Marketing attribution models: how to choose the right one in 2025?

Attribution: it’s a word that makes marketers the world over tremble. It’s the most talked-about headache in marketing communities. Everyone has an opinion, or not, but everyone also frequently questions it. A marketing attribution model is a method for distributing the credit for a conversion among the various touchpoints in the customer journey. In principle, then, it’s quite simple: the attribution model answers the question: which lever(s) actually contributed to the conversion?
In a context where customer journeys are multi-channel (search, social, e-mail, web push, display, etc.), the attribution model becomes indispensable for measuring the effectiveness of each contact point and better steering marketing investments. So we’re all familiar with the first-touch, last-touch, linear, time-dependent, U-shaped and W-shaped models. But, spoiler alert: in 2025, Google is clearly pushing Data-Driven Attribution (DDA) as standard, notably in Google Ads and GA4, as it distributes credit according to actual observed contributions (and not fixed rules).

What is a marketing attribution model?

A marketing attribution model is a rule (or algorithm) that distributes the credit for a conversion among the touchpoints of a path (clicks, views, interactions). Objective: measure for better decision-making (strategy, budget reports, bids, content).

Why is it important?

• To understand the real role of each lever in the funnel.
• To avoid over-funding some channels and under-funding others.
• To inform strategic decisions: budget, bids, content, messages, pathways.

👉 Without attribution, you’re flying blind.

Advantages in marketing strategies:

• Highlights the levers used.
• Makes budget decisions more objective.
• Facilitates alignment between marketing, media and management teams around clear KPIs.

Limits and pitfalls:

• Insufficient data volume: some models, such as DDA, require a robust history.
• Double counting possible if deduplication between tools (Ads, GA4, CRM) is poorly parameterized.
• Increasing complexity: the more advanced the model, the more expertise and maintenance it requires.

Historically, marketers relied on simple, easy-to-apply models: last-click (100% of credit on last click) or first-click (100% on first click). These approaches, widely used with Universal Analytics and advertising tools, were reassuringly simple, but masked the complexity of customer journeys.
Gradually, other rule-based models (linear, time-decay, U or W position-based) were proposed to provide a more balanced vision.
But by 2025, the trend was clear: Google was pushing DDA as standard in GA4 and Google Ads. This algorithmic approach distributes credit according to actual observed contributions, rather than relying on arbitrary rules.

So there are now two types of attribution model: rules-based and algorithmic.

Rule-based attribution models

Last touch

The Last Touch model attributes 100% of the conversion credit to the last point of contact before the decisive action.

👉 Example: a user discovers a brand via a Facebook ad, receives a newsletter, clicks on a Web Push, then converts by typing the brand name into Google and clicking on an ads → Google Ads receives 100% of the credit.

Advantages of the Last Touch model

• Ultra-simple to understand and use.
• Stability: always the same measurement criterion (the last click).
• Historically the standard in Google Analytics (Universal Analytics) and still often the default in certain CRM or advertising platforms.
• Useful for measuring closing channels (retargeting, final email, Web Push follow-up).

Limits of the Last Touch model

• Major bias towards the bottom of the funnel: totally ignores discovery and nurturing levers.
• Encourages over-investment in retargeting and search brands, to the detriment of channels that create demand.
• Doesn’t reflect the reality of a multi-touch journey (especially in 2025, when users will go through several touch points before buying).
• Not compatible with cookieless or cross-device strategies (loss of tracking).

When to use the Last Touch model?

• As a comparison model (basic benchmark in GA4).
• For low data volumes, when DDA is not yet available.
• For quick, minimalist reporting for managers who want a simplified reading.
• As a complement to another model, to identify closing levers.

Adrena’tip – Last Touch can remain useful as a safety net when data is insufficient for an algorithmic model. But it should never be the only prism through which we read data, at the risk of under-funding strategic channels such as content, social or Web Push.

First touch

The First Touch model attributes 100% of the conversion credit to the first point of contact between a prospect and the brand.

👉 Example: a user discovers a brand via a Facebook ad, receives a newsletter, clicks on a Web Push, then converts by typing the brand name into Google and clicking on an ads → Facebook receives 100% of the credit.

Advantages of the First Touch model

• Extreme simplicity: easy to understand and explain.
• Values the discovery phase and therefore awareness channels (organic search, social networks, display, Web Push…).
• Can be useful for measuring the impact of top-of-funnel campaigns in acquisition strategies.
• Interesting for branding teams who want to prove their role in lead generation.

Limitations of the First Touch model

• Completely ignores the nurturing and closing stages: follow-up emails, retargeting, post-panel Web Push, etc.
• Very partial vision: values discovery, but ignores the levers that really convert.
• Can lead to over-investment in unprofitable top-of-funnel channels, if used alone for budget reallocation.
• Less and less offered natively in tools like GA4 (replaced by DDA).

When to use the First Touch model?

• To analyze the effectiveness of awareness campaigns (e.g. display campaigns, social networks, Web Push).
• To map entry into the funnel and identify which levers trigger the first contact.
• In a strategy where the main objective is to acquire new leads/visitors.
• As a point of comparison (but not as a single steering model).

Adrena’tip – Use First Touch only as a supplement. It is useful for proving that branding investments create value, but it must always be set against a more balanced model (U, W or ideally DDA).

Linear

The linear attribution model distributes conversion credit equally across all touchpoints in the customer journey.

👉 Example: if a user converts after 4 interactions (Search → Social → Web Push → Email), each lever receives 25% of the credit.

Advantages of the linear model

• Simple and easy to explain to all stakeholders.
• Adds value to the entire customer journey: no lever is invisible.
• Ideal for demonstrating multi-touch logic to a team still accustomed to last-click.
• Promotes a collaborative vision between teams and marketing channels.

Limits of the linear model

• Dilutes the importance of key touchpoints: discovery or final conversion receive as much weight as a minor click .
• Arbitrary: 100% equal distribution does not necessarily correspond to the reality of contributions.
• May undervalue decisive channels (last click, basket relaunch, direct call to action).
• Not relevant in very long campaigns where dozens of interactions have taken place (e.g. complex B2B SaaS).

When to use the linear model?

• For internal education: to show concretely that several levers contribute to conversion.
• For short paths with few contact points (2 to 5 interactions).
• In an exploratory phase where the aim is to test and balance investments before adopting a more advanced model (DDA).

Adrena’tip – The linear model can be useful as a point of comparison in GA4 (via model comparison reports). But don’t use it alone to steer your budget, as it may mask the reality of the channels that are really performing.

Time-decay

The Time-Decay model gives more credit to the interactions closest to conversion in time. The principle is similar to a“curve”: older interactions lose importance.

👉 Example: the closer you get to the act of purchase or conversion, the more weight is given to the point of contact. Click at D-30 → 10% credit / Click at D-7 → 20% credit / Click at D-1 → 40% credit / Conversion → 30% credit.

Advantages of the Time-decay model

• Reflects the reality of many customer journeys, where the last interactions are often decisive.
• Adds value to nurturing and follow-up channels (e.g. retargeting, email reminders, Web Push re-engagement).
• More nuanced than last-click, as it retains some credit for the top of the funnel.
• Easy to defend in short-term optimization contexts (seasonal campaigns, flash promotions).

Limits of the time-decay model

• Under-values discovery points (top of funnel), which may be essential for acquisition.
• Like all rule-based models, it remains arbitrary: distribution depends on a default decay rule, not on actual contribution.
• Less relevant for very long or complex customer journeys (B2B, SaaS).
• Less and less offered by default in platforms (GA4 favors DDA).

When to use the Time-decay model?

• When the bottom of the funnel is critical (e.g. e-commerce with a lot of basket relaunches, time-limited promotions).
• To enhance the value of reactivation interactions such as automatic emails, SMS or Web Push reminders.
• In a context where conversion cycles are short to medium (a few days to a few weeks).

Adrena’tip – Time Decay is interesting for demonstrating the value of reminder channels (email, SMS, Web Push), but it should not be used alone to arbitrate acquisition budgets. Combine it with incrementality tests or comparative analyses (DDA vs. Time Decay) to get a more balanced view.

Position-based (U-shaped)

The U-shaped (or position-based) attribution model distributes the credit for a conversion over two key moments:

1. The first point of contact (brand discovery).
2. The last point of contact (the decisive action before conversion).

Each generally receives 40% of the credit.
The remaining 20% is divided equally between all intermediate contact points.

👉 Visually, this forms a “U”: lots of weight at the beginning and end, less in the middle.

Advantages of the U-shaped model

• Promotes discovery and conclusion: useful for showing the importance of awareness campaigns AND closing channels.
• More balanced than last-click or first-click, which make half the journey invisible.
• Simple to understand and explain to marketing teams.

Limitations of the U-shaped model

• Under-values the mid-funnel (nurturing, follow-ups, educational content, intermediate Web Push).
• Arbitrary: 40/40/20 is not based on real data, but on a fixed rule.
• May give a biased view if your funnel is long and complex (purchase cycles of several weeks, as in SaaS or B2B).
• Less and less accessible natively: GA4 has removed this model from its interface in favor of DDA.

When to use the U-shaped model?

• When you want to demonstrate the importance of awareness (top of funnel) and closing (bottom of funnel).
• For sales paths that are not too long, where intermediate stages play a secondary role.
• In companies that are starting to move away from last-click and need an intermediate step before adopting DDA.

Adrena’tip – If your aim is to convince management to invest both at the top of the funnel (awareness) and at the bottom (retargeting, CRM, re-engagement web push), the U-shaped model is a good pedagogical option. But as soon as the volume of data allows, switch to DDA to capture the real value of intermediate touch points.

Position-based (W-shaped)

The W-shaped attribution model distributes the credit for a conversion primarily over three key moments in the customer journey:

1. The first point of contact (discovery).
2. The contact point that generated the lead/qualification (often a form or registration).
3. The last point of contact before conversion (closing).

Each of these three points generally receives 30% of the credit. The remaining 10% is distributed equally among the other interactions in the customer journey.

Advantages of the W model

• Values the mid-funnel, often underestimated in other models (especially the contact that generates a qualified lead).
• Balances top, middle and bottom of funnel: recognition of awareness, nurturing and conversion levers.
• Suitable for B2B or long cycles, where lead generation (MQL/SQL) is a key stage.

Limits of the W model

• Arbitrary: like the U-shaped, it is based on a pre-determined rule (30/30/30/10) and not on the actual contribution measured.
• Not suitable for short runs (1-2 interactions only).
• May dilute the role of certain important touch points if they do not correspond to the three defined milestones.
• Not natively available in GA4 or Google Ads (requires third-party tools or specific configuration).

When to use the W model?

• In B2B SaaS: when lead generation (form, demo request) is a strategic step in the funnel.
• In complex lead paths (multi-interactions over several weeks/months).
• In the training phase: to show teams that the middle of the funnel has a real impact on conversion.

Adrena’tip – If you want to give visibility to the mid-funnel (like the activation email, the post-trial Web Push or a content campaign), but don’t yet have enough data to activate the (DDA), the W model is a good temporary alternative.

Algorithmic attribution models

• Data-Driven Attribution (DDA): calculation based on actual contributions measured in path data.
• In some cases: use of more global methods such as Marketing Mix Modeling (MMM) for an aggregated view (annual budget, offline/online trade-offs).

How to choose the right attribution model?

The tricky part is choosing the right attribution model, the one that’s most relevant to your business model. The impact of the wrong choice? Over-investment in low-funnel channels, under-funded content & social, invisible Web Push, biased ROAS. Here’s some food for thought:

• Data volume: low → temporary last-click+ tests (control/exposure) for key channels; sufficient → DDA.
• Lots of offline → QR/coupons, conversion imports + aggregated view (MMM) for annual arbitrage.
• Marketing objectives: awareness, conversion, retention.
• Journey complexity: mono-touch → simple; multi-touch → algorithmic.
• Business stakes: e-commerce, retail, B2B SaaS do not have the same constraints.
• Ability to test: possibility of launching exposure/control groups, QR codes, incrementality.

But to delve deeper into the practicalities, it’s useful to put the different marketing attribution models side by side, to better grasp their strengths and limitations. This summary comparison enables you to see at a glance how each model works, its advantages and disadvantages, and the contexts in which it is most relevant.

Adrena’tip – A good model never excuses overexposure: set a cross-channel course (email/SMS/Web Push).

Real-life use cases

E-commerce: rebalancing the budget between search and social/web push

“The last click fools me”: An e-merchant finds that in last-click, paid search takes all the credit. In DDA, social networks and Web Push finally come into their own: 15-20% of the budget is reallocated to mid-funnel strategies = content & push that prepare for conversion. (Method: model comparison + control/exposure test).

Drive-to-store: QR codes and offline attribution

A retailer deploys a geolocated push + QR coupon for offline measurement of visits. DDA attribution in GA4 + geo-side lift Ads: adjustment of local advertising pressure according to defined criteria (e.g. weather).

B2B SaaS: highlighting the role of onboarding

In DDA, activation e-mail and post-trial Web Push stand out as critical levers; activation content and in-app messages are prioritized over increasing spending at the bottom of the funnel.

Adrena’tip – Always back up an incrementality test with a budget swing (geotest, control groups).

Want to learn more about Web Push?

I want a demo!

FAQ – Marketing attribution model

How do you choose the right attribution model for your marketing campaigns?

• Analyze your data: volumes, attribution windows, journeys.
• Map your channels: search, social, Web Push, email.
• In Analytics (GA4), use the Model Comparison reports to estimate the impact on conversion and CPA/ROAS.
• Choose according to the sales cycle: short → temporary last-click; multi-touch → data-driven.
• Document deduplication: rules across channels/platforms.
• Translate into actions: budget reallocation, new audiences, new messages.

How to deploy attribution in GA4 and align it with HubSpot (or CRM tools)?

• In GA4: mark your conversions, activate Model Comparison, align windows and check UTMs.
• In HubSpot (or CRM): sync events & campaigns to leverage first-party data (forms, emails, deals).
• Align attribution rules (windows, deduplication) across tools.
• Create shared reports: pipeline, sources, ROAS to connect acquisition → conversion.
• Result: a unified view that informs strategy by channel, message, and segment.

How to implement attribution in GA4?

• 1. Go to settings: Admin → Attribution settings (reporting). Select Data-Driven.
• 2. Attribution reports: Advertising → Attribution → Models to compare models.
• 3. Mark “key events” (e.g. conversions) and link Google Ads; GA4 recommends DDA for multi-channel performance.
• 4. Export to Google Ads: create Google Ads conversions based on GA4 events to align measurement and Smart Bidding.
• 5. Stability: note that GA4 may take a few weeks to stabilize modeling & attribution on new signals.

How to be operational with attribution in 30 days?

• Week 1 — Map & clean: journeys, windows, deduplication, Ads↔GA4 links. Deliverables: attribution doc + last-click vs DDA comparison table. KPI: conversion consistency.
• Week 2 — Switch to DDA & launch 1 test: DDA activated, geo test control/exposed on one media lever (e.g. Web Push network) + QR in-store. KPI: incremental lift, validated store visits.
• Week 3 — Highlight the mid-funnel: Web Push scenarios (post-purchase J+7, back-in-stock), mapping of intermediate conversions (cart adds, product views). KPI: mid-funnel credit share.
• Week 4 — Arbitrate & plan Q+1: “DDA vs incrementality” table by lever; if lift < X%, reduce. Launch a mini-MMM (aggregated view) for the quarterly budget.