Marketing Attribution Models Explained: Picking One You Can Defend (2026)

A marketing attribution model is the rule that decides which touchpoints get credit for a conversion. The six standard models are first-touch, last-touch, linear, time-decay, position-based (U-shaped), and data-driven — and for most B2B teams the practical answer is: use a multi-touch model, but worry less about which one and more about whether your touchpoints are captured at all. A perfect model applied to 30% of the real journey loses to a crude model applied to 90% of it. This guide explains each model, when it misleads, and how to pick one you can defend in a pipeline meeting.

What an attribution model actually does

Every closed deal has a history: a LinkedIn ad seen in March, a comparison blog post read in April, a pricing page visit in May, a demo call in June. Attribution models answer one question — when revenue lands, how do we split the credit across that history? The split you choose changes which channels look profitable, which changes where next quarter's budget goes. That's the entire stakes: attribution models are budget-allocation machines wearing an analytics costume.

The six models, honestly assessed

First-touch attribution

100% of credit goes to the first recorded interaction. It answers "what creates demand?" and flatters top-of-funnel channels — content, ads, social.

• Good for: understanding demand creation; young pipelines where awareness is the bottleneck
• Misleads because: the first *recorded* touch is rarely the first *real* touch. A podcast mention that drove a branded Google search gets logged as "organic search"

Last-touch attribution

100% of credit to the final interaction before conversion. The default in most CRMs and, historically, most analytics tools.

• Good for: understanding what closes; short sales cycles
• Misleads because: it systematically over-credits bottom-funnel channels. Branded search and direct traffic "win" constantly because they harvest demand other channels created

Linear attribution

Equal credit to every touchpoint. Five touches, 20% each.

• Good for: a neutral starting point when you have no basis for weighting
• Misleads because: a 30-second blog skim counts the same as an hour-long demo

Time-decay attribution

Credit increases the closer a touch sits to conversion, usually on an exponential curve.

• Good for: long cycles where you want recency to matter without erasing early touches
• Misleads because: it structurally undervalues the demand-creation work that started the journey — the same bias as last-touch, just softened

Position-based (U-shaped and W-shaped)

U-shaped gives 40% to the first touch, 40% to the last, 20% spread across the middle. W-shaped adds a third 30% spike at lead creation.

• Good for: B2B funnels with meaningful stage gates; it matches how marketers actually narrate a journey
• Misleads because: the weights are arbitrary. Nothing in your data says the first touch is worth exactly 40%

Data-driven attribution

An algorithm (GA4 uses a Shapley-value approach) estimates each channel's incremental contribution from your actual conversion paths.

• Good for: high-volume funnels where the algorithm has enough conversions to learn from
• Misleads because: it's a black box you can't interrogate, it needs conversion volume most B2B sites don't have, and in GA4 it only sees what GA4 sees — which excludes your CRM, your sales calls, and most of the dark funnel

Side-by-side

Why B2B attribution breaks in practice

Model choice is the visible 20% of the problem. The invisible 80%:

• The dark funnel. Slack communities, podcast listens, word of mouth, LinkedIn lurking — influential touches that never generate a trackable click. Self-reported attribution (a "how did you hear about us?" field on your demo form) is the cheapest fix and routinely contradicts the click data in useful ways.
• Anonymous research. Around 97% of your website visitors never identify themselves, so their touchpoints attach to nobody. This is where website visitor identification becomes attribution infrastructure: identifying the company (or, on US traffic, the person) behind pre-form visits recovers journey segments that every model otherwise ignores.
• Buying committees. Six people research; one fills the form. Account-level attribution — rolling every identified visit from a company into one journey — matches how B2B actually buys.
• CRM-analytics divorce. GA4 attributes form fills; revenue lives in the CRM. If the two never reconcile, you're attributing leads, not dollars — and lead-optimized budgets buy cheap leads, not revenue. We walk through the fix in how to attribute revenue to marketing channels.

How to choose (a decision path, not a shrug)

1. Under ~50 conversions a month? Skip data-driven — it has nothing to learn from. Use U-shaped for reporting and pair it with self-reported attribution.
2. Sales-led with a CRM? Attribute to closed revenue, not form fills, and use W-shaped or time-decay across the account's touches. VisiLead is building exactly this chain (channel to CRM to closed-won) for its Scale plan ($299/mo); the same logic can be built manually with UTM discipline and CRM reports if you have the ops time.
3. PLG or high-volume? Data-driven models finally have the volume to work; run GA4's data-driven model against a U-shaped baseline and investigate where they disagree.
4. Whatever you pick — keep it stable for at least a quarter. Attribution is a measuring stick; a stick that changes length every month measures nothing. Channel decisions come from trends within one consistent model, not from the model that flatters this month's favorite channel.

Implementation checklist

• UTM-tag every paid and owned link, with an enforced naming convention
• Add "How did you hear about us?" to your demo form and read the answers monthly
• Deploy visitor identification so anonymous research visits attach to accounts
• Reconcile marketing source with CRM closed-won quarterly, and report revenue by channel — not leads by channel
• Sanity-check any model against a channel-level view of campaign effectiveness

We go deeper on the multi-touch mechanics — including the software landscape — in our multi-touch attribution guide.

Worked example: one funnel, three models, three different budget decisions

Abstract model descriptions hide how much the choice matters, so here's a concrete quarter. Imagine 10 closed-won deals worth $200k total, where the typical journey was: LinkedIn ad (first touch) → organic blog visit → webinar → branded search → demo form (last touch).

Now watch the budget meeting: under last-touch, the team doubles the search budget and cuts LinkedIn — and demand quietly dries up two quarters later, because branded search was harvesting demand LinkedIn created. Under first-touch, LinkedIn looks heroic and the webinar program gets cut, even though it was moving stuck deals. The U-shaped column is the only one that would have kept all four channels funded roughly in proportion to their actual role.

The numbers are illustrative, but the mechanism is not: the model you pick is a budget-allocation algorithm. Run your own last-quarter numbers through two models before trusting either — if the rankings agree, great; where they disagree is exactly where you need judgment (or better data) before moving spend.

Frequently Asked Questions

Q: Which attribution model is best for B2B?

A: Position-based (U-shaped or W-shaped) is the most defensible default for sales-led B2B: it credits demand creation and closing while acknowledging the middle. Data-driven models outperform it only when you have the conversion volume (roughly 50+ per month) to train them. The bigger lever is coverage — capturing CRM outcomes and anonymous-visitor touches — not model selection.

Q: What is the difference between first-touch and last-touch attribution?

A: First-touch gives 100% of conversion credit to the earliest recorded interaction (answering "what created this demand?"); last-touch gives it all to the final interaction before converting (answering "what closed it?"). Run both on the same funnel and the channel rankings usually invert — which is the clearest demonstration that single-touch models describe endpoints, not journeys.

Q: What attribution model does Google Analytics 4 use?

A: GA4 defaults to data-driven attribution (a Shapley-value algorithm) for most reports, with paid-and-organic last-click available as an alternative. Note that GA4 attributes GA4-visible conversions only — it cannot see CRM revenue, sales conversations, or visitors it never tagged, so B2B teams should treat it as one input rather than the system of record.

Q: Do I need multi-touch attribution software or can I build it myself?

A: You can get surprisingly far manually: disciplined UTMs, a self-reported attribution field, and a quarterly CRM-to-source reconciliation. Dedicated software earns its keep when journeys span months and committees — tools range from VisiLead ($29/mo multi-channel attribution; closed-revenue attribution in development) to Factors.ai ($199/mo) up to enterprise platforms. Start manual, buy when the spreadsheet becomes the bottleneck.

Q: How many touchpoints does a B2B buyer have before converting?

A: Research and vendor studies consistently land in the dozens of touches across multiple people — but the honest answer for your business comes from your own data: identify visitors, roll journeys up to the account, and count. Most teams that do this find far more early-journey touches than their CRM recorded, which is precisely why last-touch models misallocate budget so badly in B2B.

Q: Should marketing and sales touches both count in attribution?

A: For revenue attribution, yes — a journey that ends in closed-won ran through SDR emails, demos, and follow-ups as well as campaigns, and models that ignore sales touches systematically over-credit whatever marketing channel happened to sit closest to the handoff. The pragmatic version: include sales activities as touchpoints in the account journey, but report marketing-sourced and marketing-influenced revenue as separate lines so neither team feels robbed.

Q: How often should I review my attribution model?

A: Review outputs monthly, revisit the model itself annually — or when something structural changes (new motion, new market, a channel you couldn't previously measure). The failure mode to avoid is re-picking the model whenever a stakeholder dislikes a number; within-model trends are only meaningful if the model holds still. Date-stamp the model choice and treat changes like schema migrations: deliberate, documented, rare.