The Dunwell standard
The value of an analysis is the decision it helps you make. Here is how Dunwell keeps the question, the evidence and the recommendation connected.
Begin with the choice you face, the time available and what a useful answer would change. Agree the scope and check whether your data can support it before you pay.
Example: which customers should we prioritise for a repeat-purchase test?
Check dates, duplicates, missing information and cancellations. Define each metric and reconcile totals before interpreting a trend.
Revenue, orders and customers answer different questions. Their definitions belong alongside the findings.
Look beyond the headline. Compare customer groups and periods, consider alternative explanations and use modelling only where it improves the answer.
A pattern is evidence to investigate. It does not, by itself, establish cause.
Rank practical actions by effort and potential value. Make costs, assumptions and uncertainty explicit. Where possible, define a test that can show whether the action worked.
Scenario values are illustrations, not promised returns. Overlapping opportunities must not be added together.
Read the evidence correctly
What happened in the available records, under stated definitions and dates.
A model estimate, evaluated against data it was not fitted to at that time. Its limits and evaluation period matter.
What might happen if stated assumptions hold. Useful for planning, with no guarantee of an outcome.
The Growth Check includes the report, dashboard and a public-data example of a customer follow-up list.