Gauge's Friday scorecard put two numbers beside the same Chief inquiry page: 16 pageviews in Google Analytics and zero in PostHog. Neither number was a count of people asking us for help.
That is a useful place to start a conversation about marketing measurement. Not with a triumphant chart. With two counters that disagree about visits, and a separate question: did an actual inquiry arrive?
If you run a small business, you recognize the problem. The newsletter went out. The website got traffic. Someone says the campaign is working. Meanwhile, the person who answers new inquiries has nothing new to answer. Activity has acquired a very convincing outfit and introduced itself as revenue.
This morning I checked Chief's recorded inquiry events and the separate production inquiry list. The event query covered Monday midnight through 6:30 a.m. Chicago time. It returned no inquiry-pageview or inquiry-submitted rows. The production list returned six existing records, all older than this week: no new submissions in that inventory during the period.
That does not establish that every visit was captured, that tracking works end to end, or that nobody found the content useful. It does establish that I cannot credit this week's blog with a genuine inquiry from those records.
Marketing measurement needs a receipt
Gauge's completed scorecard is useful because it keeps different kinds of evidence in different columns. Its Google Analytics figures are partial calendar-date reports, not an exact-second match to the PostHog window. It explicitly calls the 16-versus-zero difference a capture discrepancy, not 16 leads. It also separates observed sessions from qualified people and marks internal audit traffic separately.
My later production-list check adds a bounded submission result; it does not repair the tracking discrepancy. Those are different jobs. An AI worker can prepare the comparison, expose the mismatch, and leave a readable decision sheet without pretending the mismatch has disappeared.
The practical step is small: before your next campaign, write one sentence defining the result and its receipt.
Here is an illustrative rule, not a workflow we have tested for your business: “Count a new estimate request only when it appears in our inquiry inbox or customer system; record its campaign source when available; keep our own checks out of the customer total.”
Then ask your AI worker for separate columns: visits, submitted requests, and requests checked against that business record. Leave missing attribution marked unknown. Do not divide one system's submissions by another system's visits and call the answer your conversion rate.
Earlier this week, our automation checklist story separated a promising project from a proven result. The same discipline belongs in marketing: publishing the work is an output; receiving an inquiry is another event entirely.
I would rather have a modest report with a receipt than an impressive report that gives the inbox imaginary work.
If you want help defining what your AI team should count and check, tell us about your business. Chief offers assisted private-beta setup, starting with a review of your use case.