First-Touch vs Last-Touch Attribution

Your first-touch report says content created the customer. Your last-touch report says branded search closed the sale. Neither number is necessarily wrong. Each model answers a different question—and becomes dangerous when you treat its answer as the whole customer journey.

First-touch attribution gives 100% of the credit to the first known interaction. Last-touch attribution gives 100% to the final known interaction before conversion. First touch is an acquisition lens; last touch is a conversion lens. The practical answer is not to declare one the winner. It is to compare both, quantify their disagreement, and verify important budget decisions against revenue.

This guide adds something the leading comparisons do not: an Attribution Gap Score that turns first-versus-last disagreement into a repeatable channel diagnostic. It also adds the missing third dimension—whether the on-site variant between those touches increased revenue. That is where attribution becomes an optimization system rather than a reporting argument.

First-touch vs last-touch attribution at a glance

Question	First touch	Last touch
Who receives credit?	The first identifiable source or campaign	The final identifiable source or campaign before conversion
Best question	“Where did customers discover us?”	“What brought customers back to convert?”
Usually favors	SEO, social, partnerships, display	Email, retargeting, branded search, direct response
Usually ignores	Nurture and closing interactions	Demand creation and early education
Best use	Top-of-funnel diagnosis	Conversion-path diagnosis

Suppose a buyer discovers you through a founder’s LinkedIn post, later reads an organic comparison article, receives an email, clicks a branded Google result, and buys a $500 annual plan. First touch assigns $500 to LinkedIn. Last touch assigns $500 to branded search. The article and email get nothing under either model.

That simplification matters because journeys are rarely one click long. A July 2026 attribution guide, summarizing Gartner’s 2025 UK Digital Marketing Survey, reports that only 24% of UK B2B organizations use multi-touch attribution, while typical B2B journeys involve six to eight touches. Treat that as a warning about model limits, not as permission to buy the most complicated attribution software available.

What first-touch attribution gets right—and wrong

First touch is valuable when the decision is about discovery. It can show whether a podcast sponsorship, non-branded search campaign, affiliate, or community is introducing future customers. It is simple enough to audit: preserve the original referrer and UTM values, associate them with the visitor or lead, then connect the eventual payment.

Its bias is equally clear. The first measurable click may be incidental, months old, or unrelated to the purchase. It also misses dark social, word of mouth, and interactions that occurred on another device. Most importantly, it tells you where a journey began—not whether increasing spend there will create incremental revenue.

Use first touch to evaluate reach into qualified audiences, compare acquisition cohorts, and protect channels that create demand. Do not use it alone to judge nurture emails, checkout campaigns, or retargeting.

What last-touch attribution gets right—and wrong

Last touch is useful for examining the final observable step. It identifies which emails, landing pages, offers, or campaigns are present when intent becomes action. For short buying cycles that happen in one session, the difference between first and last may be negligible.

But proximity is not causality. Branded search often acts as navigation for a buyer who already decided. Retargeting can harvest demand another channel created. “Direct” may merely mean the earlier source was lost. Cutting early-stage content because it has weak last-touch revenue can destroy the demand that closing channels capture.

GA4 also deserves a precise note. Google’s current attribution documentation lists three report models: data-driven, paid and organic last click, and Google paid channels last click. Google removed first-click, linear, time-decay, and position-based models in November 2023. So a team that wants a durable first-touch field must often preserve it in its own visitor or customer record rather than assume the default analytics report will.

The original Attribution Gap Score

Side-by-side reports are useful, but “these numbers look different” is not an operating rule. Calculate a normalized score for every channel:

Attribution Gap Score = (first-touch revenue − last-touch revenue)
                        ÷ (first-touch revenue + last-touch revenue)

The score ranges from -1 to +1. A positive score means the channel leans toward discovery; a negative score means it leans toward closing. A score near zero means the two lenses agree—or the channel participates similarly at both ends.

• +0.40 to +1.00: demand creator. Judge it on qualified first-time visitors, first-touch customer acquisition cost, cohort revenue, and assisted journeys. Do not cut it using last-touch ROAS alone.
• -0.39 to +0.39: balanced or inconclusive. Inspect volume and middle touches. Agreement is more persuasive when both models assign substantial revenue, not when both assign almost none.
• -1.00 to -0.40: demand closer. Measure conversion efficiency, but test incrementality before scaling. The channel may be capturing buyers who would have converted anyway.

Example: organic content receives $30,000 in first-touch revenue and $10,000 in last-touch revenue. Its score is (30,000 − 10,000) ÷ 40,000 = +0.50, a demand-creation signal. Branded search at $5,000 first touch and $25,000 last touch scores -0.67, a closing signal. That does not prove content caused $30,000 or search caused $25,000. It tells you what to investigate and which KPI is least unfair.

A decision rule for choosing your primary view

1. If purchases usually happen in one session, last touch is a workable primary operational view; retain first touch as a check.
2. If you are testing awareness channels, lead with first touch and compare cohort revenue after a fixed window.
3. If the journey spans sessions or stakeholders, report both and use the Gap Score. Add a multi-touch view only after identity, UTMs, and revenue events are reliable. See the multi-touch attribution guide for SaaS for model selection.
4. If a budget decision is large, run a geo, audience, or time-based holdout where feasible. Attribution describes paths; experiments estimate causality.
5. If the “conversion” is only a signup, fix that first. Connect the eventual payment using the workflow in this SaaS revenue attribution guide.

The missing third axis: which page variant made the money?

Channel attribution can tell you that a customer arrived first through SEO and last through email. It cannot tell you whether the pricing-page headline, proof block, or checkout treatment changed the outcome. That distinction changes the action. A campaign may look stronger because it sent visitors to a better experience; a page variant may work for one source and fail for another.

This is the gap between analytics, experimentation, and revenue attribution. GA4 measures journeys and offers attribution reports. Plausible can attach revenue to custom events and filter it by sources, campaigns, and entry pages. Matomo provides a broader self-hosted analytics route. DataFast focuses on connecting payment data to traffic sources and customer journeys. Those are credible options depending on privacy, data ownership, and reporting needs.

PageDuel is designed around a different closed loop in one snippet: measure traffic, test a page change, and prove revenue by source, campaign, and variant. Instead of asking only whether first or last touch gets credit, you can ask which experience earned more revenue within each acquisition cohort.

Implementation checklist

• Define one revenue event from a confirmed payment, not a checkout click.
• Store original source, medium, campaign, landing page, and timestamp as immutable first-touch fields.
• Update separate last-touch fields on qualified non-direct visits.
• Keep payment processors from overwriting the genuine referral source.
• Join visitor, signup, customer, and payment identifiers.
• Preserve experiment ID and variant alongside both attribution records.
• Use the same attribution window for every channel in a comparison.
• Calculate the Gap Score only after checking missing UTMs, duplicate purchases, cross-domain checkout, and consent loss.
• Review first touch, last touch, variant revenue, refunds, and customer quality together.

A useful companion is the UTM revenue attribution guide, which explains how to keep campaign identity intact from click through payment.

Which model should you use?

Use first touch to understand where customers begin. Use last touch to understand where they finish. Use the gap between them to classify each channel’s role. Then validate major decisions with experiments and actual revenue rather than pretending a credit rule proves causation.

The contrarian conclusion is simple: choosing one “correct” attribution model is the wrong objective. Build a measurement system that makes each model’s bias visible—and connects both to the page experience customers actually saw. Start your 14-day free trial to measure traffic, test changes, and prove which sources, campaigns, and variants make revenue. No credit card required.

Sources

• Google Analytics: Get started with attribution
• Marketing Mary: Marketing Attribution Models 2026
• MarketingOps: 2025 State of Marketing Attribution
• Plausible: Ecommerce revenue and attribution tracking

Related Reading

• Revenue Attribution for SaaS
• Multi-Touch Attribution for SaaS
• UTM Revenue Attribution
• The Best Google Analytics 4 Alternative