Start with the question
Analytics answers six questions a site owner asks out loud. Everything else is a dimension in search of a decision.
The map, filled in
Brightwater Dental, September 2026. One property, consent banner live, no ecommerce.
| Question asked | Read | Value | Denominator | Caveat |
|---|---|---|---|---|
| How many arrived | Organic overview | 118 | n/a | consent applies |
| Which pages caught them | Organic landing pages | /emergency 34 | 118 sessions | consent applies |
| What they did there | Page performance | 61 engaged | 118 sessions | consent applies |
| Where they came from | Traffic acquisition | search 71 | 118 sessions | consent applies |
| Did they act | Key events | 9 | 118 sessions | one event type |
| Is it even measuring | Measurement health | healthy | one page | checked by hand |
The three readings to distrust
- An average of two sources. Search Console and Analytics count different things. Averaging them produces a number that means nothing.
- A count without a denominator. Nine events from 118 sessions is 7.6%. Nine from 900 is 1%.
- A conversion rate after a consent change. The denominator moved, so the ratio moved, and nothing about the business changed.
What to do before the next report
1. read measurement health on one page, by hand
2. confirm one key event fires for one real submission
3. write the denominator beside every count
4. name the consent rate, so a fall is explainable
5. list the actions that will never appear: phone calls, walk-ins
Acceptance: the property is readable when
- measurement health is healthy on a page you checked yourself
- a key event fired once for one real submission
- every value in a report has a denominator
- the consent rate is visible beside the session count
- offline actions that never appear are named in the report
Troubleshooting
| Symptom | Likely cause | What to do |
|---|---|---|
| No data for a day | The tag was blocked, or consent denied | Check one page’s HTML for the tag |
| Sessions far below search clicks | Consent, or an unmeasured channel | Compare with the consent rate |
| Events double | Two tags, or a retry | Count the tag on the page |
| The client asks for revenue | The property does not record it | Say so, and report enquiries instead |
| Every report looks flat | Nothing changed on the site | Propose one test, then read it |
Limits
Analytics is consent-dependent and configured by somebody. A low number can be a broken tag rather than a quiet month.
Read measurement health before any other report, and treat a missing value as missing rather than as zero.