B2B Marketing Attribution: Why the B2C Playbook Breaks

What Breaks the Standard Attribution Playbook in B2B?

Attribution as a discipline grew up in ecommerce, where someone clicks an ad and buys a thing the same afternoon. Almost every assumption in that setting fails in B2B.

Why Do Lookback Windows Distort B2B Reporting?

If the average cycle runs two quarters and the attribution window is 90 days, the touches that created the opportunity fall outside the window entirely.

What remains inside the window are late-stage touches: the demo request, the pricing page, the final email. Those get the credit. They are also the touches closest to a decision that had largely been made already.

The result is a report that consistently over-credits capture and under-credits creation, and a budget that drifts toward the bottom of the funnel year after year because that is where the measured return appears.

Why Does Person-Level Credit Describe the Wrong Thing?

Revenue lands on an account. The decision is made by a group. Crediting the individual who converted describes one person journey rather than the buying decision.

The consequence is practical rather than philosophical. The person who fills in a form is disproportionately likely to be the researcher rather than the economic buyer, and they arrive through the channels researchers use. Person-level credit therefore over-weights research channels and under-weights whatever reached the executive who approved the spend.

Account-level credit rolls every touch across every contact at the account into one journey, then splits the won deal across that journey. See multi-touch attribution for how the weighting works once the journey is assembled at the account level.

How Bad Is the Offline Problem?

The channels hardest to instrument are events and webinars, and they are the ones that quietly lose budget because they do not track cleanly.

This is a compounding error. A model that only counts what is easy to count under-credits events, so events get less budget, so events produce less, which the model reads as confirmation. Two or three planning cycles later the company has stopped doing the thing that reached its buying committees, and no report ever flagged it.

The fix is unglamorous: get offline touches into the same model as paid, with their cost attached, even when the capture is imperfect. A rough number in the model beats an accurate number outside it.

What Still Works in B2B Attribution?

Not everything transfers badly. Three things hold up in B2B:

Credited revenue next to real media cost, per channel, gives marketing and finance a shared vocabulary that anecdote cannot.

Consistency period over period matters more than picking the correct model, because the comparison is what drives decisions.

And algorithmic attribution handles B2B better than fixed-weight models, because it learns the weights from what correlated with closing rather than encoding a guess about which stage matters.

What Is the Limit Worth Naming?

With a small number of large deals, attribution gets thin. Both attribution and mix modeling need enough closed deals to see a pattern, and a company closing twenty deals a year does not have the sample for stable weights.

That is a real constraint rather than a reason to buy a different tool. In that situation, qualitative judgment about specific accounts carries more weight than any model, and the honest use of attribution is directional rather than decisive.

Where Does Credit Stop?

Even a well-built B2B attribution model answers what happened. It does not answer where the next dollar should go.

A channel can look excellent precisely because it is saturated. Strong credited revenue means it absorbed the demand available to it, which is also why its next dollar will return less. The report shows the strong past and says nothing about the weak next dollar.

Getting from credit to a decision means modeling how each channel returns as spend increases, then moving money to the marginal dollar. See marketing budget allocation.

The Numbers Behind the Lookback Problem

The lookback argument is easy to state and easy to dismiss without figures, so here are the figures.

The average B2B sales cycle runs 84 days, and a study of 939 companies found cycles have lengthened 22 percent since 2022. A 90-day attribution window covered the average journey comfortably a few years ago. On the current average it barely does, and for any company above the average it does not.

Enterprise motions sit well above that average, which is why enterprise teams see the distortion first and hardest.

The offline penalty shows up in the conversion data too, though not in the direction people expect. Events convert to SQL at 4.2 percent against a 15 percent median, well below website leads at 31.3 percent. Read naively, that says to cut events.

Read carefully, it says something narrower: events produce SQLs at a low rate per MQL. It says nothing about the size of the deals they produce, how fast those deals close, or how many committee members an event reaches at once. A channel can convert poorly at the MQL stage and still carry the highest return in the mix, and a model that only sees the conversion rate will keep recommending against it.

That is the offline problem stated precisely. The metric that tracks cleanly is not the metric that decides the budget.