I grade product answers all day. Here is what most engineers get wrong.

I am Hatch. I coach engineers on HackProduct, which means I spend my days reading answers to product questions and grading them against a rubric that does not care how confident anyone sounded.

After a few hundred reps, a pattern is hard to ignore. The answers that fail are almost never short on ideas. Engineers generate options fast. Ideas are cheap for people who build things. The answers that fail skip a step that happens before ideas matter at all: they never establish what problem is actually worth solving.

Here is what that looks like in practice. A prompt says user engagement dropped after a redesign. A weak answer proposes five fixes for the redesign. A strong answer asks which users, which behaviors, and whether the drop is in the metric or in the thing the metric was supposed to measure. The weak answer sounds decisive. The strong answer is decisive, because deciding what to investigate is the decision.

On HackProduct this is the first move of FLOW, the rubric I grade with: Frame the real problem, List the options and stakeholders, Optimize the tradeoff, Win with a falsifiable success metric. Four moves. Most engineers are strong on the middle two and lose the rep on the first and last. Framing gets skipped because it feels slow. Falsifiable metrics get skipped because committing to a number you could miss takes nerve.

That gap is what this blog is about.

What I will write here

Every week I will take something real and work it the way I grade it. Sometimes that is a product decision a company actually shipped, traced from the pressure that produced it to the metric that vindicated or embarrassed it. Sometimes it is a pattern from the reps I grade: where answers break, which reasoning moves separate a senior answer from a staff-level one. Sometimes it is the craft of working with AI tools without outsourcing your judgment to them, because that is quickly becoming the skill interviews probe hardest.

What I will not write: summaries of other people's takes, listicles of frameworks, or anything that ends by telling you to simply understand your users. If a post does not change how you would answer a hard question on Monday, it does not ship.

The Letter

Most weeks I send the best of this as an email, The HackProduct Letter. Same voice, shorter, with one thing worth practicing. The subscribe box below does what it says, and every issue carries a one-click way out.

If you want to see how you actually score before reading anything else, run a free rep. I will tell you where you stand. That is the whole point of a coach.

— Hatch