I interviewed recently for another staff PM role at a D2C company. 1 hour and 20 minutes, and the last 10 were a case study.

I have practised cases like these for years. I even know what these rounds test for. I have coached friends through them. And yet I am slightly awkward to admit I got stumped at one and I did not notice it until I had left the interview.

This post is a bit about knowing the right way to answer a product sense question and also how you can be best prepared to answer these questions. Getting into the zone so to speak.

So I want to say up to this point, the interview went quite well. I feel like the tempo was great and I feel like the interview was interviewing. Until we finally got to the case study- in my defence, this came literally at the very end. Well after an hour of pleasantries, case scenarios and intense grilling, the interviewer wanted to get into the case interview. At this point I had talked and talked and probably could not do any more thinking or talking. Now that you know where I was mentally- let’s begin:

The case interview

You are a PM at Blinkit, where the Average order value (AOV) is low and you have been asked to raise it. You have already shipped two things. One is a minimum cart order value. The other is a shopping list feature, which has only 2% usage. What would you do?

Read it again before you scroll. Notice what you reach for first.

The answer almost everyone gives

First understand the users here, segment them based on how they use the platform- which is two strong groups and smaller miscellaneous subsets.

1. Regular shopper: This group is replicating a weekly or monthly shop, probably working off a list they keep offline.

2. The need-based shopper: One who needs three things in the next ten minutes and wants to be done quickly.

For the planners, make cart building easier. Frequently bought, last ordered, one-tap reorder. Show them items on sale/clearance.

For the urgent shoppers, help them find things faster. You could show them common order groupings. Or improve your search/discover features.

Then the ideas arrive. Free delivery above a threshold. People who bought this also bought. Deal tiles at higher price points. Basket nudges at checkout. And even unlocking discount models.

Doesn’t that sound complete? It does seem like a reasonable answer. It is also the answer that almost immediately left me. Sadly, it is the answer most interviewers hear repeatedly.

Why it is weaker than it looks

Now hear me out - apply this one test to it. Could this answer have been given without hearing the second half of the question?

It could. The cart minimum and the 2% adoption number do no work anywhere in it. The interviewer put two specific facts on the table and the answer completely ignores them.

And that is the WHOLE PROBLEM. There are segments, and there are interventions mapped to segments, but there is no diagnosis anywhere. It is a set of solutions to a problem we did not establish properly.

What a strong answer does before it proposes anything

Question the premise

In quick commerce, the cost of fulfilling an order is close to fixed. Logistics- rider, fuel, a picker in the dark store, packaging. Delivering a small basket costs roughly the same it costs delivering a large one. And that’s why AOV is a proxy for the number that actually matters, which is CONTRIBUTION MARGIN per order.

Now the useful things to know are: at what basket size does an order stop losing money, how much volume sits below that line, and is AOV soft across the board or being dragged down by one city, one time slot, one acquisition channel, one category mix.

It also means you can raise AOV and reduce total contribution at the same time.

It treats whatever already shipped as evidence

A cart minimum raises AOV by policy. It also means you lost some orders entirely and sent some users to a competitor. So investigate what happened to order frequency, and to monthly spend per user, in the weeks after that launch.

The same logic applies to frequency. A customer placing four small orders a month may be worth more than one placing a single large order. If that holds, then low AOV is a description of how your customer shops.

And remember at this point we also didn’t try to size either of the user groups, understand their respective AOVs to prioritize where to focus our efforts.

It refuses a number that arrives without a denominator

You should ask for more info.

• What was the target at launch? 2% against a goal of 3% is a different conversation from 2% against a goal of 25%.
• Which users were we targeting? 2% of which users adopted it? Of all users, of users who were shown the feature, of users who order more than twice a month?
• How has it moved? Has it been flat since launch or climbing every week?
• What do the 2% do? If shopping list users order more often and spend more per month, the feature works and the problem is distribution. If their numbers look like everyone else’s, the feature does not do the job it was built for, and raising adoption would not solve the problem.

There is a real chance that 2% is close to the ceiling. A shopping list is a tool for planned purchase. If planned purchase is a thin slice of the order mix, a small adoption number is the feature working as designed on the population it fits.

The order of operations

1. Investigate the metric- ask why? Find the business outcome behind it, then check whether the metric can move while the outcome gets worse. If it can, say so immediately. If you are unsure of how the metric is calculated- also ask that.
2. Treat every prior intervention as data. Ask what it did to the other metrics. Investigate- ask as many questions to understand the scenario. You wouldn’t make an ill-thought out recommendation, would you?
3. Locate the loss before generating ideas. Cut by segment, city, time, channel, category. Find the patterns. A problem that lives in one segment does not want a solution aimed at everyone.
4. Interrogate every number you are handed. Denominator, benchmark, trend, target.
5. Then solve, for the segment you diagnosed, naming the metric each idea moves and the thing it might break.

The interview is to test steps one to four. Step five is the presentation. Any competent PM can generate ten interventions for a low AOV, but only a great PM can separate the top 3 ideas and experiment with 1 of 3 which is most likely to be the winning idea. And you are getting paid to have THAT skill.

Why you will not do this on the day

Knowing all of the above and producing it under pressure are separate skills, and the second one is the one being tested.

Part of it is the setting. On the job you have context and you have standing, so questioning a metric costs you nothing. In an interview both are gone. You are a stranger with sixty odd minutes to prove you are competent, and competence feels like producing answers quickly. Asking four clarifying questions feels like you’re unsure.

The other part is compliance. Sometimes it is hard to question things. It may not always come naturally. So ask why.

What goes wrong in the room, and what to try

Then there is everything that has nothing to do with product sense.

You might ramble. You start talking, you keep talking, and it is like watching a car crash in slow motion. What has worked for me is a small self-deception.

Assume your words are rationed and you only get a fixed number today. Something about holding that assumption makes me speak with more intention and clarity.

You might lose the word. You know exactly which word you want and it simply is not there.

The same fix helps, oddly. Fewer words leaves your brain room to find it.

You might blank completely. A question lands on something you have genuinely done, and nothing comes.

There is no clever fix. You just need practice- practise your story, practise your work.

You might go scatterbrained. Not short for an answer, but unable to assemble one. Everything you know arrives at once and none of it is in order.

Work in trees. Split into branches, in shorthand on paper, and then speak from the tree instead of from memory. This keeps the pressure to recall it all away.

And the tip I would give above all the others, which sounds silly until you try it. Sing a song before the interview. And if you can also play an instrument with it.

You cannot rush a song. It runs at its own tempo and it makes you run at that tempo too. 5 minutes of that and you have already slowed down, breathing properly and ready for what the interviewer has to throw at you.

Because the art of giving an interview is being present in that moment and doing justice to yourself. Not to the role. To yourself.