Quick answer
First-touch attribution assigns all credit for a conversion to the earliest eligible recorded marketing interaction in the lookback path. Last-touch attribution assigns all credit to the latest eligible interaction before conversion, often with special rules for direct visits. Both are deterministic bookkeeping models, not causal estimates. First touch highlights observable acquisition entry points but favors channels visible early and inside the chosen window. Last touch is simple for operational reporting but favors closers such as branded search, affiliates and retargeting and ignores earlier demand creation. Results also depend on identity matching, channel rules, lookback windows, consent, offline touchpoints and direct-traffic treatment. Use a single-touch view for narrow source hygiene or a stable reporting convention, show first and last views together when useful, and never move large budgets on the assumption that credited conversions equal incremental conversions. Calibrate decisions with experiments, marketing mix modeling and customer research.
What are first-touch and last-touch models?
First-touch and last-touch attribution are rule-based single-touch models. First touch assigns one hundred percent of conversion credit to the earliest eligible observed interaction. Last touch assigns one hundred percent to the latest eligible interaction before the conversion.
The word eligible matters. Analytics tools define which channels, clicks, impressions, direct visits and time windows can receive credit. Two systems can use the same model name and produce different reports because their observed paths and rules differ.
Credit is an accounting convention. It describes how the system labels an observed conversion, not how much of that conversion would disappear if the credited channel were removed.
What first-touch attribution reveals
First-touch reporting highlights the earliest visible source in a path and can support acquisition-source hygiene. A B2B team might use it to see which tracked campaigns first create known leads, while a publisher might compare where new registered readers enter.
It structurally favors channels that are visible early and inside the lookback window. If prior brand exposure, word of mouth, television or untagged content is missing, the first recorded touch may be only the first event the system can see.
It also freezes later work out of the credit story. Nurture, sales, product experience and closing activity can be essential even though they receive zero. Use the model for its narrow opening-source question rather than treating the opening as sole cause.
What last-touch attribution reveals
Last-touch reporting is easy to implement and explain. It often supports campaign trafficking, source reconciliation and short-cycle optimization because the final eligible event sits close to the conversion record.
The same proximity creates closer bias. Branded search, coupon sites, affiliates, retargeting and direct-response messages can capture credit after brand, content or offline activity created demand. A person already intending to buy can click a final ad without that ad causing the purchase.
Rules for direct traffic change the result. Some tools ignore direct when an earlier non-direct touch exists; others distinguish paid-channel last click from paid-and-organic last click. Always name the exact rule rather than reporting last touch as a universal standard.
Compare the models as deliberately biased lenses
First touch answers which eligible recorded interaction opened the observed path. Last touch answers which eligible recorded interaction closed it. Both deliberately discard the rest of the path, so neither is a complete journey model.
Showing both can reveal structural sensitivity. If a channel dominates first touch but nearly vanishes at last touch, it may introduce paths that other activity closes. If it dominates last touch, it may harvest demand, resolve final friction or simply be easier to observe near purchase.
The comparison generates hypotheses, not causal coefficients. Investigate with path analysis, cohorts, experiments and qualitative research before changing strategy.
Decision
State the narrow reporting or operational question before selecting a credit rule.
- Is the need source hygiene, journey description or causal allocation?
- What action could this report support?
Path boundary
Define eligible touchpoints, lookback, identity, channels and treatment of direct and offline activity.
- What can the system actually observe?
- Which interactions are excluded by design?
Credit rule
Apply first or last touch consistently and preserve the raw path for comparison.
- Which event gets 100 percent?
- What happens when no eligible marketing touch exists?
Bias review
Identify which channels the chosen endpoint structurally favors and what it cannot claim.
- Which sources sit near this endpoint?
- Which demand creation is invisible?
Triangulate
Use alternative views and causal evidence before making material allocation decisions.
- Do first and last views disagree materially?
- What do lift tests or MMM indicate?
Attribution boundaries determine the winner
The lookback window decides how far the system searches. Short windows favor recent closers and miss long consideration. Long windows attach more earlier events but increase overlap, identity loss and the chance that unrelated activity enters the path.
Identity stitching determines whether devices and sessions become one journey. Login, consent and customer matching improve some links but do not observe every person or channel. The matched population can differ from everyone who converts.
Channel taxonomy, URL tagging, referral exclusions, time zones, conversion deduplication and offline imports also shape results. Treat these as governed model inputs and version changes so a reporting shift is not mistaken for a marketing effect.
Platform availability changes over time
Model menus are product decisions, not a canon of valid science. Google Ads currently supports last-click and data-driven attribution for relevant conversion actions and notes that first-click, linear, time-decay and position-based models are no longer supported there.
Google Analytics currently offers data-driven, paid-and-organic last click and Google-paid-channels last click in attribution reports, with stated direct-traffic rules. Other tools preserve first touch as a custom or CRM field even when their ad platform does not.
Document platform, property, conversion action, model and date. Never assume an old tutorial describes current behaviour, and do not overwrite historic model versions without explaining the break.
First-touch and last-touch example
The training provider makes observability explicit. Its first touch begins at the first eligible recorded interaction, not necessarily the learner's first awareness, while the last touch can reflect a final branded query or adviser call.
Both reports remain useful for operations because definitions are stable and linked to enrolment quality. Budget allocation relies on evidence that can address the counterfactual, not on choosing the preferred endpoint story.
A hypothetical provider of professional finance training has a long enrolment path that can include an industry podcast, downloadable checklist, webinar, branded search, adviser call and online payment.
The team defines enrolment after the refund window as the conversion. Its analytics can reliably observe tagged web interactions and CRM calls after a known lead is created, but not every podcast exposure or colleague recommendation.
The first eligible recorded interaction receives all credit. This report is useful for checking which trackable sources introduce newly known leads, while acknowledging that invisible awareness may precede them.
The final eligible marketing interaction before payment receives all credit under a documented direct-traffic rule. Branded search and adviser calls gain more reported enrolments because they often occur near purchase.
The team connects both source views to attendance, refund and later qualification outcomes. It does not conclude that the credited endpoint alone caused enrolment.
Geo tests, controlled campaign pauses, aggregate modeling and enrollee research inform material budget changes. Single-touch reports remain a stable operational lens rather than the financial truth.
This example is hypothetical. Available attribution models and channel rules vary across analytics and advertising products and can change over time.
Why single-touch credit is not incrementality
Causal impact asks how outcomes would change without the channel or treatment. First- and last-touch rules do not create that comparison. They redistribute observed credit even when every conversion would have happened anyway.
Self-selection is central: high-intent people are more likely to search a brand, revisit a site, open a sales email and convert. Their touchpoints predict conversion partly because intent caused both the interaction and the purchase.
Use randomized holdouts or geo experiments for causal lift where feasible. Marketing mix modeling can inform aggregate allocation under explicit assumptions, while customer research explains paths missing from digital logs. Attribution remains a frequent operational signal, not a substitute for these methods.
Use single-touch models for bounded jobs
First touch can support lead-source completeness, tagging audits and early-source cohort reporting. Last touch can support conversion-path reconciliation, partner operations and an easily understood baseline for model comparison.
Keep the conversion, window, channel rule and identity boundary stable enough for trend interpretation. Show unattributed and unknown shares rather than forcing every outcome into a named channel. Connect source cohorts to downstream value and quality.
For material budget changes, require corroboration. A model-sensitive channel, disputed tracking rule or large gap from lift evidence should trigger investigation and a smaller reversible action rather than an automatic reallocation.
Limitations and common mistakes
Single-touch models ignore interaction among touchpoints and overstate whichever endpoint they select. They underrepresent offline influence, word of mouth, untracked views, cross-device paths and activity outside the chosen platform.
Common mistakes include calling first recorded touch first awareness, calling last touch the channel that caused the sale, comparing tools with different windows, ignoring direct rules, mixing lead and revenue conversions and optimizing toward credited volume without customer quality.
Simple does not mean useless. A transparent biased rule can be more reliable for operations than an opaque model used beyond its evidence. The discipline is to match the tool to the question and label what it cannot answer.
First touch names the visible opener and last touch names the visible closer. Neither endpoint owns the customer journey.
Frequently asked questions
What is first-touch attribution?
It assigns all conversion credit to the earliest eligible interaction recorded within the model's defined path and lookback window.
What is last-touch attribution?
It assigns all credit to the latest eligible interaction before conversion, subject to the tool's rules for direct traffic and channel eligibility.
Which model is more accurate?
Neither is generally accurate for causal impact. Each is a transparent endpoint rule useful for different bounded reporting jobs.
Why does last touch favor branded search and retargeting?
Those interactions often occur near purchase among high-intent people, so an endpoint rule gives them all credit regardless of earlier demand creation.
Can first or last touch guide budget allocation?
Use them as diagnostics or reporting conventions, then corroborate material allocation with experiments, MMM, customer quality and research.
Sources and further reading
- Google Ads Help: About Attribution Models ↗Current official definitions, model availability and last-click behavior in Google Ads
- Google Analytics Help: Get Started with Attribution ↗Current official attribution models, direct-traffic rules and data-driven context in Analytics
- Nielsen: Methods and Models, A Guide to Multi-Touch Attribution ↗Industry definitions and comparison of first-touch, last-touch and fractional models
- IAB: Multi-Touch Attribution Implementation and Evaluation Primer ↗Industry guidance on limitations of first and last touch and advanced attribution implementation