Updated on 13 October 2026.
Use this template to give an agency client clear observations and an actionable task list. The example data explain the method and do not represent results from Semly or its clients.
1. Copy the scope record
| Field | Value to complete |
|---|---|
| Client and report owner | [name; responsible person] |
| Period and time zone | [start–end; zone] |
| Country and language | [separate values] |
| Systems and environment | [system; available settings] |
| Question set | [version ID; count; branded and unbranded groups] |
| Planned and valid observations | [counts; collection failures separately] |
| Metric definitions | [numerator; denominator; classification rule] |
| Configuration changes | [date; description; comparison impact] |
State whether you count answers, questions or occurrences. A mention identifies the brand; a recommendation presents it as an option; a citation points to a source. Keep them separate. Quote a supplier's metric definition accurately and do not label your own calculation as an official Semly metric.
2. Complete the observation record
| Field | Value to complete |
|---|---|
| Question and date | [text; observation time] |
| Observation | [what the answer actually contains] |
| Evidence | [answer identifier; checked source URL] |
| Interpretation | [what this sample supports] |
| Limitation | [what remains unconfirmed] |
| Task and existing URL | [specific correction; address] |
| Owner, deadline and acceptance | [person; date; completion condition] |
Keep full answers and client information in the agreed private location. Public examples require a separate disclosure decision. Publishing private prompts or findings is not a default part of this template.
3. Worked example using example data
In the example data, an agency checks the brand Example in one country and language. Of 12 planned observations, 10 return valid answers and 2 fail. The brand appears in 3 of the 10 valid answers. This illustrative calculation gives 30%; failures are reported separately.
One answer points to an outdated offer page. The task is to verify the condition on the existing page and correct a confirmed error. Acceptance means the responsible person has approved the information and the public URL works. This does not support a sales forecast or attribution of a later improvement to that correction alone.
4. Review before delivery
Check scope and units, recalculate the example, open sources, explain missing observations and assign task owners. Period comparisons must identify configuration changes. No universal question count or observation window guarantees a reliable result. End with a decision about work, not a promise of AI recommendations.
