Skip to content
B2B SaaS / Measurement5 months · 2026

The channel they were about to cut was carrying the business

Attribution said paid search was their worst channel. A holdout test said it was the only one that mattered. The decision to cut it was three weeks from being signed off.

-41%

Reported vs actual revenue gap

+34%

Conversions recovered

6 wk

Audit to first clean month

3.2x

True incremental return

The challenge

Ashgrove sells workflow software to mid-market operations teams at around $30,000 a year. Their dashboard showed paid search returning less than half what paid social did, and a plan was circulating to move the entire budget across. What nobody had checked was that the CRM held no click identifiers, so any deal that involved a phone call — most of them — was attributed to whatever the buyer clicked last, usually a branded search or a retargeting ad. The channel creating demand was invisible and the channels capturing it were taking the credit.

What we did

01

Reconciled platform claims against booked revenue first

Before touching a campaign we summed what every platform reported and set it against what finance had actually recorded. The gap was 41%, and that single number moved the project from a marketing initiative to a board-level one in about ten minutes.

02

Rebuilt collection server-side with the click identifier persisted

Events sent from their own infrastructure with a shared identifier, and the original click ID written into the CRM record at form submission. That one field is what made offline conversion import possible, and it had never existed.

03

Re-pointed optimisation at closed-won, not demo requests

The platforms had been optimising toward form fills for two years and had become extremely good at finding people who fill in forms. Importing actual won deals with their values changed what the algorithm was hunting for, and the composition of inbound shifted within a quarter.

04

Ran a geo holdout before anyone cut anything

We paused paid search in a set of matched metros for eight weeks — longer than their sales cycle — and compared total pipeline against control regions. Pipeline in the paused markets fell substantially. The channel attribution had ranked last was the one creating the demand the others were closing.

The result

The holdout stopped a budget reallocation that would have removed the top of their funnel. Once closed-won data was flowing back, paid search showed a true incremental return of 3.2x against the 1.4x attribution had reported, and conversion volume recovered 34% from server-side collection alone. The reported-versus-actual revenue gap closed from 41% to under 6%, which is the number their CFO now uses. Media spend rose rather than fell — but it rose against a figure both marketing and finance had agreed on first.

We were three weeks from cutting the channel that was actually feeding us, and we would have had a defensible-looking dashboard justifying it the whole way. What I found hardest was accepting that our reporting had been wrong for two years, not that it needed fixing.
Yusuf Adeyemi-ClarkeVP Marketing, Ashgrove Systems
Same playbook

Let's run this on your business.

Start with a free teardown. We'll show you where your funnel leaks and what we'd do about it — before you spend anything.