The short version
Last-touch attribution gives all the credit for a deal to the final tracked interaction before conversion. It's simple, but in B2B it mostly rewards the channels that capture demand, not the ones that create it.
Switching to a multi-touch model doesn't fix the problem. Every model only shares out credit among the touches your tools recorded, and most of what moves a B2B buyer is never recorded.
Buyers decide early. In 6sense's 2025 study of more than 4,000 buyers, 94% of buying groups had ranked preferred vendors before first contact with a seller.
A better approach combines a simple tracked model, a self-reported source field, four pipeline metrics and the occasional incrementality test. None of it needs a data team.
Every month, a founder opens the attribution report and sees the same story. Branded search and direct traffic closed most of the deals. LinkedIn, the podcast and the webinars look like they did almost nothing. So the budget drifts toward search, and six months later pipeline is thinner and nobody can say why.
The report wasn't lying on purpose. It was answering a narrower question than the one being asked. This guide explains what last-touch attribution actually measures, why multi-touch reports are wrong in B2B for the same reasons, which model misleads least, the signal most firms ignore, the four metrics that predict revenue whatever model you use and the measurement stack we run for clients.
What last-touch attribution actually measures
Last-touch attribution credits 100% of a conversion to the last tracked interaction before it happened. If a buyer heard you on a podcast, read three LinkedIn posts, attended a webinar and then searched your brand name and booked a demo, branded search gets the whole deal.
It's still the most common model in practice, partly because it's the easiest to explain and partly because the platforms have narrowed the choice. In 2023 Google retired the first click, linear, time decay and position-based models across Google Ads and GA4, moving advertisers to data-driven attribution and leaving "the last click model" as the alternative (Google Ads Help).
Last-touch answers one question well: which channel was present when an already-interested buyer raised their hand. It answers a different question badly: where that interest came from in the first place.
Why every B2B attribution report is wrong
The problem isn't only last-touch. Every attribution report in B2B, however sophisticated, has the same blind spots.
Buyers decide before they click. 6sense's 2025 Buyer Experience Report found that 94% of buying groups ranked preferred vendors before first contact, and bought from that early favourite 77% of the time (6sense via Business Wire). By the time a tracked conversion happens, the decision is largely made.
Most of the market isn't buying yet. Research with Professor John Dawes at the Ehrenberg-Bass Institute suggests only around 5% of B2B buyers are in-market at any time (Ehrenberg-Bass Institute). Marketing aimed at the other 95% pays off months later, long after any click window has closed.
The biggest influences leave no trail. Podcasts, peer recommendations, private Slack groups, LinkedIn posts read without clicking and AI assistant answers all shape demand. This dark social activity is invisible to tracking.
Committees, not individuals, buy. The person who clicked the ad is often not the person who signed. Tracking follows browsers, not companies.
Platforms report on themselves. Google, LinkedIn and Meta each claim credit for the same deal using their own rules, which is why the platform totals add up to more conversions than you actually had.
Why multi-touch attribution doesn't fix it
The usual answer to "last-touch is too simple" is to buy or build a multi-touch attribution model. It feels more rigorous, and the reports look more balanced. But a multi-touch model only redistributes credit among the touches it can see. If the podcast, the peer recommendation and the AI answer were never recorded, no weighting scheme will give them credit.
Multi-touch also adds false precision. A report saying LinkedIn drove 23.4% of revenue looks exact, but the number depends on which model you chose, how long the lookback window is and how many touches were lost along the way. Change any one of those and the figure moves. Data-driven models go further, but they need large conversion volumes to work and don't show their reasoning, which makes them hard to trust for a firm closing 20 deals a quarter.
None of this means multi-touch is useless. It's a good way to see which tracked channels tend to appear together. It just shouldn't be treated as the answer to where revenue comes from.
Last-touch vs first-touch vs weighted: which misleads least
Model | What it over-credits | What it misses | Useful for |
|---|---|---|---|
Last-touch | Branded search, direct traffic, retargeting and sales emails | Everything that created the interest | Seeing which channels convert existing intent |
Whatever was tracked first, often organic search or a paid ad | Untracked first exposures and everything after the first visit | Seeing which channels bring new people in | |
Weighted (U-shaped, W-shaped, time decay) | Tracked touches near the milestones the model favours | Untracked touches, plus it hides assumptions inside the weights | Comparing tracked channels that work together |
Data-driven | Channels with lots of tracked conversion data | Low-volume and offline activity; reasoning is hidden | High-volume paid programmes |
For most B2B firms between $1M and $20M in revenue, our view is simple. Use first-touch to see where new demand enters and last-touch to see what converts it, read them side by side and don't spend months debating weights. Our guide to attribution models explained covers every model in detail, and our guide to choosing a B2B attribution model matches models to sales cycles.
// The Rule
Attribution shows where credit was recorded, not where demand was created. Use it to compare channels, never as the final word on budget.
Self-reported attribution: the signal most firms ignore
The cheapest fix for the blind spots is to ask. Add a required, open-text "how did you hear about us?" field to your demo and contact forms, and have sales confirm the answer on the first call.
Self-reported attribution picks up what tracking can't: the podcast episode, the founder's LinkedIn post, the colleague who recommended you, the AI assistant that listed you. It's imperfect, because buyers forget and simplify, but its errors are different from tracking's errors. That's what makes it useful.
Keep it open text. A dropdown pushes people toward "Google" because it's the first sensible option.
Code the answers monthly into a short list: podcast, LinkedIn, referral, search, AI assistant, event, other.
Compare with tracked source. When a deal shows "direct" in the CRM but the buyer wrote "your podcast", the podcast is doing work your reports can't see.
Across the B2B firms we work with, the gap between the two is usually largest for founder-led LinkedIn, podcasts and referrals. Those are exactly the channels last-touch reports tell you to cut.
The four metrics that predict revenue
Whatever attribution model you pick, these four numbers tell you whether marketing is working. They don't depend on getting credit right, so they're much harder to argue with.
Metric | What it tells you | Watch for |
|---|---|---|
Qualified pipeline created | The value of new sales opportunities accepted by sales each month | A rolling three-month trend, not single months |
Whether the pipeline is the right kind of buyer | Win rate falling as volume rises, a sign of weaker fit | |
How warm buyers are when they arrive | Shorter cycles for deals with a strong self-reported source | |
What each new customer costs across all sales and marketing spend | Blended CAC, not per-channel CAC built on attribution |
Together they make up pipeline velocity. If pipeline, win rate and cycle length all improve after you invest in a channel, that channel is working, whatever the attribution report says.
Incrementality testing: the closest thing to proof
Attribution tells you who was present. Incrementality testing tells you what would have happened without a channel. You switch a channel off, or hold back a region or audience segment, and compare pipeline against a control group over a set period.
It's the only method that measures cause rather than correlation, so it's the right tool for big budget questions: "Would we lose pipeline if we cut LinkedIn ads by half?" It's slower and needs discipline, so run one or two tests a year on your largest channels rather than testing everything. Our guide to incrementality testing shows how to design one at B2B volumes.
The mrktrs measurement stack
This is the setup we build for B2B services and SaaS firms between $1M and $20M in revenue. Each layer covers a blind spot in the one before it.
Layer | What it does | Review |
|---|---|---|
1. Clean capture | Consistent UTM parameters and hidden form fields so every lead arrives with a source | Set once, audited quarterly |
2. CRM as the record of truth | Original source on every contact and deal, with contacts linked to opportunities | Monthly data quality check |
3. First-touch and last-touch side by side | Shows where demand enters and what converts it | Monthly, on rolling quarters |
4. Self-reported source | Catches dark social, referrals, podcasts and AI assistants | Monthly, compared with tracked source |
5. Four pipeline metrics | Pipeline created, win rate, cycle length and CAC | Monthly leadership review |
6. Incrementality tests | Confirms whether the biggest channels cause new pipeline | One or two a year |
If you run HubSpot, our guide to HubSpot attribution shows how to set up layers one to four without a data team. For the data foundation underneath, see our guide to first-party data attribution, and for the errors that undermine all of this, our guide to common attribution mistakes.
Is your attribution report steering budget away from the channels that create demand? Book a 30-minute call and we'll show you what your current reports are missing.
Frequently asked questions
What is last-touch attribution?
It is a model that assigns all conversion credit to the final recorded interaction before a lead or sale, such as a branded search, a retargeting ad or a direct visit, and ignores every earlier interaction.
What is the main weakness of last-touch attribution?
It favours channels that harvest existing interest and undervalues those that generate it. In B2B, where buyers form preferences long before contacting a vendor, this can lead teams to cut the activity that fills future pipeline.
Is last-touch or first-touch attribution better for B2B?
Neither is complete. First-touch is more useful for judging how new buyers discover a company, while last-touch is more useful for judging conversion. Reading both together gives a fairer picture than either alone.
Does Google Ads still offer last-click attribution?
Yes. Last click remains available in both Google Ads and GA4, although data-driven is now the default and the other rules-based options were retired in 2023.
How accurate is self-reported attribution?
It is imprecise at the level of individual answers, since buyers sometimes misremember, but reliable in aggregate for spotting channels that tracking cannot see, such as word of mouth, podcasts and private communities.