The recorded enquiry rate rises after a consent dialog changes. The service page stays the same. Before crediting that page with an improvement, check who is now included in the report and when collection starts for those visits.
A comparison based on consented data initially describes the groups actually observed in each period. You do not know whether unobserved visitors behave similarly. Their activity cannot therefore be filled in with a fixed multiplier.
The rate needs a consistent denominator
Fictional calculation: Before the dialog change, a report records 12 confirmed enquiry events in 600 observed sessions. Afterwards, it records 9 in 300. The observed rate rises from 2 to 3 percent while the number of recorded enquiries falls. Neither fact explains the behaviour of every visitor.
The calculation only makes sense when its numerator and denominator belong together: the same property, period and collection rules for those sessions. Dividing all backend enquiries by analytics sessions from consenting visitors would produce a different, unsuitable metric. An enquiry remains an enquiry, not a contract or revenue.
With tracking blocked until consent, earlier activity is missing. Matomo’s documentation explains this distinction between tracking consent and cookie consent. For a comparison, later consent may therefore reveal a different section of the journey from early consent. Missing earlier steps should not be filled in retrospectively as observed actions.
Look beyond one overall percentage
Record when the dialog and collection logic changed. Examine existing, permitted breakdowns, such as device category, with the number of observed sessions in each group. If rates within a group remain similar while its share of the overall report changes, composition is one possible explanation for the aggregate movement. That does not prove a particular selection effect.
Even identical consent percentages do not establish an identical group. Newly included visitors may have different tasks. Where no sound reference total exists, describe overall coverage as unknown. Server requests, banner displays, people and sessions are different units. Collecting additional activity from people who declined solely to simplify a comparison is not a solution.
Keep observed and modeled values distinct
Some reports incorporate estimates. Google explains for GA4 how modeled and observed data can appear together and how the data-quality indicator identifies this. That does not mean another report or your own tool uses the same modeling.
Record whether each period contains only observations or also estimates. A model does not turn unknown individual journeys into observed sessions. Avoid comparing a modeled overall report with an exclusively observed subset as though they shared the same basis.
A useful conclusion may be: “Recorded rate increased; the service page’s effect remains unresolved because collection coverage changed.” Keep the periods separate and establish a documented baseline under stable rules. If the concern is a technical banner fault, test the four consent states. For reconciling the metric itself, consider the Analytics data check.