Google says 50, Meta says 40, the CRM says 35. Who is right?
Paste what each source claims per channel - each ad platform's own count, GA4 or your analytics, your CRM or store, a "How did you hear about us?" survey, and spend - and, if you have them, your converting touch paths. Your browser picks one source of truth for the count, measures how far the platforms over-claim against it, prices each channel per real conversion, scores six attribution models side by side and shows where direct, branded search and dark social hide the real driver. Free, before you sign in, and nothing is uploaded. Then the readout makes the allocation call, and the test lane designs the incrementality test that settles the channel two sources fight over.
The companies and numbers in the examples are invented. Three are readouts (the B2B one also carries its test, one click after the readout) and one is a test design; each comes with a saved model reply, so you can see the whole page without signing in or spending a credit.
What this does, and what it does not
Every platform counts the same sale in its own window, so the sum of what Google, Meta and LinkedIn report almost always exceeds what really happened. Analytics loses cross-device and cookie-blocked visitors and files them under direct. The CRM holds whatever the form or the rep captured. Buyers remember the memorable touch. Nobody is right; the job is one honest count, an explained gap and a confident direction. This page does the arithmetic first - the source of truth, the over-claim, each channel's cost per real conversion, the direct and branded share, what buyers say against what tracking credits, and every model over your journeys - and reads your notes for the mistakes the skill warns about: summing platforms, a single-touch model on a long cycle, data-driven attribution at low volume, no self-reported data, redirect referrers, view-through windows, a media mix model below its scale.
The paid lanes see your sources (the middle is cut past 9,000 characters, and the page says so), your paths (whole journeys past 4,000) and the page's scan. Your input stays in this browser unless you run a lane. Every reply is held to your numbers: quotes are checked word for word and no figure, window or duration may be invented or rounded. Attribution is directional, not truth. Derived from the agent skill @coreyhaines31/attribution by Corey Haines (coreyhaines31/marketingskills, MIT): its source-of-truth and triangulation rules, the six models and how each lies, MTA, MMM and incrementality, self-reported attribution, the blind spots and the B2B and ecommerce playbooks. Sources are listed in llms.txt.
After the first result: save the scan, and next month paste the new numbers and compare them with it - the source of truth, the over-claim and each channel's share and cost per real conversion, then and now. When the incrementality test has run, type the two groups' conversions under the test design and the page reads it: incremental conversions, the lift with its 95% interval, the incremental cost per conversion and which of the three readings applies. Both are free and happen in your browser.