How to Attack a Strategy Round

Strategy questions reward a decision, not a survey. The fastest way to fail this round is to list considerations and never commit to anything. The interviewer wants to watch you frame a real call, weigh real options, and pick one, then defend it when they push. Fence-sitting loses even when the analysis underneath it is good. So what's really being scored is judgement under ambiguity, plus business literacy.

Can you decide with incomplete information the way a real leadership team has to? And do you understand the money, the margin, the moat, or are you only reasoning as a product person? Both have to show up. Step one, frame the real decision and the goal. State what is actually being decided, and what winning looks like. Honestly, half of a good strategy answer is defining the question well, because a vague question invites a vague answer.

If someone asks should we build our own model or use a vendor, the real decision is about control and cost and speed, so say that first. Step two, lay out two or three real options, each with its upside and its cost. Not a list of ten, which is just hiding. And not a single obvious one, which isn't a choice.

Real alternatives a leadership team would genuinely debate. Build versus buy versus partner. Go broad versus go deep on one vertical. Premium versus free-and-scale. Step three, commit to a recommendation. Say which one, and why. It depends is only acceptable if you then say what it depends on, and make the call for the most likely case anyway. The interviewer is specifically testing whether you can pick, so pick.

Step four, name the risks and the moat. What could go wrong, and why your position holds up over time. And this is where you bring real business reasoning, not vibes. Unit economics. A data moat that compounds. Switching costs. Distribution advantage. Regulatory exposure. If you name a moat, it has to be a real one. And a quick word on what a moat actually is in AI, because this is where a lot of answers fall down.

The base model is almost never the moat, because your competitor can license the same one, or a cheaper open one that's nearly as good six months later. The durable moats are the things that get harder to copy the longer you run. Proprietary data that improves your output and that a rival can't buy. Deep integration into a customer's workflow that makes switching painful.

Distribution you already own, so you reach users your competitor has to pay to acquire. A brand people trust with something sensitive. When you name a moat, ask, could a well-funded competitor buy or build this in a quarter? If yes, it isn't a moat. Say the question is, a competitor just launched a free version of your paid AI feature, what do you do?

The weak answer lists five things you could consider. The strong answer says, I'd hold the paid tier and lean into depth, because our moat is the proprietary customer data that makes our output better, and a free clone can't match that. Then names the risk, that if the free version is good enough for most users, our addressable market shrinks, so I'd watch conversion from their free tier as my early-warning metric.

That's a frame, a committed call, a moat, and a risk, in four sentences. So what earns the points? First, you commit, you make a recommendation and defend it under pressure. Second, you reason about defensibility with something concrete, not our model is better, but something that actually compounds. And third, you show you understand the money, the cost, the margin, and what builds up over time into an advantage.

Now the ways people lose it. The biggest is fence-sitting, listing pros and cons and never choosing, which reads as, this person can't make a call. Second trap, naming a moat that isn't one. A model anyone can buy API access to is not a moat, so don't claim it as one. And the third, ignoring the business entirely and answering only as a product person, because this round specifically wants the commercial reasoning on top of the product instinct.

So here's the whole attack. Frame the real decision and say what winning looks like. Lay out two or three genuine options with their costs. Commit to one and defend it. Then name the risks and the moat, with real business reasoning underneath, unit economics, a data moat, switching costs, distribution. Do that, and you've given a leadership-team answer instead of a student essay.

Memorise this. Frame the decision, give two or three real options, commit to one, then name the risks and the moat. Commit, don't survey. ---