The two-sources-of-truth problem, and how to end it

There's a meeting that happens every month in every company that spends money on marketing. Marketing shares a screen: the campaigns drove $2.4M in revenue. Finance shares a screen: revenue from those same campaigns was $1.7M. Both screens say "revenue". The numbers are forty per cent apart.

What happens next is always the same. The meeting stops being about what to do and becomes about whose spreadsheet is telling the truth. It won't be the last time. This is the two-sources-of-truth problem, and it's quietly the most expensive data issue in marketing – not because of the discrepancy itself, but because of what the discrepancy does to every decision that follows.

The uncomfortable truth is that both numbers are usually "right" – they're just answering different questions on different clocks.

Marketing's number comes from the platforms: conversions attributed within click and view windows, counted gross, each platform grading its own homework. It's a number built for optimisation – fast, directional, generous. Finance's number comes from the ledger: orders recognised, net of cancellations and returns, deduplicated across channels, booked when the money settles. It's a number built for stewardship – slow, conservative, exact.

Different definitions (what counts as a conversion), different timing (when the conversion is counted), different deduplication (how many systems claim the same customer). Three gaps, one word – "revenue" – stretched over all of them. Neither side is lying. But a word that means two things in one meeting is a word that means nothing at all.

The obvious cost is the meeting itself: the monthly relitigation of whose data is true, burning the hour that was meant to decide something.

The bigger cost is the discount. Once finance has been burned by marketing math a few times, every marketing number gets mentally marked down – including the true ones. Show a genuinely incremental result, a clean holdout, a channel that really did drive growth, and it lands with the same credibility as the inflated platform claims. The two-sources problem doesn't just waste time; it's how good measurement loses budget to bad intuition.

And there's a third cost, quieter still: analysts stop reconciling because reconciliation is thankless, so the gap drifts wider each quarter, and each quarter the two rooms trust each other a little less.

The instinct is to demand a single source of truth: one dashboard to rule them all. It fails every time, because both numbers are doing real jobs. Platform numbers are how you steer within a channel; the ledger is how you account for the business. The fix isn't choosing one. It's agreeing the translation between the two.

Then the rule that makes it stick: one metric of record per decision. In-platform numbers optimise inside channels. The reconciled number allocates between them. No number gets to do both jobs – that's exactly the rule that keeps attribution honest everywhere else, applied at the boundary where it matters most: the one between marketing and finance.

Publish the bridge monthly, next to both numbers, so the gap is explained rather than discovered. The first reconciliation is uncomfortable – the adjustments have owners, and nobody enjoys seeing their favourite channel's claim shrink on paper. After that it's a utility, like the bank reconciliation nobody throws a meeting about.

The end state isn't agreement on every number. It's that marketing can stand up and say "our platform number will run higher than the ledger, and here's exactly why, to the dollar" – and be right every single month. Predictable variance is trustworthy. Surprise variance isn't.

Notice what changed: not the platforms, not the dashboards, not the definitions. The two numbers still differ by forty per cent. But the difference now arrives with its reasons attached, and the meeting goes back to deciding things.

Marketing and finance will never count the same way, and they shouldn't – they're different instruments built for different jobs. The job of measurement is to make the difference boring: named, sized, and reconciled every month. Boring numbers are the ones that get believed.