Three numbers. One ledger. No mystery.
What this stage is for
Measurement is listed fourth but runs first and last. A baseline is taken before any work begins, because a number produced after the fact cannot tell you whether anything changed. Then it runs continuously, which is what makes the other four stages accountable rather than assertive.
The reason this stage carries so much weight at Mindflow is that the industry's normal practice here is bad. Most agency reporting reports activity, things done, hours spent, tasks closed, or reports a single composite score whose derivation the client cannot inspect. Neither tells an owner whether the work is working.
The work in this stage
Search Visibility Share
Map-grid position measured city by city rather than as a single average, plus the rankings that actually matter for your trade. A city-level average hides the two neighbourhoods where you are invisible, which are usually the ones worth fixing.
Qualified Demand
Calls and enquiries that are actually plausible work, separated from the noise. An unfiltered call count is not a business metric; a spam call and a booked estimate are not the same event and should never sit in the same total.
AI Share of Answer, scored X/12
Twelve real buying questions, run three times each, on Google AI Overviews, ChatGPT and Perplexity, scored and reported per surface and never blended. Reported as X/12 rather than a percentage, because twelve questions cannot honestly produce a percentage.
The Work Ledger
Every completed task, dated. Not a summary of activity, a record specific enough that you could check it. The Ledger sits alongside the numbers rather than substituting for them, which is the distinction most agency reporting collapses.
A quarter where the number went down
Illustrative scenario: a representative pattern rather than a specific client engagement. No client data is described here.
A client's AI Share of Answer went from 7/12 to 5/12 between quarters, on Perplexity, with no corresponding fall on the other two surfaces.
The report said so. The evidence file showed the two lost questions had both been answered by a comparison page on a national directory that had not appeared in the prior run at all, a source change, not a decline in the client's own position, and visible only because the raw responses were captured rather than summarised.
That is the argument for publishing the protocol and keeping the evidence. A composite score would have shown a decline and offered no way to tell whether it was the client's problem, a competitor's gain, or a change in what the engine consulted. Those three call for entirely different responses.
The complete X/12 Protocol, covering question selection, personalisation controls, the exact calculation, repeat-run handling, evidence capture and known limitations, is published on the methodology page. It is the measurement half of the Visibility Engine, and it is public so the number can be checked rather than trusted.
This stage proves itself: the protocol behind AI Share of Answer is published in full, including how the twelve questions are chosen, how personalisation is controlled, the exact calculation, how inconsistent runs are handled, and the known limitations. You can reproduce the number, or have someone else reproduce it.
The failure mode here is the composite score, one number that blends everything and can therefore always be made to look like progress. Three separate numbers can disagree with each other, and a quarter where one falls is reported as a number that fell.
Fair questions
Why X/12 and not a percentage?
Because twelve questions cannot produce a meaningful percentage. Reporting 8/12 as 66.7% implies a precision the sample does not support.
How often does the AI number update?
Quarterly. These surfaces move on their own between runs, and a monthly cadence produces noise an owner would reasonably mistake for progress.
Can we see the raw data?
Yes. The evidence file, full-text responses, timestamped, per run, is delivered with the report and belongs to you.
What if the numbers go down?
They get reported as having gone down, with what we know about why. A measurement system that can only show improvement is not a measurement system.
Start with a measurement, not a pitch
A free Visibility Check shows where you actually stand across Google and AI answers, before any conversation about scope or price.
Get my free Visibility Check