Strategy
Business-to-business marketing attribution models compared: first-touch, last-touch and multi-touch
B2B marketing attribution models compared: first-touch, last-touch and multi-touch, with the criteria that decide which one fits a US sales cycle.
What to take away
- First-touch credits the campaign that created the contact. Last-touch credits the final touch before the deal closes. Multi-touch divides credit across several.
- The model you pick decides which channel gets the next budget cycle, so it is a spending decision before it is a reporting one.
- Multi-touch suits long US B2B cycles with six or more stakeholders, but it depends on clean identity data across systems.
- None of the three proves that marketing caused revenue; each measures correlation inside systems you control.
What is being compared
Attribution answers a single question: which marketing touch gets credit for a closed deal. First-touch gives all of it to the earliest recorded interaction. Last-touch gives all of it to the final one before the opportunity closes. Multi-touch divides credit across a set of touches, using weighting rules your team writes.
The three are not rival philosophies so much as different guesses about where influence sits. Most teams inherit their model from whatever sales funnel stages their reporting already tracks.
The criteria that matter
Compare the models on five tests before you shortlist software. The five are cost to run and maintain, availability of clean identity data across systems, length of the sales cycle, size of the buying group, and whether the output survives a finance review.
Three attribution models compared
First-touch
- Credit goes to
- Earliest tracked touch
- Data required
- One source field
- Six to twelve month cycle
- Weak
- Illustrative annual cost
- Under $1,000
- Common failure mode
- Overpays top-of-funnel
Last-touch
- Credit goes to
- Final touch before close
- Data required
- One opportunity field
- Six to twelve month cycle
- Moderate
- Illustrative annual cost
- Under $5,000
- Common failure mode
- Overpays search and retargeting
Multi-touch
- Credit goes to
- Defined set of touches
- Data required
- Stitched identity across systems
- Six to twelve month cycle
- Strongest
- Illustrative annual cost
- $25,000 to $180,000
- Common failure mode
- Weighting rules become negotiation
First-touch
- Cost to run and maintain
- Lowest. One field to read.
- Clean identity data
- Barely needed.
- Sales cycle length
- Short, under a month.
- Buying group size
- One or two people.
- Survives a finance review
- Simple, but easy to attack.
Last-touch
- Cost to run and maintain
- Low. Usually already stored on the opportunity.
- Clean identity data
- Barely needed.
- Sales cycle length
- Short to medium.
- Buying group size
- One or two people.
- Survives a finance review
- Simple, but flatters closing channels.
Multi-touch
- Cost to run and maintain
- Highest. Needs stitching across systems.
- Clean identity data
- Essential. One account must resolve to one record.
- Sales cycle length
- Long, six months or more.
- Buying group size
- Six or more.
- Survives a finance review
- Defensible only if the model is documented.
The table is a starting point, not a verdict. Cost matters less than whether the data behind each row survives a finance review, and the counting problems are set out in this guide to B2B marketing analytics.
First-touch and last-touch in practice
First-touch is cheap and simple. You read one field and you are finished. It suits a US team selling a low-priced product on a short cycle, where one channel creates most of the pipeline.
Last-touch is the default in most CRMs because the opportunity record already stores it. It flatters paid search, retargeting and any landing page that sits beside a form fill. Teams that sell through a demo request often find last-touch matches how the sales floor talks about a deal.
Neither model sees the four research visits that happened before anyone filled in a form. Both depend on stage definitions, and the ones that survive scrutiny are described in these sales funnels that hold up.
Multi-touch attribution for long US sales cycles
Multi-touch exists because long cycles break the simple models. A US enterprise deal can run six to twelve months and involve a buying group of six to ten people, each arriving through a different door.
Linear, position-based, U-shaped and data-driven weighting are the common variants. Position-based gives 40 percent to first touch and 40 percent to last. Data-driven models fit weights to your closed-won history, so they describe what your team did before rather than what will work next.
The stitching work is the real cost. A typical mid-market build takes four to eight weeks of engineering time to join CRM, marketing automation and product data, and it needs a named owner after launch.
All of them need identity resolution, and that stitching raises privacy obligations. California's consumer privacy law covers business contact data, and the state's removal of the B2B carve-out took effect in 2023, so opt-out and deletion requests reach marketing records. The NIST Privacy Framework offers a structure for assessing that risk.
Where each one wins
Choose on the shape of your cycle, not on which model sounds most advanced.
Which model fits your cycle
Does one channel create nearly all new contacts?
First-touch
Does a sales team close on a demo request?
- First-touch is right when one channel creates nearly all new contacts and the cycle closes inside a month. A self-serve product at a low monthly price fits.
- Last-touch is right when a sales team closes on a demo request or a pricing page visit, and marketing wants credit for feeding that motion. It is the easiest model to defend to a CFO.
- Multi-touch is right when deals run for months, several departments sign off, and the team can fund the data plumbing. Enterprise software and capital equipment are the usual cases.
Account-based programs complicate the choice, because the unit of analysis is an account rather than a lead; the selection and measurement side is covered in account-based marketing.
What none of them solve
Every model shares one blind spot. They only see touches recorded in your own systems. Word of mouth, a peer recommendation, an analyst report and a hallway conversation at a conference are invisible to all three.
Attribution software reports which touches were logged, not which ones moved the decision.
Multi-touch does not repair bad data. If two systems disagree about an account name, more touchpoints produce more noise. Teams that skip the stitching step get a longer report with the same gaps. Matching a funnel to the buying group is the harder exercise, and it is where B2B marketing automation platforms tend to overpromise.
