Launch and growth prompts

Here is what this feature actually does, technically and in terms of what a user can now do that they couldn't before. Write three versions of a launch announcement: an in-app banner (under 20 words), a changelog entry (2 to 3 sentences), and an email (under 100 words with a single clear action).

Each version should name the specific, concrete thing that changed for the user. None should use "exciting," "new and improved," or similar language that doesn't tell the reader what actually changed if they skip the rest of the sentence.

What changed: [describe it plainly]

Here is a metric before and after a launch, plus context on what else was happening at the same time (other launches, seasonality, marketing pushes, anything relevant). Before concluding this launch caused the move, list what else could plausibly explain it, and for each, state whether the timing and magnitude of the move is more consistent with the launch or with the alternative.

Then give me a plain statement: how confident should I be that this launch specifically caused this move, and what additional data would raise that confidence.

Metric: [before, after, and dates]
Other context: [anything else happening in the same window]

Here is data or research on why users in this segment actually churned: [paste your churn reasons, quotes, or data]. Write a win-back message that addresses the specific reason, not a generic "we miss you" message.

Write two versions: one that leads with what's changed since they left (if something has), and one that asks a single question to find out if their specific reason for leaving has been resolved for them, without assuming it has.

Do not write a message that would make sense for someone who churned for a completely different reason. If the message is generic enough to send to any churned user, it needs to be more specific.

Here is a growth loop described step by step: [describe each step, for example "user completes a budget, sees a shareable summary card, shares it, a friend clicks the link, a friend signs up"]. For each step, estimate what would have to be true for that step to have a high completion rate, and flag which step is most likely to be the actual bottleneck based on how these kinds of steps typically perform, and why.

Then suggest, for the step you flagged, two ways to test whether that's really where the loop breaks, using only data we could plausibly already have or cheaply add, not a large new build.

Loop: [describe your loop]

Here is an overall metric that stayed roughly flat after a change. Before concluding the change had no effect, suggest 3 to 4 ways to segment the user base that might reveal it moved differently in different groups and cancelled out overall (for example, by signup cohort, by platform, by whether they'd already done the related action before the change).

For each segmentation, state what result would suggest the change actually worked for a subgroup even though the top-line number looks flat.

Metric and change: [describe what changed and the overall metric result]

Help me run a post-launch retro for [feature]. Here is what we predicted would happen (the hypothesis and target metric) and what actually happened (the real numbers).

Write five specific questions for the retro that compare the prediction to the actual result, not generic questions like "what went well." At least one question should address what we'd do differently in the experiment design itself if we ran this again, and at least one should address whether the stopping rule or guardrail we set in advance actually got used the way we planned.

Prediction: [your original hypothesis and target]
Actual result: [what happened]