KPI drop diagnosis and action plan

Investigate why a key metric moved: a structured hypothesis tree, the data cuts to run, and a prioritized plan to confirm and fix.

Change note: Initial version

Prompt

Help me diagnose a change in a business metric.

Metric and change: {{metric}}
Business context: {{context}}
What changed recently (releases, campaigns, pricing, seasonality, team, market): {{changes}}
Data I can access: {{data}}

Act as an analyst working with me step by step:
1. **Restate the problem** precisely: metric definition, size of change, time window, baseline, and whether the change is outside normal variation. Tell me how to check that quickly.
2. **Metric decomposition**: break the metric into its drivers (for example revenue = visitors x conversion x order value). Show the formula.
3. **Hypothesis tree**: list 8 to 10 candidate causes grouped by Data issue, Internal change, Customer/segment mix, Competition/market and Seasonality. For each, state what we would see in the data if it were true.
4. **Analysis plan**: a ranked list of the cuts or queries to run, in order of speed and diagnostic power, with the exact segmentation (by channel, product, cohort, region, device, time). State what result confirms or rules out each hypothesis.
5. **Most likely explanations**: your top 3 based on my details, with your confidence.
6. **Actions**: what to fix now, what to monitor, and who to talk to.
7. **Summary for leadership**: 3 sentences, written so it stays accurate whatever the analysis finds.

Do not claim a cause as fact until the data supports it.

Variables

• {{metric}}Metric and changetextrequired

  Default: Weekly new customer sign-ups fell 22% over the last 4 weeks

• {{context}}Business contextlongrequired

  Default: B2C meal-planning app, subscription model, most sign-ups come from paid social and search.

• {{changes}}What changed recentlylongrequired

  Default: Website redesign went live 5 weeks ago. Ad budget unchanged. A competitor launched a free tier. School holidays started 3 weeks ago.

• {{data}}Data availabletextrequired

  Default: Analytics funnel by channel and device, ad platform reports, CRM sign-up records

Example output

**Problem:** Sign-ups down 22% for 4 weeks, starting roughly when the redesign launched. First check: compare against the same weeks last year and the weekly standard deviation.

**Decomposition:** sign-ups = traffic x landing conversion x sign-up completion, by channel.

**Hypotheses and what we would see**
- Data: tracking broke in redesign - sudden step change on one day, ad platform conversions unchanged.
- Internal: redesign hurt conversion - traffic flat, conversion down, especially on mobile.
- Mix: paid traffic quality fell - click-through steady, conversion down only for paid social.
- Market: free competitor tier - search impressions steady, brand search down.
- Seasonality: holidays - last year showed a similar dip in the same weeks.

**Analysis plan (in order)**
1. Conversion by device before and after the redesign. Confirms redesign if mobile fell more than desktop.
2. Ad platform conversions vs analytics sign-ups, to rule out a tracking issue.
3. Same weeks last year.

**Likeliest:** redesign impact (medium-high), tracking gap (medium), seasonality (low-medium).

**Actions:** audit sign-up events today; run A/B of old and new page for a week.

**Leadership summary:** Sign-ups have dropped about a fifth since the redesign. We are checking tracking, mobile conversion and seasonality and will report the cause within three working days.

Notes

The hypothesis tree states what each cause would look like in data, which makes the analysis falsifiable rather than a hunt for a story. Add real numbers and it will rank the cuts to run.