The blog · GTM Systems · From The Demand Compass · 7 min
Why is last-click attribution wrong for B2B?
It hands the whole deal to the final form fill and throws away the nine months of movement that made it possible. What a board can fund is the account's timeline, which shows cause instead of credit.
Last-click attribution is wrong for B2B because it fights over credit for the final form fill and throws away the six to nine months of movement that made it possible. It measures the harvest and ignores the farm. A considered purchase runs on a clock that long, while the budget lives upstairs with a board running a ninety-day clock, so the work that pays off three quarters out is judged against what closed this quarter and loses every time. What wins the argument is not a better attribution model but a translation layer: each account's story over time, what it touched and how it moved from state to state toward a deal, which is what a board can fund because it shows cause rather than credit.
What does last-click attribution get wrong?
Last-click attribution assigns the whole outcome to the final touch before the form fill, which in B2B is almost never the touch that decided anything. A considered purchase forms slowly: intent builds, the signer leaves no signal, a committee assembles, and the deal sits still until a trigger moves it. Six to nine months from first faint signal to closed deal is normal, not slow. The last click lands at the very end of that, usually a branded search or a direct visit by someone who already decided, and the model hands it all the credit.
Everything before it disappears from the record: the newsletter opens, the roundtable seat, the post reactions, the pricing visits, the community mention. The awareness work that made the last click possible reads as zero, and a line item that reads as zero is the first one cut.
It measures the harvest and ignores the farm.
Why does the ninety-day clock make it worse?
Because the sale and the budget run on two different clocks, and the gap between them is where good marketing leaders quietly lose their jobs. The first clock is the sale's: six to nine months for a considered purchase. The second runs every ninety days, when the board meets, the CEO wants a number, and the question is some version of "what did marketing produce this quarter?" The work you do this quarter mostly pays off two or three quarters out, but you are asked to justify it against what closed this quarter.
The trap rarely springs as a dramatic decision. Pipeline gets tight, the pressure comes down, and the awareness work, the content, the newsletter, the events, the nurture of accounts that are not ready yet, looks like the most cuttable thing on the list. It is not producing a meeting this week, so you cut it and chase the accounts closest to closing. It works for a quarter. Then the pipeline is worse, because the awareness work you cut was what made next quarter's outreach land. You ate the seed corn to make this season's bread. We have done this to ourselves, and the cost showed up later every time.
What is a translation layer?
A translation layer is the instrument that tells each account's story over time, what signals it emitted, what it touched, and how it moved from one state to the next, so that brand work becomes legible in the language the budget responds to. Last-click attribution assigns credit; a translation layer shows cause. It is built on account-level data: every touch tied to a named account that fits your profile, dated, and read in order.
Brand investment that cannot be measured is not a strategy. It is a hope, and hope is a difficult line item to defend. Every marketing leader already believes brand matters; what they lack is the instrument to prove it, and without it brand is cut the moment pipeline gets tight, exactly when cutting it does the most damage.
Which conversations does account-level data change?
Three, and each one is a conversation a marketing leader is having right now. The board conversation stops being a defense and becomes an explanation of a compounding asset. The hard question is not "how is awareness trending?" but "what did brand spend do for pipeline?" Aggregate metrics have no answer. Account-level data does: these forty accounts sat in our awareness orbit for sixty days before they entered pipeline, converted at two and a half times the rate of cold accounts, and brand-touched deals moved through the cycle twenty-five to thirty percent faster. Named accounts sit behind it, and you become someone explaining an investment rather than someone protecting a line item.
The sales conversation becomes an efficiency argument. A rep reaching an account that attended two roundtables, opened a dozen newsletters and visited the pricing page three times this month is not cold-calling. Account-level data, with the timing attached, makes outreach relevant by default.
And the attribution conversation stops being philosophical and becomes one of data architecture. The reason brand could never be tied to revenue is that the touchpoints were never tied to named accounts filtered to the profile you sell to, so the thread was never traceable. That is a plumbing problem, not a belief problem, and plumbing can be fixed.
| Conversation | With last-click | With account-level data |
|---|---|---|
| Board | Defending a line item | Explaining a compounding asset |
| Sales | "Marketing's leads are cold" | Outreach relevant by default |
| Attribution | Philosophy | Data architecture |
What does it look like when it lands?
We watched it change a room once. A founder was ready to cut the slow awareness work for more outbound, because it had never produced a number he could point to. Instead of arguing, we showed him his own deals as timelines: the months each account spent quietly engaging before anyone raised a hand. The conversation flipped from "justify this spend" to "how do we do more of what's working." The work had not changed. The visibility had.
The same move answers the question a board asks after a win: skill or luck? In the book, Northwind's board asked it, and the answer was a timeline, not a story. Sector breach, new CISO hired, first compliance-page visit, pricing visits, a stack-named job post, the In-Market score, the opener to the new CISO, the opportunity. Read top to bottom, the win stopped looking like a lucky reply and started looking like a process that caught a real moment and moved on it. Anyone can claim a deal. Showing the chain of evidence that preceded it is how you prove the system works and justify pointing it at the next account.
What number should replace closed pipeline in the board update?
Not replace: add. Closed pipeline is the easiest number to count, not the only real one. A reply is a real number. A move from never-heard-of-you to actively-engaged is a real number. The count of target accounts that entered your orbit this quarter is a real number, and a better predictor of next quarter's pipeline than this quarter's closes. One buyer told us she wanted to walk into the board meeting with a real number, not a story, and she believed awareness lived in the story column. It does not, once it is instrumented. It is a measurable signal that runs ahead of revenue, and the brand awareness axis of the Demand Compass is what turns it from a feeling into a count.
Related questions
Is multi-touch attribution the fix?
It is better than last-click and still a credit model: it splits the deal among touches instead of handing it to one. A translation layer is not about splitting credit at all. It shows each account's movement in order, which is what a board can read as cause.
Doesn't this require an expensive attribution platform?
It requires that every touch be tied to a named account in one place. That is data architecture more than a product: web visits resolved to companies, event scans, newsletter opens and post reactions landing on the same account record with dates. Many teams already have the pieces in separate tools.
How long before the translation layer shows anything?
One quarter to build the timelines from data you already have, since the touches happened whether or not you recorded them well. The first timelines are backward-looking, drawn from closed deals, and they are usually enough to change the conversation.
What if our brand work really is not producing pipeline?
Then the timelines will show it, and that is useful too. A translation layer is honest in both directions. It defends the awareness work that is working and exposes the awareness work that is not, which is a better position than defending all of it on faith.
