Usage Data Investigation Template
Numbers come from supplied data or are [NEEDS SOURCE]. Every statement carries a label.
Quality checks at the bottom.
1. The question
Question: · Unit (account / user / unit): · Window: · Decision it informs: Key event definitions: [exactly what counts as activated, active, commissioned...]
2. Data check
| Check | Result | Limits what we can say |
|---|---|---|
| Tracking fires correctly? | ||
| Definition stable over the window? | ||
| Who is missing? | ||
| Enough sample to read? |
3. Funnel
| Step (a behavior) | Count | Conversion to next | Notes |
|---|---|---|---|
| Biggest drop: |
4. Cohorts
| Cohort (shared start) | Count | Month 1 | Month 3 | Month 6 | Metric |
|---|---|---|---|---|---|
| Show counts with rates. No conclusions from a cohort of three. |
5. Segments
| Segment | Count | Behavior | Differs from others? |
|---|
6. Findings (labeled)
| # | Statement | Label (Fact / Inference / Assumption) | Basis |
|---|
7. What people must answer
| Question the data raised | Who can answer | Interview question or test |
|---|
8. Next
[Interview / fix tracking / recheck / experiment]
Quality checks
- A question and a decision were written first
- The data was checked before it was read
- Key events are defined and stable
- Counts appear beside every rate
- Every statement is labeled Fact, Inference, or Assumption
- No correlation is presented as cause
- The "why" questions are handed to people
- Nothing invented: every number traces to supplied data