A category-defining vertical SaaS, scaling past Series B, had accumulated four years of legacy demand-gen architecture. Compounding ad spend was no longer compounding pipeline.
A vertical SaaS platform, recently Series-B funded, was approaching the awkward middle of growth. Blended customer acquisition cost had climbed 2.4× over eighteen months. Marketing felt like an expensive black box; the leadership team disagreed about what was working and why.
The brief, from the CFO directly: "We need to know what every dollar is doing. If a channel is dead, we want to bury it. If a channel is alive, we want to feed it."
Our brief, internally: rebuild the attribution architecture from first principles, then redesign demand generation around the truths it revealed.
Three months of forensic work revealed what marketing leaders quietly suspect but rarely test: the attribution dashboard was telling a coherent story that was, in significant part, fictional. Last-touch attribution credited the channels closest to the buy button while obscuring the channels that created the demand to begin with.
When we ran proper geo-incrementality tests against the most expensive paid channels, two of the top five showed near-zero incremental lift. They were claiming credit they hadn't earned.
Equally surprising in the other direction: a small content programme, dismissed by the previous team as a vanity expense, was the single most productive top-of-funnel asset in the portfolio.
The most expensive line item on a CFO's marketing P&L is the channel you only think is working.
Step one: build a measurement spine the team could defend in a board meeting. We unified ad-platform data, CRM stages, product analytics, and revenue into one warehouse with one attribution model — the definition of which we wrote down, signed, and agreed not to change without explicit decision.
Step two: a rolling incrementality programme. Every paid channel above a defined spend threshold was tested on a 12-week rotation. Channels that earned their place stayed and grew. Channels that didn't were paused — politely, but firmly.
Step three: rebuild creative around what the measurement actually rewarded. Long-form editorial content, technical documentation as marketing, founder-led video — all of which the old last-click model would have killed within a quarter.