Marketing attribution in B2B: why last click is quietly costing you budget

Last click attribution gives all the credit for a sale to whatever the buyer touched immediately before converting. In a business with a ninety day cycle and a buying group of eight people, that is almost always a branded search or a direct visit, which means your reporting quietly credits the channel that finished the process and starves the ones that started it. The model is not just inaccurate. It changes what you fund.

What does last click actually miss in B2B?

Three things, and they compound.

It misses everyone who never clicks. A significant share of B2B influence happens in places that produce no trackable click at all. A slide seen in a webinar, a video watched on a phone, a colleague forwarding something in a private channel, a conversation at a stand. Those moments change who gets shortlisted, and they arrive at your analytics as direct traffic six weeks later.

It misses everyone who is not the buyer. Gartner's B2B buying research puts a typical complex purchase in the hands of six to ten people. Your content might have persuaded the engineer, the security lead and the operations manager, and none of them will ever be the last click. The last click belongs to whoever happens to be at a keyboard at the moment the form gets filled in.

It misses the time gap. Attribution windows in most ad platforms are measured in days. A ninety day B2B cycle outruns the window entirely, so the first touch has fallen out of the model before the deal even reaches a proposal.

The result is predictable. Branded search, retargeting and direct look extraordinary. Video, thought leadership, events and partner marketing look like waste. Budget moves accordingly, and eighteen months later your branded search volume starts falling because nothing is feeding it, and nobody can explain why.

How much does this actually change the numbers?

Enough to reverse a decision.

Arrow ECS ran a seven week video amplification campaign around VMware during the Broadcom transition, with Joe Baguley, VMware's CTO for Field Sales in EMEA, as the trust anchor. The results:

Read that campaign on last click and you see 323 clicks. Read it on cost per view and you see attention bought roughly 42 times more efficiently than the benchmark, delivered to a specialist technical audience at the exact moment their vendor relationship was in question.

The single Shorts asset, "Containers Need Full Stack", took 169,660 views on its own. County Dublin alone accounted for around 121,700. That is a lot of specific, addressable, in market attention that a click based report would have valued at close to nothing.

Nobody in that audience was going to click an advert and buy enterprise infrastructure. They were going to remember who was talking sensibly during a period when their supplier landscape was moving under them, and act on it later through a channel that would show up in analytics as direct.

Which attribution model should you use instead?

There is no clean answer, which is why this is a strategy decision rather than a settings change. Four workable options, with what each costs you.

Multi touch attribution. Distribute credit across every recorded touchpoint, weighted linearly, by position, or by an algorithm. Better than last click, but it can only ever credit touchpoints it recorded. Everything off platform stays invisible, so it improves accuracy without fixing the underlying blindness.

Marketing mix modelling. Regress business outcomes against spend and external variables at an aggregate level. It sees channels that produce no click, and it does not depend on cookies. It needs two to three years of clean spend and outcome data, and it will not tell you which campaign to stop on Thursday.

Self reported attribution. Ask on the form: how did you hear about us? Crude, biased towards recency and towards whatever is memorable, and consistently more revealing than anything in your analytics. Almost free to implement. If you do one thing after reading this, do this one.

Incrementality testing. Hold out a region or a segment, run the activity everywhere else, compare. The only method that produces a genuine causal answer. It costs you the revenue from the holdout and needs enough volume to read, which rules it out for smaller programmes.

Most B2B businesses of any size end up running self reported attribution and multi touch continuously, with incrementality tests on the two or three channels where the budget is large enough to justify the exercise.

Where should the measurement actually live?

Not inside the platform selling you the media.

Every ad platform reports on its own contribution using its own attribution window and its own definition of a conversion. Run three platforms and you will routinely find they claim more conversions between them than you had. That is not fraud, it is three referees each awarding themselves the goal.

There is a second problem, which is that the history is not yours. Stop paying, and the record goes with it.

On the HMS Networks programme, measurement runs on infrastructure HMS owns rather than a rented platform, using Matomo alongside CRM data. The immediate benefit is one number everyone argues from instead of three that disagree. The longer benefit is that the log accumulates. After two years of it, HMS can see which early signals in their own business actually preceded revenue, and that is not something a competitor can buy.

The commercial effect showed up where it matters. Cost per qualified lead fell from around £760 through trade events to around £72, and the programme returned 35:1 on media. You do not find a saving that size by optimising bids. You find it by being able to see, honestly, which activity was producing qualified people and which was producing volume.

How do you make this survive a board review?

By separating the question your board is actually asking from the one attribution answers.

Your board wants to know whether marketing spend is producing pipeline. Attribution answers a narrower question: which touchpoints preceded a conversion. Presenting the second as an answer to the first is what leads to the awkward silence.

Three things make the conversation work.

Agree one definition of a qualified lead with sales, in writing, before the meeting. Most attribution arguments are definition arguments wearing a costume.

Report on a signal ladder rather than a single number. Attention today, intent within a week, qualification within a month, revenue at 60 to 180 days. Show all four. It makes the lag visible and stops anyone judging a January programme on a February revenue column.

Say what you cannot measure, out loud, first. A report that claims complete attribution in a business with a ninety day cycle and an eight person buying group is not credible, and any finance director worth their salary knows it. Naming the blind spot before it is found buys you more trust than a tidier number ever will.

Where to start this quarter

Add the self reported attribution question to every form, and read it monthly against what your analytics claim. The gap between those two datasets is the most useful marketing document you will produce this year.

Then check what your attribution window is actually set to across each platform, and compare it against your median deal cycle. If your cycle is 120 days and your window is 30, you already know the answer to why video and content never look like they work.

Our conversion and attribution intelligence work starts with that comparison, and our unified marketing dashboards put the result somewhere your board can see it without asking you to rebuild a spreadsheet each month.

Frequently asked questions

Last click attribution assigns all the credit for a conversion to the final touchpoint before it. It is the default in most analytics and advertising platforms. In B2B, that final touchpoint is usually a branded search or direct visit, so the model tends to credit the channel that closed the process rather than the ones that created the demand.

Because B2B buying involves several people over several months, and much of the influence produces no click at all. Attribution windows measured in days cannot see the start of a cycle measured in months, so channels like video, events and thought leadership appear to underperform and lose budget, even when they are the reason the branded search happened.

No single model is complete. Most B2B businesses run self reported attribution on forms alongside multi touch attribution in analytics, then use incrementality tests on the two or three largest channels where the spend justifies a holdout. Marketing mix modelling becomes worthwhile once you have two to three years of clean spend and outcome data.

No. Each platform reports on its own contribution using its own window and definition, which is why three platforms will often claim more conversions between them than you actually had. Measurement held on infrastructure you own gives you one number to argue from, and the history survives when you stop paying.

Report on a ladder of signals rather than a single figure: attention on the day, intent within a week, qualification within a month, and revenue at 60 to 180 days. State the limits of what you can measure before anyone asks. A report that acknowledges its blind spots is more defensible than one that claims complete attribution in a long cycle business.