Use when consent or tracking loss is large enough that reported numbers need a stated margin.
consent-gap-impact-estimate.md
You are estimating what consent and tracking loss does to reported numbers. You are not giving legal advice and you are not redesigning the banner.
Consent setup, banner behaviour and defaults: {{CONSENT_SETUP}}
Consent rates by region or device where known: {{CONSENT_RATES}}
Tools in use and any modelling they apply: {{TOOLS_AND_MODELLING}}
Metrics people report from those tools: {{REPORTED_METRICS}}
Output:
1. A table: Metric from {{REPORTED_METRICS}} | Observed, modelled, or mixed | Direction of the error | Rough size as a band | What that band rests on.
2. Which of these metrics stay comparable over time, and which move when {{CONSENT_RATES}} move rather than when behaviour moves.
3. Segments hit hardest under {{CONSENT_SETUP}}, for example a region with a stricter banner, a browser that clears storage, or an app webview.
4. Three checks that would narrow the bands, each with the data needed and who holds it.
5. One sentence to put under any chart built from these metrics.
Rules:
- Never give a single percentage for unmeasured traffic. Use a band and name the assumption behind it.
- Where {{TOOLS_AND_MODELLING}} fills gaps with modelling, say which reported figures contain modelled data and which are counted.
- If consent rates were not supplied, say what the estimate cannot do rather than substituting a typical rate.
- Plain English for a marketer, not for a privacy specialist.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When do I use this rather than accepting the reported numbers?
When consent loss is large enough that a reported figure needs a band around it, or when a metric moved and you cannot tell whether behaviour or consent rates moved. It gives no legal advice and does not redesign the banner. It sizes the error already sitting inside what you report.
What do I need in front of me?
Your consent setup and banner defaults, consent rates by region or device, the tools in use with any modelling they apply, and the metrics people actually report from them. Take the rates from the banner vendor rather than the analytics tool, since that tool sits inside the gap you are trying to measure.
What comes back?
A per metric table saying whether each figure is observed, modelled or mixed, the direction and rough band of the error, which metrics stay comparable over time, the segments hit hardest, three checks that would narrow the bands, and one sentence to sit under any chart. The observed against modelled column is the useful bit.
What is the mistake that costs me here?
Quoting a single percentage for unmeasured traffic. Bands are the honest form and the prompt will not hand you a point figure. If you did not supply consent rates it says what the estimate cannot do rather than substituting a typical rate, and substituting one yourself is the same error made quietly.