Case study · built for our own companies
The ad platform said 4.87x. The ledger said 0.4.
Platform-claimed 4.87x vs a true 0.33 to 0.50 contribution-margin ROAS
Our own D2C brand was spending real money on paid social. The platform dashboard reported a 4.87x return on ad spend. By that number, we should have been scaling the budget aggressively.
The ledger disagreed. When we reconciled platform-claimed conversions against actual orders, margins, refunds, and fees, the true contribution-margin ROAS came out between 0.33 and 0.50. Not 4.87. The dashboard was not lying, exactly; it was attributing sales it did not cause and ignoring costs it never sees.
What we built
An ad-spend truth engine: a system that pulls spend and claimed results from the ad platform, pulls orders and costs from the store and the books, matches them, and publishes one daily answer to one question: what did a dollar of ad spend actually return, after everything?
The build was done the same way we now sell as the AI Workforce Sprint: fixed scope, built on our own infrastructure, documented, handed over to the operating team.
What changed
Budget decisions moved from the platform dashboard to the ledger number. Spend was cut where the truth engine showed it was underwater and held where it was defensible. The uncomfortable spreadsheet argument (“but the platform says…”) ended, because there was now a single reconciled number both sides could check.
Why we publish this
Because it was built for our own company, with our own money at stake, we can publish the numbers without a client’s permission and without polishing them. The claimed 4.87x versus the real 0.33 to 0.50 is the strongest argument we have for why operations need a truth layer before they need more automation. Figures as of July 2026.
If your dashboards and your books disagree, this is the exact category of system we install. Describe your version at /start.