metrics & analytics cases

Metrics questions are the fastest way to separate PMs who think from PMs who memorize. In an interview, you get “revenue dropped 20% — diagnose it.” On the job, you get the same question except nobody tells you the percentage, nobody hands you a dashboard, and the CEO wants the answer by Friday.

This page is five worked cases. Each one is a real scenario — the kind you encounter at Indian startups where data infrastructure is patchy, the business model is still evolving, and the answer is never just “check the funnel.” Work through them sequentially. The diagnostic pattern will start repeating, and that repetition is the point.

The diagnostic framework you will use repeatedly

Before the cases, one framework. Every metrics problem follows the same skeleton:

1. Clarify the metric. What exactly is being measured? How is it calculated? Over what time window? A “20% drop in revenue” means nothing until you know whether it is daily, weekly, or monthly — and whether it is gross or net.

2. Establish the baseline. Is this drop unusual? What does the normal range look like? Seasonality, day-of-week effects, and holiday patterns kill more investigations than bugs do.

3. Segment before you hypothesize. Break the metric by platform (iOS vs Android vs web), geography, user cohort (new vs returning), acquisition channel, and product line. The aggregate number lies. The segments tell the truth.

4. Isolate internal vs external. Did you change something (deploy, campaign, pricing)? Or did the world change (competitor launch, regulation, seasonality)?

5. Find the layer. Metrics are stacked. Revenue = Users x Transactions per User x Average Order Value. A revenue drop lives in one of those layers. Find which one before you start fixing.

This is not a framework to memorize. It is a checklist to run. Every case below uses it.

Case 1: The revenue drop that was not a revenue problem

Setup: You are a PM at a B2C e-commerce company in Bangalore — think Meesho or early-stage Flipkart competitor. The Monday morning dashboard shows gross revenue dropped 22% week-over-week. The leadership Slack channel is on fire.

The decomposition:

Revenue = Traffic x Conversion Rate x Average Order Value (AOV)

Against the pre-sale baseline, every metric is either flat or slightly up. The “crisis” is a comparison error.

The lesson: Always ask “compared to what?” before investigating. Sale periods, festivals (Diwali, Dussehra), month-end salary cycles, and even IPL match schedules distort Indian e-commerce baselines. If you do not control for these, you will spend every Monday chasing ghosts.

India-specific context: Indian e-commerce revenue is intensely cyclical. Month-end (salary credit days, 25th-1st) sees 40-60% higher transaction volume than mid-month at many platforms. Diwali week can be 5-10x normal. If your analytics tool defaults to week-over-week comparison without seasonal adjustment, it is lying to you most of the time.

Case 2: Engagement is flat but the CEO wants it up

Setup: You are a PM at an edtech platform — think Unacademy or a Pragmatic Leaders-scale operation. Monthly Active Users (MAU) has been flat at 280K for four months. The board wants 15% quarter-on-quarter growth. The CEO asks you to “fix engagement.”

The problem: “engagement is flat” is not a diagnosis. It is a symptom. And MAU is one of the laziest metrics in product management.

The decomposition:

MAU (app open) = New users who opened + Returning users who opened

But meaningful engagement = Users who completed a learning action

The real insight: The platform had two completely different products hiding inside one metric. For 95K users, the app was a curiosity — opened once from an ad, never returned. For 20K users, it was a daily habit. Building features for “280K MAU” meant building for nobody specific.

The lesson: When someone says “engagement is flat,” your first job is to challenge the metric definition. MAU with a low activation bar is a vanity metric. Redefine “active” around the core value action — the thing users came for. In edtech, that is learning. In e-commerce, that is purchasing. In payments, that is transacting. The number will be smaller and more honest, and the growth trajectory will become visible.

India-specific context: Indian edtech has a massive top-of-funnel problem. Millions download apps from Google Play ads, use the free tier for a day, and vanish. If your MAU counts these users, your metric is a marketing dashboard, not a product dashboard. Pragmatic Leaders learned this early — the metric that mattered was not “how many people signed up” but “how many people completed a module.” That number was always smaller and always more useful.

Case 3: Picking the right metric for a new feature

Setup: You are a PM at a fintech company building a new “savings goal” feature inside an existing UPI payments app — picture a PhonePe or Google Pay type product. Users can set a savings target (like “Goa trip — Rs 50,000”) and auto-transfer small amounts weekly from their UPI-linked account.

The feature is built. The Head of Product asks: “What is the success metric?”

This is a trap question. Not because it is hard, but because the obvious answer is wrong.

The metric selection framework for new features:

The lesson: For any new feature, you need a leading indicator (did they try it), a health indicator (are they using it meaningfully), and a lagging indicator (did it create the outcome). Do not pick just one. And never let the leading indicator become the team OKR — it will be gamed within a quarter.

India-specific context: Savings features in Indian fintech face a unique challenge — the Rs 500 weekly auto-debit that feels trivial to a Bangalore tech worker is a meaningful commitment for a teacher in Patna earning Rs 30,000/month. If you track “average deposit amount” as a health metric without segmenting by income tier, you will conclude the feature is failing when it is actually serving different users differently. Segment by city tier. Always.

Case 4: The conversion rate that lied

Setup: You are a PM at an online insurance aggregator — think PolicyBazaar. Your term insurance comparison page has a 4.2% click-to-quote conversion rate. Product leadership is happy. Then you segment the data.

The decomposition:

Overall conversion (click to purchase) = 4.2% x 0.8% = 0.034% (3.4 per 10,000 visitors)

But the funnel has five stages, and the bottleneck was stage 4 — not the top of funnel that leadership was celebrating.

The lesson: A high top-of-funnel conversion rate can mask a catastrophic mid-funnel drop. Never celebrate a partial metric. Always measure the full journey from first touch to completed action. When you find the bottleneck, look at the user experience at that exact step — not the data, the experience. Go through the flow yourself on a phone. Time yourself. The data told the team “medical questions page is the problem.” But only using the product on a phone revealed why — 12 fields, tiny touch targets, no progress indicator, and a question about “family history of renal disorders” that most users cannot answer confidently.

India-specific context: Insurance purchase on mobile in India has a unique trust problem. Users filling out medical details on a phone worry about two things — data privacy (“will my health data be sold?”) and disclosure consequences (“if I say my father had diabetes, will they reject me?”). The form was not just long — it was anxiety-inducing. The redesign added a single line: “Your medical information is used only for this quote and is not shared with third parties.” That line improved form completion by 11% on its own.

Case 5: The daily active user trap at a social commerce startup

Setup: You are a PM at an early-stage social commerce company — think early Meesho. The product lets resellers in tier-2 and tier-3 cities share product catalogues on WhatsApp and earn commissions on sales. DAU is 45K and growing 10% month-over-month. The board is pleased.

Then a data analyst pulls a cohort retention chart.

DAU is growing only because acquisition is outpacing churn. The company is on a treadmill — running faster to stay in the same place. At current acquisition cost (Rs 120/user) and month-6 retention (5-6%), the unit economics are negative.

The lesson: DAU and MAU are aggregate numbers. They add new users and returning users into one figure and call it “active.” But for any product where retention matters (which is every product), you must look at cohort retention — what percentage of users who signed up in month X are still active in month X+3, X+6, X+12? If the cohort curves all collapse to near-zero, your growth is an illusion sustained by marketing spend. The moment you stop spending, the number collapses.

India-specific context: Social commerce in India (Meesho, DealShare, CityMall) saw this pattern play out across the entire sector. The reseller model attracted millions of sign-ups from tier-2 and tier-3 cities — homemakers, students, small shopkeepers looking for supplemental income. But the gap between “signed up” and “earned money” was enormous. The successful companies were the ones that optimised for time-to-first-sale, not sign-ups. Meesho eventually cracked this by eliminating the reseller margin concept entirely and becoming a direct e-commerce platform — a pivot driven entirely by retention data showing that the reseller model churned 95% of users.

The patterns across all five cases

Every metrics problem you will encounter is one of these five:

1. The baseline problem (Case 1). You are comparing against the wrong reference point. Fix the comparison before you investigate the metric.

2. The definition problem (Case 2). The metric is measuring the wrong thing, or measuring the right thing too loosely. Tighten the definition and the “problem” either disappears or becomes clearer.

3. The selection problem (Case 3). You picked a metric that is easy to move but does not correlate with the outcome you care about. Go back to first principles: what behavior, if it increased, would make this product more valuable?

4. The aggregation problem (Case 4). The top-level number looks fine but hides a bottleneck at a specific funnel stage or user segment. Segment everything.

5. The leading-vs-lagging problem (Case 5). Your growth metric is growing but your retention metric is decaying. The aggregate hides the cohort truth. Always look at cohorts.

When you are in an interview and someone says “revenue dropped 20%,” start with this list. Identify which problem type it is. Then decompose. The structure will carry you further than any memorized framework.

Where to go next

• The systematic diagnostic process for metric drops: Diagnosing Metric Drops
• Choosing KPIs for your product: Metrics and KPIs
• Applying metrics thinking to growth: Growth Analytics
• Indian market context for these problems: Indian Market Cases