Sanat Jain

In your Analytics career you can choose between 1. working within a single industry or domain, or 2. jumping around between different industries There are advantages to both! Sticking to one industry: - You probably get more replies when applying - You hit the ground running faster in a new job - Your domain knowledge compounds over time Jumping around: - Wider pool of job openings available to you - Get exposed to a wider variety of tools & methods - On average, it helps you grow more as a generalist Based on your experience, what else would you add? And what do you prefer? #data #analytics #career

100

28 Comments

Here is a very interesting question asked in Probo for a senior data analyst position. We have a table which has 5 columns. a. User_id, poll_id, poll_option_id, amount, created_dt b. Users invest money on these different poll options - these entries are there in the table c. One of the options is the outcome of the event. Example : Event: How many matches will the Indian cricket team win in 2022? A) Less than 50 B) 50-60 C) 61-65 D) Greater than 65 India wins 63 matches in the end. Option C is the winner. Money invested in option A, B and D should be proportionately distributed amongst users who invested money in option C. For example - If the total money invested in option A, B and D is Rs. 1500 and there are 3 users who invested 500 in option C {250, 200 and 50} These users would receive (750, 600 and 150 respectively) - sum is 1500. Write down a query for the above settlement process. Check out my detailed explanation in the comments section. #sql #analytics

161

25 Comments

Analytics is simple in theory. Analytics in practice is another matter 🙃 Why is it simple? It's only 3 steps: 1. Understand business problems 2. Select the right technical solutions 3. Analyze, build and influence to drive outcomes What makes it so hard in practice? - Data itself is messy and something is always breaking - Implementing solutions is never straightforward - Some people are data-driven, others are not - Different people have different incentives - Stakeholders don't understand our work The list goes on...what else would you add? #data #analytics

102

23 Comments

Today's 2 Important Learnings: 💪 1) Let's say your data source is "Power BI Datasets Semantic Models", so in that case there are some calculations coming from backend which we can't see in Power BI. But yes there is way to see all calculations dax formula with "External Tool Tabular Editor". You can download all measures with this way and see calculations in excel ✅✅✅ 2) Your using one measure which was calculated at backend and is referenced by 8-10 measures, you found that, that particular measure where it is used, that page is taking more time to load. So the solution is clubbed that 8-10 measures in one measure with help of Variables and use that one measures only. It will help in 50 % of Performance Optimization ✅✅✅ One of the Most Important Learnings I have posted till date ❤️❤️❤️ #dataanalytics #powerbi #data #businessintelligence

31

4 Comments

The best analytics outcomes happen when we understand decision tradeoffs. It’s not enough to analyze data and tell our stakeholder what the best option is. We should also consider and communicate things like - what are the costs of A vs B? - how likely is A to succeed vs B? - how reversible is A as opposed to B? - what happens if we choose A instead of B? - is there an option C? How else do you like to frame a decision? #data #analytics

26

2 Comments

AVG(1) vs SUM(1) in tableau In Tableau, AVG(1) and SUM(1) are both calculations that involve a constant value (1). However, they serve different purposes in aggregating data: 1. AVG(1): - This calculation takes the average of the constant value 1 across the specified dimension or level of detail in your view. - The result is always 1, regardless of the number of records or rows. Example: AVG(1) = 1 2. SUM(1): - This calculation sums up the constant value 1 for each record within the specified dimension or level of detail. - The result is the total count of records within that particular dimension or context. Example: If you have 2323 records for Central Region in dataset the: SUM(1) = 2323 In summary, AVG(1) always results in 1, while SUM(1) gives you the count of records within the specified context. These are the hacks which helped you to create some custom charts, like if you want to create Progress Chart you can use AVG(1) along with the primary measure and then create the dual axis.

42

The ability to understand correlations & relationships between multiple factors is the most important skill to look for when hiring an analytics person. In your next round of hiring for an analyst, - Give them a problem statement - Ask them to build an exhaustive flow chart listing the hypothesis/ reasons behind the problem - Let them talk you through how each hypothesis/ reason is related to the problem statement This alone can help you easily differentiate great analysts from bad ones - because product analytics is less about data, building reports, knowing tools, etc., and more about making sense of data & drawing correlations. #anshdoesanalytics #analytics #hiring

8

#SASCuriosityCup is over and we have announced the winners. As ALL student teams have worked hard and did a really good job I think they deserve to be promoted - all of them, not only the winners.   Let's start with the runner-up in the category   Data Preparation Team: Byte Me If You Can - ISCTE Portugal Paper title: Necessitating an Understanding of Diverse Attitudes Towards the European Union   "This research explores the multi-faceted perceptions of the European Union (EU) among its citizens. By identifying positive and negative perception clusters, the study reveals the intricate relationships between demographic, political, and economic factors, and their collective impact on EU perception. The findings provide practical insights for policymakers and communicators to customise strategies aimed at improving EU perception and promoting a united European citizenship."   What a relevant topic in these days, and I learned a new term: "Eurosceptisim" Well done ! Raul Laureano Jos van der Velden Francisca Abecasis Francisca Passos ISCTE - Instituto Universitário de Lisboa

42

4 Comments

Visualizing IPL Insights with Tableau: I recently created a Tableau dashboard to analyze key statistics from the Indian Premier League (IPL). This interactive visualization covers various aspects such as: Player Performance: Correlation between balls faced and dismissals, with insights into key players like Virat Kohli and Chris Gayle. Team Success: A breakdown of the top teams, with Mumbai Indians leading with 109 wins. Batsmen Strike Rates: Highlighting players like B Stanlake and KMDN Kulasekara with the highest strike rates. Popular IPL Venues: Wankhede Stadium in Mumbai tops the list for most matches hosted. This dashboard offers an intuitive way to understand player performance, team dominance, and match venues in the IPL. Check out the project on my GitHub: [https://lnkd.in/gYHTFBYP]

1

What are the most common calculations you will do in Tableau 90% of the time? 1. IF-THEN-ELSE Statements Yes! Most of the time you will be writing these if-else statements rather than any complex calculation 2. Simple Calculations like addition, subtraction, and division are the most common. Make sure you can translate formulas into calculations in Tableau and make sure you understand the difference between row-level and aggregated calculations. 3. Date Calculations Master them before it's too late. No report or dashboard is useful if you can't slice or dice your data concerning dates. 4. Boolean Calculations These are my favorites and you can write any logical statements to return TRUE or FALSE and can be used for coloring, filtering, and creating categories based on your logic. Do not miss out on learning them as they are performance efficient too. That's it! LODs or Table calculations aren't always necessary and are usually needed for specific needs only or should I say to go beyond. Don't let others scare you with complex terms! Master the simplest things and you will be far ahead of the crowd.

88

6 Comments

Possibly One of the Most Underrated Features of Tableau Cloud: ↳Tableau Pulse It won't: - Provide the full interactive experience of dashboards - Replace custom reporting for complex analysis - Offer in-depth visual exploration for every dataset But it does: - Speed insights with rapid access to critical metrics - Empower users by focusing on key, actionable information - Boost decision-making with easily consumable data at a glance Tableau Pulse is perfect for those moments when fast, impactful insights are the priority.

4

Difference between Dimensions and Measures in Tableau

3

🤔 Have you ever wondered "What the heck is ATTR()?!" in Tableau?   It basically works like this:   IF ISNULL([field]) THEN   NULL ELSEIF MIN([field]) = MAX([field]) THEN   [field] ELSE   "*" END.   In other words, within all the variables defining a Mark within the Rows, Columns and Marks shelf (in other words, your SQL "GROUP BY"), this function tells you whether the Dimension in question is a unique ATTRibute of that Dimension grouping.   Get it?   If it has 1 value, you get that distinct value. It is an *attribute* of your Mark.   If it has multiple values, you get a *.   If all are null, you get null.   I created an example to clarify (link in comment).   An NBA team can have more than 1 All Star.   Example Viz: ✔First tab shows the number of teams (30) and players per team in 2019. ✔The second tab shows teams that have 1, 2, or 3 All Stars (not all teams). ✔The third tab (the post image) shows for which teams Player Name is, and is not, an ATTRibute. In other words, is a team had a single All Star, that name is listed. Teams with multiple All Stars have a * listed. The table also shows the MIN and MAX to show the names of available All Stars for comparison to the IF/THEN above in this post. The combination of these two fields lists all All Star players except for The Golden State Warriors, who have 3. Since there is no such thing as a "MID" text function, you lose sight of Klay Thompson unless you create some sort of crazy LOD calc like I did for "Remaining Player" ✔The last tab shows how you can use ATTR as a meaningful filter - in this case, for teams with >1 All Star.   HTH   🏗→🧠 Build to Learn!   💭🚶‍♀️🚶‍♂️ Follow for more.   #tableau #data #analytics #VizoftheRay

20

4 Comments

Tableau is known for its efficient data processing and calculation speed compared to many other data visualization and analytics tools. There are several reasons why Tableau can calculate faster: In-Memory Data Engine: Tableau uses an in-memory data engine called Hyper (formerly known as VizQL), which loads and processes data into memory for rapid querying and visualization. This allows for quick data retrieval and manipulation without constant trips to the database. Data Compression and Optimization: Hyper employs advanced data compression techniques that reduce the amount of disk space required and speed up data access. This optimization ensures that calculations and aggregations are performed swiftly. Parallel Processing: Tableau's architecture supports parallel processing, allowing it to execute multiple queries or calculations simultaneously. This capability enhances performance, especially when dealing with large datasets or complex calculations. Query Optimization: Tableau automatically optimizes queries to minimize the amount of data transferred between the data source and the visualization engine. It intelligently pushes queries to databases whenever possible, leveraging database optimizations for faster data retrieval. Incremental Data Updates: Tableau supports incremental data refreshes, where only new or updated data is retrieved from the source. This reduces the amount of data processing required and speeds up dashboard refresh times. Hardware Utilization: Tableau can leverage multi-core processors and high-memory configurations effectively, utilizing available hardware resources to maximize calculation speed and performance. Data Extracts: Tableau allows for creating data extracts (extracting data from the original source and storing it locally). Extracts are highly optimized for querying and can significantly improve performance, especially for large datasets. Continuous Improvements: Tableau regularly updates its software with performance enhancements and optimizations based on user feedback and technological advancements. New releases often include improvements in calculation speed and efficiency. Overall, Tableau's focus on efficient data handling, in-memory processing, query optimization, and leveraging hardware capabilities contributes to its reputation for fast calculation speeds compared to many other analytics and visualization tools.

2

We often ignore STATISTICS, but it's concept clarity is very important for Data Analyst. So, here i m going to start series on stat topic and it's relevance in interview 📌 Scenario-based interview question focused on hypothesis testing: Scenario: You are a data analyst at a marketing company that recently launched a new advertising campaign. The marketing team claims that this new campaign will increase the average click-through rate (CTR) on ads from the current average of 2% to 3%. To verify this claim, you decide to conduct a hypothesis test. Steps:- 1. State the null and alternative hypotheses. 2. Explain the type of test you would use and why. 3. Describe the data you would need to collect. Assume you collected the following data from a sample of 1,000 ads: 35 ads had a click-through. 4. Describe the steps you would take to conduct the hypothesis test. 5. Explain how you would determine the significance level (alpha) for your test. Calculate the test statistic and p-value. Compare the p-value to the significance level to decide whether to reject or fail to reject the null hypothesis. 6. Based on your calculations, what conclusion would you draw regarding the marketing team's claim? Approach:- -> State the null and alternative hypotheses. *Null Hypothesis (H0): The new advertising campaign does not increase the average CTR, so the average CTR is still 2%. *Alternative Hypothesis (H1): The new advertising campaign increases the average CTR to 3%. -> Explain the type of test you would use and why. A one-sample z-test for proportions would be appropriate because we are testing a single population proportion (CTR) against a known value (2%). -> Describe the data you would need to collect. You need to collect the number of ads that were clicked (successes) and the total number of ads displayed (trials). ->Explain how you would determine the significance level (alpha) for your test. The significance level (alpha) is typically chosen based on how strict we want to be about avoiding a Type I error (rejecting a true null hypothesis). Common choices are 0.05, 0.01, or 0.10. For this scenario, let’s choose α=0.05. ->Assume you collected the following data from a sample of 1,000 ads: 35 ads had a click-through. Calculate the test statistic and p-value. Standard error SE≈0.0044 Test statistic z≈3.41 Using standard normal distribution tables or software, the p-value corresponding to a z-score of 3.41 is very small, much less than 0.05. -> Based on your calculations, what conclusion would you draw regarding the marketing team's claim? Since the p-value is much smaller than the significance level of 0.05, we reject the null hypothesis. This provides strong evidence to support the marketing team's claim that the new advertising campaign has increased the average CTR to 3%. for more such content, plz like and share the post #interview #statistics #hypothesistesting #dataanalysis #dataanalyst #datascience #datascientist #Machinelearning #python

19

✅ Hot Take: "Self-Serve" Analytics is a major red flag to avoid in BI landscape To understand self-serve analytics, Let's understand who is "SELF" here and what are they trying to "SERVE". ➡️ Who is the "self" ? 90% of the time, this "self" will be the finance/marketing, and compliance teams because they are Excel wizards. 1️⃣ They would like to export the report's data to excel and build their own department-specific aggregations and measures on the fly if they can't find the data at a granularity which they need for analysis. 2️⃣ Sooner, these custom built excel sheets will be shared within the teams. 3️⃣ One year down the lane, There will be tons of these excels floating around with no automation, no governance, and no version control. 4️⃣ Plus, Everyone will be creating multiple reports (some that may already exist) that could lead to one person doing it one way and another person doing it another way. 5️⃣ This can lead to mismatching KPIs and when data doesn’t match, the executives tend to not trust you anymore. Losing their trust in your data is what will be your downfall 6️⃣ Plus, having "multiple" "local" "excel" sheets goes against first principle rule of analytics. You should have a single source of truth. And then next year, there will be a new BI initiative to automate this process. Back to square one. Even if we implement self-service analytics in our organisation, what do we want to serve? ➡️ What are these self-serving folks "serving" ? 1️⃣ You can only "serve" specific use-cases. 2️⃣ If the use case changes or is new, then it will not be self-serve anymore unless the "self" themselves is a data / analytics engineer. ➡️ If the executives are reluctant to implement "self service", you really need to know what your stakeholders look for in the data, make sure they already have established KPI's and success metrics, and then limit what they can actually get from the self service tool. You absolutely cannot give them everything under the sun, you will exhaust yourself trying. Giving users everything is similar to giving the users "admin" access to your data warehouse To be honest, If I can get the clients to actually login and view the dashboards we've prepared for them, that would be a win. Getting them to explore the data on their own is a whole different animal. 📚 P.S. Want to learn data ops? I will be launching a new course next month, here is the link for early notification: https://lnkd.in/grXyHDQS #analytics #selfserve

34

3 Comments

Here’s how to measure bounce rate using Mixpanel. Start by defining the bounce rate for your purpose. Your criteria could be based on: Time spent: A user exits after spending less than a predetermined amount of seconds. Action count: A user performs fewer than a certain number of actions after landing. Head over to Funnels & select your starting point → Event that signifies a user’s arrival on your site, such as a landing page visit. Make sure to apply the “First Time Filter” to this event, to ensure you’re only measuring bounce rate for new users, & not repeat ones. You can find this filter by clicking on the 3 dots next to the event name. Select Mixpanel’s default event “Session End” as your second event. Apply filters for Session Duration (Seconds)" or "Session Event Count" on this event to align with your bounce rate definition. Set the conversion criteria to occur within a single session to reflect true bounce behavior. That’s it. You’re good to go! Look at the attached pic to see how I’ve calculated the bounce rate for users that leave the landing page within 30 seconds, with less than 2 actions. #anshdoesanalytics #analytics #insights #earlystagestartups #mixpanel #bouncerate

6

Scatter plots, Trendline, Correlation, RSquare, Linear Regression, non-linear regression, Decision trees are just some of the statistical techniques powering Market Mix Modeling. Impact of input/independent variables (Marketing tactics/promotion/price etc) on output/dependent variable (Sales) studied using Market Mix Modeling (MMM) to identify optimal level of advertising spend or price change impact on sales volume proves to be a powerful tool to enable Data-driven decision-making for all businesses alike (Ecom/offline/Omni). Do you incorporate MMM tool to power your biz insights?✨ Share in comments 👍 #MarketMixModeling #AdvancedMarketingAnalytics #DiminishingReturnAnalysis #LinearRegression #DataScience #dataanalytics Scaler Cassandra

9

2 Comments

𝐏𝐨𝐰𝐞𝐫 𝐁𝐈 𝐢𝐬 𝐞𝐚𝐬𝐲 IF: - Use data from only CSV or Excel - Use one or two data - have only One to Many relations - No or less cleaning needed - Simple DAX is in used - Filters are simple It is hard IF: - Use data from different data sources - Use multiple data - have multiple relationships with many to many - Data is uncleaned - Complex DAX is needed - Filters are scenario wise Next time choose your Power BI wisely! Shubhrajit Basu WhatsApp channel -> https://lnkd.in/gJ2cBp_X #PowerBI

27

4 Comments