in a visual format. (All the other facts like the situation of Canada or the influence of the number of networks are simple annotations not evidence in any data)</p><h3>2. The visualization is not adequate.</h3><p>With only one metric I would use a simple bar chart ordering the countries by price. A ranking would be easier to read and understand and if you want you can keep the flags attached to each bar. This way you create a hierarchy that is clearly communicated. Currently, it’s impossible to see all the countries and their values.</p><h3>3. The visualization is redundant</h3><p>By using bubbles and their sizes to communicate the prices the authors are repeating the same information twice, as the position in the Y-axis informs the same. As there is no other data available, the X-axis has no meaning except trying to put in as many countries as you can.</p><h3>How to make it better</h3><p>Here it comes: <a href="https://public.tableau.com/views/Competititionorinfrastructurewhatdrivescheapermobilecosts/Competititionorinfrastructurewhatdrivescheapermobilecosts">this is my proposal</a> using a “Contrasting Values” data story type. And to do it this way I tried to validate which of my hypotheses (derived from the previous issue tree) was true.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vcoFSFG0SX30JtkR" /></figure><p>But wait. With only one metric. How could you convert the visualization into a scatter plot? Well, I had to find more data but as Andy Kriebel and Eva Murray say in their book “Makeovermonday” :</p><blockquote>“Aside from the initial research, it can be helpful to seek out information from secondary sources, such as those used by the original authors or commentary included with the original analysis. ”</blockquote><p>I wanted to tell the story, mobile data prices can tell, not only ranking prices and countries. Finding more data wasn’t difficult.</p><p>Not too far I found “<a href="https://www.cable.co.uk/broadband/speed/worldwide-speed-league/">Worldwide broadband speed league 2021</a>” another research published by Cable.co.uk that was very easy to join with “internet network speeds derived from over 1.1 billion speed tests taken in the 12 months up to 30 June 2021 and spanning 224 countries”.</p><p>I was looking for some data about network infrastructure, its development, level of maturity, etc, but the average speed seems to be a good proxy to measure the situation in each country from the user perspective.</p><p>Then I found<a href="https://www.mcc-mnc.com/"> a big list of network providers</a> so I could have the reference of the number of competitors by using a simple count distinct.</p><h3><strong>Good infrastructure is important to reduce mobile data prices but not enough.</strong></h3><p>My initial hypothesis was that countries with a more developed infrastructure would present cheaper prices. Using the values and the indexes available in both datasets (Prices Rank for prices and Speed position for infrastructure) you can easily appreciate that most countries are in the middle but many others have low prices without high speed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1010/0*5Wi74TR_Va-gViDJ" /></figure><p>So the idea of bad infrastructure higher prices is true but only for some countries ( in particular it’s true for those with the highest prices)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/848/0*UDvdrtd0KhzFh2v3" /></figure><p>Adding the number of networks operating in the country only confirmed that low competition is a constant for these countries</p><p>But in other cases, the available bandwidth is not correlated with cheaper prices</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/922/0*vqVmAPkJzjnlNwVA" /></figure><p>These countries have modern infrastructures expensive to maintain but customers can afford to pay more. Why should companies offer cheaper prices?</p><p>Then we have geographies where networks are not so fast and modern, but the population is big and adoption is almost complete. These countries are in the middle but with big differences between moderated prices in Europe (having big differences between european countries, as usual) and the cheapest prices in Asia.</p><h3><strong>If adoption is near-ubiquitous prices are cheaper even with no high bandwidth</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/908/0*vZLkbXqJC4aV986t" /></figure><p>Competition is important as the market forces prices down (see the number of networks in India for example)</p><p>As the original article denoted (without data demonstrating it):</p><p>“Countries with little to no fixed-line broadband availability, therefore, rely heavily on mobile data provision. In these cases, mobile data is the primary means the population has of getting online, and adoption is often near-ubiquitous. With a saturated market and many competing providers, often accompanied by a low average wage, data pricing in such countries can be exceptionally cheap when compared globally.”</p><p>So my three hypotheses have something to do with explaining the cost variation for mobile data between countries but customer usage seems to be the key driver.</p><p>Nevertheless, take care of the way you analyze the data. I’m not an expert in mobile telecommunications so using raw data without a clear context could lead to bad decisions. Always ask the experts as the original authors that advised us in this regard when explaining highest prices: “Countries with poor infrastructure tend to use fewer data. With mobile plans that offer smaller data limits, the overall average cost per GB tends to skew higher.”</p><p>Next time you see data like this I hope you would think about it this way. You and of course, the friends you will ask to do so. :)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=66cad34a2897" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[If you want results, give me some actions]]></title>
            <link>https://medium.com/analytics-vidhya/if-you-want-results-give-me-some-actions-909a2fa41266?source=rss-4c79ef981eae------2</link>
            <guid isPermaLink="false">https://medium.com/p/909a2fa41266</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[growth]]></category>
            <category><![CDATA[agile]]></category>
            <dc:creator><![CDATA[javier godoy]]></dc:creator>
            <pubDate>Wed, 04 Dec 2019 22:52:54 GMT</pubDate>
            <atom:updated>2019-12-09T04:33:25.705Z</atom:updated>
            <content:encoded><![CDATA[<h4>How to reach your business goals by focusing on your customer behavior.</h4><p>Following <a href="https://medium.com/@javiergodoy/mapping-digital-behavior-to-business-value-through-analytics-67694e7d4901">my previous post</a>, now its time to go into details. My goal here is to exhibit the essential ingredients that you, the product owner, will use to take advantage of combining behavior and value.</p><p>My short recap of the initial post is that every company should combine it’s traditional sales data with customer behavioral data, and use the last as a leading indicator for business results.</p><h3>Input metrics vs. output metrics</h3><p>Input metrics represents actions, output metrics represent results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*Wtz2RWJziGq8S_Yp" /><figcaption>Source: Reforge.com</figcaption></figure><p>A general approach like this lets you assign the appropriate role to each piece of your data. Results could be anything coming from your customer actions, although not all results are desirable, and not all are revenue and profit.</p><h3>Formulate your vision: follow your north star</h3><p>The growth hacking community and many start-ups worldwide use a metric called ”North Star” as the most important KPI to evaluate initiatives and align team efforts.</p><blockquote><strong>The North Star Metric is the single metric that best captures the core value that your product delivers to customers</strong>. Optimizing your efforts to grow this metric is key to driving sustainable growth across your full customer base.</blockquote><p>The north star is a leading indicator of future success. Results are difficult to influence but easy to measure. They either exist or not. The actions someone has to take to produce those results are not always known, not so easy to measure, but once you measure it, they are easier to influence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZIZPXQmjaiWagRp17Y-ouw.png" /><figcaption>Adapted from “Using Data to Set Product Strategy” by Justin Bauer</figcaption></figure><p>The north star metric responds to the question: <em>What actions do you need your users to do that leads you to reach your business goals? </em>Or in other words: <em>what inputs are required to get to the desired output?</em></p><p>As an example, Spotify uses time spent listening to music as their north star.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*-akK5dtMBJNoVm7V" /><figcaption>Source: Reforge.com</figcaption></figure><p>They understand their users and knows why they come and use their products. They focus on time spent as a signal of a more engaged user that is more willing to pay for the service. In the end, why would you pay for something you don’t use heavily?</p><p>So they focus on time spent to increase revenue, but to increase the time spent, they concentrate on some input metrics like how often users back to the service or the average time spent per person.</p><p>And to increase those metrics, again, they look for other input metrics and how different initiatives have an impact on the next level.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*d6E4eIfB3zKkRhcr" /><figcaption>Source: Reforge.com</figcaption></figure><p>That’s how the Spotify team ends, creating features like playlists or recommendations. Because those features lead the user to come back more often or to increase the average time per person, that leads to more time spent listening to music that leads to more subscribers and then more revenue.</p><h3>Set strategy: find the mathematical equation for your business.</h3><p>Premium subscribers to Spotify consume an average of <strong>75 hours per month</strong>, while free users only 25. (Impressive if you consider that <strong>90% of Spotify users are premium</strong>)</p><p>The reason Spotify is getting outstanding results looking at time spent is that it reflects the company’s <a href="https://www.strategyzer.com/canvas/value-proposition-canvas">value proposition</a> is working for their customers, while aligned with the company’s vision (“<em>Give people access to all the music they want, all the time — in a completely legal and accessible way”</em> )</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5V3H0POUHnCRL2RLWb6D1w.png" /><figcaption>My Behaviour to Value Framework</figcaption></figure><p>Again, this is a demonstration of how <strong>to get business value; you must provide customer value first</strong>.</p><p>Examples like this are used by the growth hacking community to assert that <strong><em>“Any business can be explained using a mathematical equation.”</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*Vtifcgcg3sl58HFu.png" /><figcaption>Andy Johns’ formula for Amazon’s growth</figcaption></figure><p>Another good example is <strong>Netflix</strong>. When Netflix was still a rental DVD service, they measured its business outcome by the number of subscribers, so they put their focus on the monthly retention rate.</p><p>As they identified that people with at least 3 DVDs in the queue tent to stay more loyal to the company, they worked to increase the number of people with 3 DVDs <strong>from 60% to 90%</strong>. That’s how they improved the monthly retention rate <strong>from 92% to 98%</strong>.</p><h3>When you know what you don’t know: put your data to work</h3><p>To identify your north star metric, you must understand your customers’ behavior quite well. You need to learn how specific actions correlate to the outcomes your business demands to generate the desired business results.</p><p>Often, the reality is that your company is asking you and your team to achieve the desired goal without knowing the resources, activities, and outputs of those activities that better influence the customer behavior.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Kby15jaxoLYFkKZEqsxrEA.png" /><figcaption>Adapted from Josh Seiden</figcaption></figure><p>What differentiates companies like Spotify or Netflix is how often they test assumptions, try new things and, measure everything they do. It is not a culture of measurement but one of experimentation and continuous learning.</p><p>The first time you try it, you don`t know what the behavior you should focus on is. First, you make an assumption and test if you are wrong or right. But you know what success looks like, or at least your boss knows. So, look into your data and try to answer this:</p><blockquote>Compare your ‘successful’ users to your ‘unsuccessful’ users — what actions are the successful users taking?</blockquote><p>The data team at Amplitude <a href="https://amplitude.com/blog/2014/07/29/find-the-key-to-your-apps-growth-without-an-army-of-data-scientists">suggests</a> a simplified method to find that correlation. They call it “<em>behavioral cohorting</em>:”</p><blockquote>Test which of the user actions may actually be a good predictor of success. A great way to dig into this is by grouping users into cohorts based on their behavior and then testing how outcomes differ by cohort.</blockquote><p>Let’s say your goal is to increase customer retention and, you suspect that opening your email messages more than three times in a month and creating a buying list are potentially the best leading indicators.</p><p>For each one of these behaviors, you will create a cohort and compare each group using a <em>binary classification test</em> on <em>sample data</em>. For that, you will calculate the <em>positive and negative predictive value</em> as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*qxTDkhlZJ-fI_vXK" /><figcaption>Source: Amplitude blog</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WFwc9f8wGJxzgBfht1_xWg.png" /></figure><p>In this toy example, opening more than three emails has a substantial positive predictive value for retention. Opening fewer than three emails has a significant negative predictive value. It means that this is an excellent leading indicator.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zxmHT7cJqaVF3iqgsN90Xg.png" /></figure><p>This other test has a strong positive predictive value, but even those that have no buying list have a good predictive value related to retention. In other words, it is less clear that this behavior leads to customer retention.</p><h3>Every problem is a nested solution to a higher level problem</h3><p>What I’m encouraging you to do is not to follow a simple recipe. That’s not going to work, even if the skills are not so complicated.</p><p>Moving from measurement to test and experiment requires more than putting a data scientist in your team. It requires a whole new team or at least a team with a whole new mindset.</p><h4><strong>1. Start with mindset.</strong></h4><p>As <a href="https://blog.leanstack.com/@ashmaurya?source=post_page-----d52ffd573f0c----------------------">Ash Maurya</a> recommends, every transformation should start with a mindset shift. Think of this as a microscopic version of the idea of inputs and outputs, but applied to the team:</p><blockquote>While we rightly place a lot of emphasis on doing, <strong>thinking precedes doing</strong>. And <strong>the quality of your output</strong> (results) is directly dependent on <strong>the quality of your inputs </strong>(ideas).</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*nCX19A63AKivo1IV.png" /><figcaption>Source: <a href="https://blog.leanstack.com/@ashmaurya?source=post_page-----d52ffd573f0c----------------------">Ash Maurya</a></figcaption></figure><h4><strong>2. Start together</strong></h4><p>I love this analogy from John Cutler that uses The Goonies (sorry millennials) as a metaphor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*d_twE8IrC3uZtJWM.jpg" /><figcaption>The Goonies, 1985</figcaption></figure><p>Your team is going to places they have never been before. Sometimes where no one has been before. They should discover the problem together for the first time, as The Goonies opening the door of the hunted house and looking inside at the same time.</p><h4><strong>3. Compensate metrics</strong></h4><p>Take care of the side effects optimizing a metric may have as people start to cheat the system. If you, for example, optimize for an increase in the average value per customer, the easiest and fastest way to do that is removing low-value customers, so numbers get better.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/0*4dJQASZ-brrO4sjl.jpg" /><figcaption>Source: Dereck Sivers <a href="https://sivers.org/compensate">Overcompensate to compensate</a></figcaption></figure><p>Introduce restrictions in the formula by adding metrics that offset each other, like increasing revenue while keeping the NPS score, for example. This practice will prevent your team from moving radically in one direction without looking behind.</p><h4><strong>4. Facilitate Psychological Safety</strong></h4><p><a href="https://journals.sagepub.com/doi/abs/10.2307/2666999">Defined</a> as <em>“The belief that the team is safe for interpersonal risk-taking. That one will not be punished or humiliated for speaking up with ideas, questions, concerns, or mistakes.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*rwMpFjHNqyw2C7xV83okoQ.png" /><figcaption>Source: <a href="http://modernagile.org/">modern agile</a></figcaption></figure><p>Create an environment in which making mistakes is not considered as a failure, but rather as experimentation and learning opportunity.</p><h4><strong>5. Master model evaluation</strong></h4><p>You don’t have to be the mathematical magician behind all kinds of sophisticated models, but as a business leader, you must learn to evaluate the models created by your team.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/740/0*CEzSoHaU5HTiVKcC.jpg" /></figure><p>There are lots of resources available online, but I will recommend you to read one and only one book that will bring you the superpowers you need: <a href="https://www.amazon.es/Data-Science-Business-data-analytic-thinking/dp/1449361323">Data Science for Business </a>by Foster Provost.</p><p>Chapter 7 introduces how to evaluate a model, with references to true and false positives and the confusion matrix I used earlier.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=909a2fa41266" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/if-you-want-results-give-me-some-actions-909a2fa41266">If you want results, give me some actions</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mapping [digital] behavior to business value through analytics]]></title>
            <link>https://javiergodoy.medium.com/mapping-digital-behavior-to-business-value-through-analytics-67694e7d4901?source=rss-4c79ef981eae------2</link>
            <guid isPermaLink="false">https://medium.com/p/67694e7d4901</guid>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[agile]]></category>
            <category><![CDATA[analytics-platforms]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[innovation]]></category>
            <dc:creator><![CDATA[javier godoy]]></dc:creator>
            <pubDate>Sun, 24 Nov 2019 16:18:35 GMT</pubDate>
            <atom:updated>2021-03-18T10:08:41.419Z</atom:updated>
            <content:encoded><![CDATA[<p>As the product owner of an omnichannel analytics platform, I’m frequently asked: “<em>what value do you provide to our business with this new tool?</em>” My answer is usually a combination of reasons I’m putting here all together for the first time. For a follow up post, please <a href="https://medium.com/analytics-vidhya/if-you-want-results-give-me-some-actions-909a2fa41266">read this</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OTwR3FFfaMqMs-o_H5Sugg.png" /></figure><h3>What we know we don’t know</h3><p>For large enterprises with substantial legacy systems and data, the starting scenario usually describes two separated spaces.</p><p><strong>The perfectly-known:</strong> where you have all that you know about your customers after years of business operations, customer relationships, and industry practices.</p><p><strong>The imperfectly-known:</strong> where you have all your clients’ behavior while interacting with your owned channels, but not all the digital channels, and sometimes not linked to the same customer ID. (Think cookies or third-party data)</p><p>I’m using the word “imperfectly” here because even when companies deploy all that is needed to capture the data, some things remain unknown for the enterprise. Digital behavior is fuzzy, unpredictable, and rapidly changing. We can’t pretend to have the same level of perfection that we got with internal business data. Nevertheless, we can work to improve it continuously.</p><p>That is not a problem for smart organizations that assume perfection is the enemy of usefulness. And it is the starting point to those who understand that mapping one space to the other should be of great value, even if the result is not as perfect as the initially separated spaces.</p><h3>Lagging indicators vs. leading indicators</h3><p>On one side of the newly created space, there is <em>customer behavior</em>. On the other hand, we find <em>business results.</em></p><p>Our goal as analysts and especially as creators of tools for analysts is to help users understand the relationship between one and the other.</p><p>In doing that, we have to use the same conceptual tool: <em>metrics</em>. But we must understand the differences between those used in one space and the other.</p><p>Sales data, correctly recorded in our internal systems provide us <em>lagging indicators </em>like<em> </em>total numbers of sales, average revenue, etc</p><p>Digital data offer us <em>leading indicators </em>like<em> </em>number of visitors, number of applications, percentage of repeating users, etc</p><p>Lagging indicators are “output” oriented, easy to measure, but hard to improve or influence while leading indicators are typically input-oriented, hard to measure, and easy to shape.</p><p>Tradicional strategies focus on the outputs, establishing a goal for the metric, and asking the departments involved to define a plan that should reach that goal.</p><p>Without the perspective of customer behavior, historically, companies tried a lot of things to improve their metrics, not knowing with precision what initiatives changed the trend and improved results. (<em>“I know 50% of my advertising is useless but don’t know what 50%”</em>)</p><p>Now that customer behavior is fundamentally digital; our opportunity resides in the use of customer behavior as a predictor of business results.</p><h3>From customers outcomes to business outcomes</h3><p>The biggest problem in many organizations is how deeply they focus on <em>outputs</em> more than <em>outcomes</em> when measuring and evaluating initiatives.</p><p>As <a href="https://medium.com/u/7f786c230ca9">Josh Seiden</a> gratefully explains in his book <a href="https://www.senseandrespondpress.com/managing-outcomes"><em>Outcomes over outputs</em></a>,</p><blockquote>An outcome is a change in human behavior that drives business results. Outcomes have nothing to do with making stuff — though they sometimes are created by making the right stuff.</blockquote><p><em>Outputs</em> instead are <em>features</em>. Setting your targets as outputs it’s ok if you are <em>“confident that the thing you’re going to make will solve the problem. On the other hand, we often work on problems where the answer is not so clear.”</em></p><p>For those of you more familiar with the idea of outcomes as business results, here we differentiate between <em>customer outcomes</em> (behaviors) and <em>business outcomes</em> (results, revenue, etc.)</p><p>Following this idea, if we can create the right features that change or help to change the customer behavior, we can “predict” the results it will generate to the business.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*W0aJt_sUheniymWo.png" /></figure><p>In other words, if your product generates value to the customer (making the experience easy, convenient, fast, secure, relevant, and intelligent), this will correlate to more sales, more renewals, and more recommendations. And that’s value for the business.</p><h3>From measuring to test and experiment</h3><p>For the analyst, reporting the metrics that demonstrate the number of sales is critical when you are working only on the perfect-known space. Her work is no other than telling the manager what happened, and if we got the result, we had forecasted. (Plus finding good arguments if the result is not good)</p><p>If we play on the more complex but more promising space of behavior and results, the first thing to know is what we should build to change the customer behaviors. Or as Josh Seiden put it:</p><blockquote>What are the measurable changes in behavior that will create this result?</blockquote><p>Here the analyst can help by providing the manager a picture of the current customer actions, showing the data. Then she can filter the behavior of those customers already creating the result we are seeking.</p><p>Showing the correlation between behavior and results, managers don’t have to try ten different things hoping to get it right, at least in a couple of them. But remember, knowing what worked well in the past does not constitute a future success by default; it informs a hypothesis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XAuqUDAzZAJ4yT_X" /></figure><p>This new approach is a fundamental change in the mindset of managers, but for the analysts too. Even better, when learning about customer outcomes and testing new business ideas becomes so crucial to the company, should we left the task in the hands of only a few, or should we all think like a data-influenced, experiment-driven team?.</p><h3>Designing teams around data and experiments</h3><p>Data should be our best ally to better and faster formulate a hypothesis and of course, to better and more accurate validate our analysis.</p><p>Many experts in product management ask us to use hypothesizes as the basics of our product roadmap and so as the reason to produce every new feature. Again Josh gave us a simple template to use when generating a new hypothesis:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7VckZFbpp4ORVY6OBSzLxQ.png" /></figure><p>But to do that, implement it, test it and, learn from it we need a team. And teams have their behavior too.</p><p>David J. Bland, in his recently published book <a href="https://www.strategyzer.com/books/testing-business-ideas-david-j-bland">Testing business ideas, </a>maps the team behavior as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Pz6Gzo6xeJbq4ULHmKNgw.png" /></figure><h3>Designing tools for teams around data and experiments</h3><p>If our business colleagues should use data and experiments to identify “<em>what measurable behaviors will create business results</em>” while focused on customer outcomes and creating the solutions/features that better produce them. Shouldn’t we, the tool creators, help them achieve this goal, the better we can?</p><p>And in doing so, shouldn’t we develop the same mindset, and put our focus on how they (our internal customers) “attain this [<em>user outcome</em>] with [<em>this solution/feature</em>]?</p><p>As creators of analytics tools and platforms:</p><p><strong><em>We are responsible for creating features that better change our co-workers’ behavior.</em></strong></p><p><strong><em>We should build tools that make it easy for them to provide business value to the company.</em></strong></p><p><strong><em>We should focus on employee outcomes: data-based decisions, and an experiment-driven mindset, with an interactive approach — not</em> dashboards and alerts.</strong></p><blockquote>We are the ones that help to change the employee behaviour that help to change the customer behavior that generates business outcomes.</blockquote><p>Read my follow up post “<a href="https://medium.com/analytics-vidhya/if-you-want-results-give-me-some-actions-909a2fa41266">If you want results give me some actions</a>”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=67694e7d4901" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>