Field guide · Promotion measurement

Published work is the beginning of the measurement.

A content receipt tells you an action happened. It does not tell you whether the right prospect arrived or whether an engagement paid for the work. Keep those questions separate.

1. Define the outcome and the window

Choose the service, audience and decision you want to support. Record the reporting dates and what data is available. A short experiment with few inquiries needs interpretation, not automatic budget escalation.

2. Count distinct things in distinct columns

Keep the original inquiry reference when an operator qualifies an opportunity.

  • Work: approved drafts and actual publication receipts, including failed or uncertain actions.
  • Response: observed visits and persisted inquiries; separate bots, previews and internal QA.
  • Opportunity: a reviewed fit, discovery booked and attended, proposals and signed engagements.
  • Money: verified payments and refunds. Quotes and pipeline values are not collected cash.

Keep the observation surface with the evidence

For AI answer observations, retain query, engine/product, exposed model version, locale, observation timestamp, session conditions, cited URLs, method and repeated-run count. A consumer answer and an API answer are different surfaces. Neither supplies a universal AI rank.

  • Search Console imports are partial source reports, not automatic live synchronization or per-lead AI Overview attribution.
  • Manual-verified, API-verified, fixture, uncertain and not-connected results must stay distinguishable.
  • Research prospects are not submitted inquiries, and an inquiry becomes qualified only after private review.

3. Keep attribution modest

A permitted campaign tag may explain an observed visit or form source. It cannot reveal every earlier interaction or identify a person behind a search query. Keep unknown and self-reported sources visible. Use one named additive attribution model; assists are separate.

4. Use denominators people can inspect

Cost per qualified opportunity divides the defined period’s allocated cost by its qualified opportunities. When none are recorded, show N/A and the counts. Separate estimated labor, reconciled expenses and actual provider charges; allocate shared research only once.

5. An illustrative journey

Synthetic example only: an owner approves a resource summary, publishes it manually, verifies the link, and later reviews a saved project inquiry. The operator qualifies the project and records a proposal. Until a payment is verified, the journey has no collected revenue. This is not a real customer or a measured conversion result.

6. Review what needs attention

Check failed or uncertain publications, access health, inquiries without a next action, available search data and delivery effort. Decide whether to improve the offer, page or follow-up before changing the distribution budget. Missing analytics must remain unavailable, never replaced with a plausible-looking number.

A useful next step

Discuss your measurement process

A rough outline is enough to start a scope conversation.

Discuss your measurement process ↗

Content revised 2026-09-27.