Analysis prompts

I'm going to paste a table of customer feedback: one row per quote, with a problem label, a severity score, and a user type. Turn this into a ranked list of claims a product manager could defend to a founder.

For each distinct label, calculate: row count, average severity, and how many rows are from new versus existing versus churned users. Rank the labels by a combined score of reach (row count) and severity, and for the top three, write a one sentence claim in this exact shape: "[X percent / X out of Y] of [feedback type] describe [problem], most severely for [user type], for example: '[a real quote from the data]'."

Do not round the numbers in a way that makes the claim sound stronger than the data supports.

Data:
[paste your signal log]

I have a hypothesis about why a metric changed. I'm going to give you the hypothesis and the data behind it. Do not evaluate whether my reasoning is internally consistent. Instead, generate two alternative explanations for the same data that I have not considered, and for each, describe what additional evidence would distinguish it from my hypothesis.

Then tell me: based only on the data I've given you, could someone reasonably prefer one of your alternative explanations over mine? If so, say which one and why.

My hypothesis: [state your hypothesis]
The data: [paste the data or describe what you observed]

I have a cohort finding: [describe the correlation, for example "users who set a budget in week 1 retain at 41 percent versus 19 percent for those who don't"]. Help me think through whether this is likely to be causal or a shared cause behind both.

List at least two plausible reasons this correlation could exist without budget-setting causing retention (for example, a third trait that makes someone both more likely to set a budget and more likely to stick around). For each, suggest what data I would need to check to rule it in or out, using only data a small startup's analytics could realistically produce, not an idealised experiment.

End with a plain statement of how confident I should be in a causal read given only what I currently have.

Here is a conclusion I've reached and the evidence behind it. Argue against it as persuasively as you can, using only the evidence I've given you plus reasonable inferences, not invented data. Do not soften the counter-argument to be polite. Assume you are a colleague who genuinely disagrees and has to convince a skeptical founder.

After the counter-argument, tell me honestly: does my original evidence hold up against it, or does the counter-argument expose a real gap I should address before I send this?

My conclusion: [state it]
My evidence: [summarise or paste it]

I'm considering [metric] as a metric a team will be evaluated on. Help me stress test it before we commit to it.

List the ways someone could improve this number without improving the thing it's meant to represent, being specific to how our product actually works, not generic gaming strategies. For each way, suggest a counter-metric that would catch it, something we could realistically track alongside the primary metric without much extra instrumentation.

Then tell me: of the gaming strategies you listed, which is the one a team under quarterly pressure would most likely reach for first, and why that one specifically.

Metric: [your metric]
Context on how the product works: [brief description]

I'm going to paste a data summary or set of numbers. Write a two to three sentence summary of the overall pattern, and separately, in its own paragraph, describe the single most surprising or largest deviation from that pattern, even if it's a small subgroup. Name it specifically: which segment, what number, how far it departs from the average.

Do not fold the outlier into the overall summary as a caveat. Give it its own space, because it's often the more important finding than the average.

Data: [paste your numbers or summary]