Flipkart PM Interview — Grocery North-Star Metric

It tests whether you can define one north-star metric for the Flipkart Minutes grocery vertical and defend it under pressure. The interviewer probes whether your metric reflects real customer value, whether it has explicit guardrails so it cannot be gamed, and whether you connect it to Indian grocery economics like basket size, repeat rate, and contribution margin. You are also tested on whether you can build a small input-metric tree and reason about second-order effects before being prompted.

Start by clarifying the launch goal and the customer, then propose one crisp north-star metric tied to value and frequency. Build a short tree of input metrics that mechanically move it. Name guardrails and counter-metrics that stop the metric from being gamed by discounting or margin destruction. State the denominator and the customer segment for every number. Close by naming the tradeoff between GMV, retention, and margin you are consciously accepting.

The biggest mistake is picking raw GMV as the north-star metric with no counterweight. Others include naming zero guardrails, reciting a framework name without applying it to Flipkart grocery, quoting a metric with no denominator or segment, treating all grocery buyers as one base, going fully qualitative when numbers are asked for, and backtracking the moment the interviewer pushes back instead of defending the choice with reasoning.

It behaves like a real loop interviewer but is consistent and never tired or disengaged. It interrupts when an answer drifts, never offers praise mid-round, and probes every claim at least once before moving on. It will deliberately propose a way to game your metric to see if you catch it. Unlike a human, it produces a transcript-backed scorecard naming the exact tradeoff you could not justify.

Scoring is derived only from your transcript. Graded dimensions include how clearly you choose and justify one metric, how well your input tree mechanically connects to it, the quality of your guardrails against gaming, whether you ground numbers in denominators and segments, and how you defend tradeoffs under pushback. Applied numeracy and structured decomposition weigh more than framework recall.

Do not jump to a metric. Spend the opening clarifying the launch objective, the time horizon, and which customer you are optimising for. State that grocery is thin-margin and high-frequency in India so frequency and retention matter more than one-time value. Then commit to one north-star metric and signal that you will defend it with input metrics and guardrails. This framing signals outcome-thinking before the interviewer has to drag it out of you.

Do not freeze or abandon the metric. Acknowledge the specific gaming path the interviewer named, then show the guardrail or counter-metric that catches it, for example a contribution-margin floor or a repeat-rate guardrail that would move the wrong way if a team discounted to inflate the headline. If you have no guardrail for that path, add one out loud and explain the second-order effect it controls.

A strong answer commits to one metric like weekly ordering households or repeat purchase rate, explains why it captures grocery habit value better than GMV, lays out three or four input metrics that mechanically move it, names guardrails such as a contribution-margin floor and a cancellation-rate ceiling, segments tier-two versus metro cohorts, and states the explicit GMV versus retention versus margin tradeoff being accepted. It stays numeric and defends choices calmly under pushback.

GMV is a lagging, gameable total. A team can inflate it with deep discounts that destroy contribution margin, or with one-time festive spikes that mask collapsing retention. For a thin-margin, high-frequency grocery vertical in India, value comes from repeat behaviour and healthy unit economics, so a frequency or retention metric paired with a margin guardrail reflects real product health far better than a raw value total.

Very. Roughly forty percent of Flipkart quick-commerce orders already come from outside the top eight metros, and tier-two cohorts behave differently on basket size and frequency than metro cohorts. A blended north-star number can look healthy while a key segment quietly underperforms, so segmenting the metric by city tier and stating the denominator for each is a clear signal of seniority in this round.

No. You are not graded on memorising figures. You are graded on structured reasoning: choosing one metric, connecting input metrics to it, naming guardrails, and stating denominators and segments. Reasonable estimates that you label as estimates are fine. What fails is being fully qualitative, quoting a metric with no denominator, or refusing to put any number to your reasoning when asked.