Ad attribution models: Types, examples & how to choose

Two advertising platforms can observe the same customer journey and report different results. The difference often comes from the attribution model, eligibility window, identity signals, and conversion definition used by each system.

An ad attribution model determines how conversion credit is distributed across eligible interactions. Reliable marketing attribution software must also show which channels were included and whether the outcome was a lead, purchase, opportunity, revenue, or retained customer.

This guide explains the main ad attribution models models, applies them to one customer journey, and shows how to choose a reporting view without treating any model as absolute truth.

Key takeaways

• Models distribute credit: The model determines how much credit each eligible interaction receives.

• Windows control eligibility: The window determines how long a click, view, or engaged view can receive credit.

• No universal winner: The right model depends on the campaign objective, journey length, data volume, and business question.

• Platform reports have boundaries: Each ad network measures the interactions and identities available inside its own ecosystem.

• Attribution is not causality: Credit assignment does not prove that an ad created incremental demand.

• Downstream value matters: Strong attribution follows campaigns into qualification, revenue, retention, and lifetime value.

What is an ad attribution model?

An ad attribution model is the rule or method used to assign conversion credit across eligible advertising interactions. It is a paid-media application of broader marketing attribution models, with added attention to ad clicks, impressions, engaged views, campaign identifiers, and platform-specific reporting rules. For a complete overview of attribution in advertising (including ROI considerations), read the full guide to attribution in advertising.

Models are usually grouped into three categories. Single-touch models assign all credit to one interaction, rule-based multi-touch models distribute credit using fixed weights, and algorithmic models estimate contribution from observed patterns.

Single-touch models

First-click, last-click, and last non-direct attribution place all credit on one eligible interaction. They are easy to explain, but they reduce a multi-step journey to one reported winner.

Rule-based multi-touch models

Linear, time-decay, and position-based models divide credit across several touchpoints. Their weighting is transparent, but it reflects predefined assumptions rather than measured causal impact.

Algorithmic models

Data-driven and custom algorithmic approaches use historical paths, conversion outcomes, or business-defined logic to estimate contribution. They can reflect more complexity, but their value depends on data quality, scope, and explainability.

Table 1: Ad attribution models compared

Model	Credit assignment	Best used for	Main limitation
First-click	100% to the first interaction	Discovery and acquisition	Ignores later influence
Last-click	100% to the final interaction	Conversion and closing analysis	Undervalues discovery
Last non-direct	100% to the last known non-direct source	Return and branded journeys	Ignores assisting touches
Linear	Equal credit across all touches	Complete recorded journeys	Assumes equal influence
Time decay	More credit to recent touches	Long nurturing journeys	Undervalues early demand
Position based	More credit to selected stages	B2B and lead-generation funnels	Uses fixed assumptions
Data driven	Credit based on observed patterns	Larger datasets and complex paths	Depends on data quality and scope
Custom	Business-defined rules or weights	Specialized lifecycle models	Can formalize weak assumptions

The comparison shows how each model works, not which one is universally superior. The same company may use different views for discovery, optimization, pipeline, and revenue decisions.

One journey under different attribution models

Consider one recorded path: Meta ad → organic search → Google ad → email → direct visit → purchase. Each model uses the same history but distributes credit differently.

Table 2: How each model assigns credit

Model	Credit result
First-click	Meta ad receives 100%
Last-click	Direct visit receives 100%
Last non-direct	Email receives 100%
Linear	Every interaction receives equal credit
Time decay	Email and direct receive more credit
Position based	The first and final key interactions receive more credit
Data driven	Credit depends on observed conversion patterns
Custom	Credit follows the company’s weighting rules

1. First-click attribution

How it works

First-click attribution gives all credit to the first eligible advertising interaction. In the example journey, the Meta ad receives 100% because it introduced the customer.

Best for

Use it to study initial discovery, new-customer acquisition, awareness campaigns, and the channels creating the first measurable interaction.

Main limitation

It ignores every later touchpoint that educated, nurtured, retargeted, or closed the customer. It is useful as a discovery lens, not a complete account of influence.

2. Last-click attribution

How it works

Last-click attribution gives all credit to the final eligible interaction before conversion. In the example journey, direct receives credit unless the platform excludes direct traffic.

Best for

It works for short journeys, direct-response campaigns, closing-channel analysis, and teams that need a simple reporting baseline.

Main limitation

It can overvalue branded search, retargeting, email, and other late-stage interactions while hiding the advertising that created initial demand.

3. Last non-direct attribution

How it works

Last non-direct attribution ignores a final direct return and gives credit to the most recent known source. In the example journey, email receives 100% rather than direct.

Best for

It helps with repeat sessions, branded journeys, and situations where direct traffic would otherwise overwrite a known acquisition source.

Main limitation

The model still chooses one winner. It protects a known source from direct overwrite but does not recognize the earlier ads and organic interactions that assisted the conversion.

4. Linear attribution

How it works

The linear attribution model divides credit equally across every eligible touchpoint. A five-touch journey gives each interaction 20% of the conversion.

Best for

It is useful for teams beginning with multi-touch reporting and for journeys where no lifecycle milestone is clearly more important than the others.

Main limitation

Equal credit is easy to explain but rarely reflects equal influence. A brief retargeting impression can receive the same credit as the ad that created demand or the email that closed the sale.

5. Time-decay attribution

How it works

Time-decay attribution gives progressively more credit to interactions closer to conversion. Earlier ads still receive value, but recent touches receive a larger share.

Best for

It fits longer sales cycles, nurture campaigns, subscription journeys, and situations where recent interactions are expected to have stronger conversion influence.

Main limitation

The model systematically discounts early demand generation. A high-value awareness campaign can look weaker simply because it occurred farther from the conversion date.

6. Position-based attribution

Position-based models place more credit on selected stages. The model family is popular in lead generation and B2B because teams can emphasize discovery, lead creation, opportunity creation, and conversion.

U-shaped attribution

U-shaped attribution usually emphasizes the first interaction and the lead-conversion touchpoint, with the remaining credit spread across assisting interactions.

W-shaped and Z-shaped attribution

W-shaped models add substantial credit to opportunity creation. Z-shaped models extend the logic to more lifecycle milestones, which can suit longer revenue journeys with several meaningful stage changes.

Main limitation

The weights are business rules rather than causal evidence. They can be useful when the lifecycle is clear, but they may turn assumptions into permanent reporting logic.

7. Data-driven attribution

How it works

Data-driven attribution compares historical converting and non-converting paths to estimate which eligible interactions are associated with conversion. Credit can be fractional and vary by path.

Best for

It is useful for larger datasets, complex advertising journeys, automated bidding, and teams that want to reduce dependence on fixed first-click or last-click assumptions.

Main limitation

The output is bounded by the platform’s available data, identities, channels, and conversion volume. A sophisticated algorithm cannot recover missing touchpoints or disconnected revenue outcomes.

How can AI-powered attribution simplify multi-touch analysis?

AI-driven marketing attribution can reduce the manual work required to compare long paths, channel combinations, conversion stages, and model outputs. It helps analysts move from building reports to investigating the patterns that deserve attention.

AI can group similar journeys, identify recurring assisting sequences, flag sudden changes, and summarize how credit shifts between models. A deeper guide to how multi-touch attribution works provides the measurement foundation that these automated explanations still depend on.

Natural-language interfaces can also answer questions such as which ads assisted high-value deals, which paths produce retained customers, or why a channel gained credit after a model change.

AI should support, not replace, judgment. Missing campaign tags, duplicate conversions, incomplete identities, and disconnected CRM records can produce confident-looking explanations that are not reliable enough for budget decisions.

8. Custom attribution

How it works

A custom attribution model applies company-defined weights, lifecycle milestones, or algorithmic rules. A business might emphasize first contact, qualified lead creation, opportunity creation, and closed revenue differently.

Best for

Custom models fit specialized sales processes, account-based marketing, CRM stages, and businesses with enough historical evidence to justify a tailored weighting system.

Main limitation

Customization can make internal preferences appear more scientific than they are. Teams should document every assumption and compare the custom view with simpler models and experiment results.

How Google Ads applies attribution models

Google Ads now centers model selection on data-driven attribution and last click. Google removed first-click, linear, time-decay, and position-based options from new selectable conversion models, although those approaches remain useful in independent and cross-channel reporting. See Google Ads Help on attribution models.

Data-driven attribution in Google Ads

Google uses eligible account conversion-path data to assign fractional credit across supported interactions. The model can inform automated bidding because credit is distributed beyond the final Google ad click.

Last-click attribution in Google Ads

Last click assigns the conversion to the final eligible Google Ads interaction. It remains simple and transparent but gives little visibility into earlier Google ad assistance.

Why broader models still matter

Independent platforms can compare Google with Meta, LinkedIn, Bing, email, organic search, and CRM outcomes. Teams choosing a Google-specific view can also review the best attribution model for Google Ads before applying the result to wider budget decisions.

Model vs. window vs. scope vs. outcome

Marketers often use these terms interchangeably, even though each setting answers a different question. A report is only comparable when the model, window, scope, and outcome are aligned.

Attribution model

The model controls how credit is distributed among eligible interactions. First-click gives all credit to discovery, while linear distributes it across the recorded path.

Attribution window

The attribution window controls how long a click, impression, or engaged view remains eligible. Short windows emphasize recent influence, while longer windows capture delayed conversions.

Attribution scope

The scope determines which channels, campaigns, placements, and touchpoints are included. A platform-level report may cover only its own advertising inventory, while an independent report can compare paid and organic sources.

Conversion outcome

The outcome is the event receiving credit, such as a form submission, qualified lead, opportunity, purchase, closed revenue, or retained customer. A useful model applied to the wrong outcome can still produce poor decisions.

Click-through vs. view-through attribution

Click-through and view-through attribution describe the eligible ad interaction, not the credit-distribution logic. Either interaction type can be evaluated within a platform model and window.

Click-through attribution

A click receives credit when the user selects the ad and converts within the eligible click window. This is usually stronger evidence of direct engagement than an impression alone.

View-through attribution

A view receives credit when the person sees an ad, does not click it, and later converts. Adjust’s view-through attribution guide explains why impression-based measurement is common in display, video, social, mobile, and connected-TV campaigns.

Engaged-view attribution

Engaged-view attribution is relevant to video advertising where a person watches meaningfully without clicking. It sits between a simple impression and a direct click in the platform’s evidence hierarchy.

Why the distinction matters

View-based credit can materially increase reported campaign contribution. Marketers should compare click and view results separately instead of merging them into one performance number without context.

Attribution windows explained

The lookback window determines how long an interaction remains eligible. Two reports using the same model can still disagree when one uses a one-day view window and another uses a seven-day click window.

Short windows

Short windows emphasize recent response and reduce the chance of crediting distant interactions. They can undercount channels that influence longer consideration journeys.

Long windows

Long windows capture delayed decisions but increase the chance of assigning credit to an interaction with limited influence. The window should reflect the buying cycle rather than a convenient platform default.

Interaction date vs. conversion date

Some reports place conversions on the date of the ad interaction, while others use the conversion date. The totals may match across a long period but disagree sharply by day or week

How ad platforms apply attribution differently

Google, Meta, LinkedIn, Amazon, TikTok, and other platforms use their own identities, eligible interactions, windows, modeled data, and reporting logic. This is why every platform can claim credit for the same conversion.

Google Ads

Google measures eligible interactions within its own advertising inventory. A deeper Google Ads attribution review helps separate Google-reported credit from a wider cross-channel journey.

Meta Ads

Meta can include click-through and view-through influence within its attribution settings. Marketers should interpret Facebook Ads attribution alongside independent conversion and revenue data.

Amazon and closed ecosystems

Commerce platforms can connect ad activity with purchases inside their own environments. This creates strong closed-loop reporting but does not automatically explain touchpoints that occurred outside the platform.

Why totals conflict

Different windows, identities, conversion dates, and eligible interactions create ad platform discrepancies even when every tracking tag is technically working.

Platform attribution vs. independent attribution

Platform attribution

A network evaluates performance through the ads, identities, and conversion signals available inside its own ecosystem. The result is useful for platform optimization but is not a neutral comparison of the full media mix.

Independent attribution

Cross-platform marketing attribution applies a more consistent framework across advertising networks, email, organic search, referrals, direct traffic, and downstream business outcomes.

Independent attribution is not absolute truth. It provides a consistent comparison layer, but the result still depends on tracking coverage, identity resolution, model settings, and the selected conversion outcome.

Which ad attribution model is best?

No model is best for every decision. The useful question is which model matches the business question, data coverage, journey length, and action the team plans to take.

Table 3: Choose a model by business question

Business question	Useful model
Which ads introduce new customers?	First-click
Which ads close conversions?	Last-click
Which source preceded a direct return?	Last non-direct
Which interactions appeared across the journey?	Linear
Which recent touches influenced conversion?	Time decay
Which funnel stages matter most?	Position based
Which interactions contribute based on observed data?	Data driven
Which lifecycle stages need custom weights?	Custom

Factors to consider

1. Campaign objective and funnel stage

2. Journey length and number of interactions

3. Conversion volume and identity coverage

4. CRM, offline, and revenue integration

5. Reporting purpose and audience

6. Model transparency and implementation capacity

Ad attribution models by channel

Paid search

Paid search attribution should distinguish branded and non-branded queries, repeat searches, closing keywords, and automated bidding inputs. Last-click can overvalue branded search when another channel created demand.

Paid social

Paid social often influences discovery through both clicks and impressions. View-through eligibility, retargeting, modeled conversions, and cross-platform overlap require careful interpretation.

Ecommerce advertising

Ecommerce attribution should connect ads with orders, revenue, repeat purchases, products, retention, and LTV rather than treating every purchase as equally valuable.

Mobile advertising

iOS attribution and other mobile measurement approaches must handle installs, deep links, re-engagement, app events, consent, and privacy-aware identity methods.

CTV and video advertising

CTV and video measurement often relies on impressions, engaged views, household exposure, delayed response, and cross-device matching. These journeys require more uncertainty than click-heavy search campaigns.

B2B advertising

B2B models should connect ads with contacts, companies, qualified leads, opportunities, sales activity, and closed revenue. Marketing attribution CRM integration prevents the website form from becoming the final reported outcome.

Attribution models and CRM outcomes

Ad attribution should not stop at the first form submission. The useful lifecycle is ad click → lead → qualified lead → opportunity → customer → revenue.

Ad-spend attribution for CRM systems

A connected system should preserve the original source, campaign history, contact and company identity, opportunity stages, deal value, and closed status. This allows marketing to compare cost per lead with cost per qualified opportunity and revenue.

Revenue attribution

Revenue attribution evaluates campaigns through closed revenue, repeat purchases, expansion, retention, and customer value. It prevents high-volume but low-quality sources from appearing more successful than they are.

How models affect advertising decisions

Changing the attribution model can change reported conversions, fractional credit, CPA, ROAS, keyword performance, campaign rankings, budget recommendations, and automated bidding inputs.

The model changes how recorded credit is distributed. It does not change the historical customer journey, create missing interactions, or prove that the credited campaign caused the outcome.

Attribution vs. incrementality

Attribution asks

Which recorded interactions should receive conversion credit?

Incrementality asks

Did the advertising create additional results that would not otherwise have happened?

Why both are needed

A campaign can appear frequently in converting journeys because it reaches people already likely to buy. Attribution identifies the journey pattern, while experiments test whether the campaign created lift.

Marketing mix modeling can add an aggregate planning layer where user-level tracking is incomplete. Larger budget decisions are stronger when attribution, incrementality, and planning models support one another.

What common mistakes should you avoid with attribution models?

Most failures come from the measurement setup rather than the model name. The following challenges with attribution models should be checked before a team changes budget or declares one channel the winner.

• Using one model for every decision: Discovery, conversion, pipeline, and revenue questions may require different views.

• Ignoring attribution windows: Two reports can use the same model and disagree because their eligibility windows differ.

• Trusting platform totals blindly: Each network measures within its own identities, inventory, and reporting logic.

• Stopping at leads: Lead volume does not show qualification, revenue, retention, or lifetime value.

• Overvaluing view-through credit: Impression-based evidence should be reviewed separately from direct click engagement.

• Letting direct overwrite known sources: Return visits can hide the earlier campaign that introduced or nurtured the customer.

• Changing settings without documentation: Historical comparisons become misleading when models or windows change silently.

• Treating attribution as causality: Conversion credit does not prove incremental impact.

• Using incomplete conversion data: Duplicate events, missing UTMs, and disconnected CRM records weaken every model.

A practical attribution checklist can help teams validate sources, identities, conversions, windows, and business outcomes before relying on model results.

How Usermaven supports ad attribution

Usermaven acts as an independent attribution layer that connects paid advertising with organic acquisition, website behavior, product activity, customer journeys, CRM outcomes, revenue, and retention.

Cross-channel source and campaign tracking

Usermaven applies consistent source and campaign definitions across Google, Meta, LinkedIn, Bing, email, referrals, organic search, and direct traffic. This gives teams a neutral comparison beyond each ad platform’s self-reported totals.

First-touch, last-touch, and multi-touch views

Attribution reporting shows how discovery, closing, and assisting touchpoints connect to the same customer and conversion data, without piecing together separate spreadsheets.

Complete customer journeys

User journeys preserve the sequence of sources, pages, return visits, product events, and conversions. Marketers can inspect the path behind a model result rather than viewing credit in isolation.

Website and product behavior

Website analytics explains landing-page and content behavior, while product analytics connects acquisition with activation, adoption, upgrades, and retention.

CRM, pipeline, and revenue attribution

Campaigns can be evaluated through qualified opportunities, deal values, closed revenue, retention, and customer quality. This gives paid-media teams a stronger outcome than clicks or form submissions alone.

Conversion synchronization

Conversion syncs return verified first-party outcomes to ad platforms so bidding can optimize toward qualified leads, purchases, customers, or other meaningful events.

AI-assisted analysis

Maven AI helps teams query connected journeys, attribution paths, campaigns, and outcomes in natural language. The result supports faster investigation without replacing data validation or strategic judgment.

Final verdict

An ad attribution model controls how conversion credit is divided, but the model is only one part of the measurement setup. Reliable decisions also depend on windows, platform scope, identity coverage, conversion definitions, CRM integration, and data quality.

First-click explains discovery, last-click explains closing, rule-based multi-touch models show the recorded path, and data-driven models estimate contribution from observed patterns. Comparing several views is usually more useful than forcing every decision into one model.

Usermaven connects advertising interactions with the wider customer journey, website and product behavior, CRM outcomes, revenue, and retention. Start a free 14-day Usermaven trial to test attribution models against real customer and business outcomes.

FAQs