Interpret a Merchant Cohort Retention Table and Sketch Its Curves by Hand

A commerce platform groups new merchants by the month they signed up for a paid plan. A slide shows the cohort retention table below. You are asked to interpret it and, with only pen and paper, to draw a chart from it. The report says one slide asked how to interpret a retention data table and another asked for a chart to be drawn by hand from given data. It gives neither the data nor the chart type, so this version combines the two on one illustrative table. Each cell is the percentage of a cohort's merchants that still had an active paid plan at the end of the given month after signup; month 1 is the first full calendar month after the signup month. The data run through the end of July, so newer cohorts have fewer months, and the table shows months 1 to 5 only. | Signup cohort | Merchants | Month 1 | Month 2 | Month 3 | Month 4 | Month 5 | | --- | --- | --- | --- | --- | --- | --- | | January | 1,000 | 80% | 71% | 60% | 57% | 55% | | February | 1,050 | 81% | 66% | 61% | 58% | 56% | | March | 1,100 | 74% | 65% | 60% | 57% | | | April | 1,150 | 82% | 72% | 67% | | | | May | 2,400 | 68% | 57% | | | | | June | 2,600 | 67% | | | | | ### Clarifying Questions - Does a merchant who cancels and later resubscribes count as retained again? - Is a merchant on a free trial or a paused plan counted as active? - Did anything change in acquisition, such as a new channel, promotion or trial offer, in the months covered? - Did anything change for existing merchants, such as pricing, billing or policy, during the months covered? ### Part 1 — Read the table Describe the main patterns in the table and what each one suggests. Cite the cells that support each point. ```hint Three directions Read along a row, down a column, and along the diagonals, whose cells share a calendar month; each direction answers a different question. ``` ```hint Size and rate together Put each cohort's size next to its month-1 retention. ``` #### What This Part Should Cover - The shape of a typical cohort's retention curve - Differences between cohorts, read together with cohort size - Any effect tied to a calendar month rather than to how long merchants have been on the platform ### Part 2 — Draw it by hand Choose the chart or charts that best show what you found, sketch them, and say what the axes and the lines or bars represent. ```hint Choose the x-axis Plotting against months since signup and plotting against calendar month put different cells next to each other; each makes a different pattern visible. ``` #### What This Part Should Cover - A chart choice justified by the pattern it is meant to show - Correct axes, scales and labels, including how incomplete cohorts appear - A sketch accurate enough that the main patterns are visible ### Part 3 — Investigate and recommend What would you investigate next, what data would you pull, and what would you tell the business? ```hint Mix or quality A falling overall rate can come from a change in who signs up or from a change in how the same kind of merchant fares. ``` #### What This Part Should Cover - Hypotheses for each pattern and the data that would test them - The difference between the retention rate and the number of retained merchants - Concrete actions or experiments that follow ### What a Strong Answer Covers - An accurate reading of rows, columns and diagonals, with numbers cited - Cohort effects separated from calendar effects - Awareness of mix shifts, cohort size and the limits of immature cohorts - A hand-drawn chart chosen to make the main findings obvious - Hypotheses tied to specific data, and a recommendation ### Follow-up Questions - A dashboard shows overall month-1 retention falling from about 80% for the early cohorts to about 67% for June. How would you tell an executive whether the product got worse? - How would you project 12-month retention for the May cohort from two months of data? - If "retained" meant "made at least one sale in the month" instead of "still on a paid plan", how would the curves and your reading change?