Attribution models: analytics for multichannel campaigns.

An attribution model is the set of rules that decides how much weight is assigned to each touchpoint in the customer journey when a conversion happens — a sale, a lead, a download. It is the basis on which the decision is made of which channels will receive more budget the following month. Without a defined and consistent model, the marketing department is optimising blind.

The topic matters because digital advertising investment in Spain reached €5.030 billion in 2024, according to IAB Spain's Digital Media Advertising Investment Study. Of that figure, an average Spanish advertiser dedicates between 30% and 55% to direct-response channels (search, social, programmatic display, affiliation). The criterion to distribute that amount across channels depends, almost always, on the active attribution model in its analytics tool. Most alarmingly, 61% of Spanish advertisers still use last-click attribution by default (IAB-Smartme survey 2024), despite it being the model with the worst explanatory capacity.

What exactly is an attribution model and why do you need one?

When a user reaches a purchase, they almost never do it through a single channel. The typical journey of a B2C digital sale in Spain in 2024 went through 8.3 brand touchpoints on average before converting (Wavemaker study 2024). In B2B the figure rises to 14.7 touchpoints. Each touchpoint is a cost: a click bought on Google Ads, a display impression, an organic visit, an email opened, a direct visit, a social network click.

If the marketing department only attributes the conversion to the last click before the sale, all previous touchpoints are left without value in the report. This has three consequences:

1. Channels that generate demand in the early funnel phase (display, branded content, organic social, induced branded search) are undervalued and receive less investment.
2. Channels that close the sale in the final phase (brand search, retargeting) are overvalued and receive more investment than they justify.
3. The budget migrates from what builds to what harvests, until there is almost nothing left to harvest.

The attribution model is the tool that corrects — or aggravates — this imbalance.

What are the most used attribution models and when does each fit?

There are seven standard attribution models recognised by the main analytics tools (Google Analytics 4, Adobe Analytics, Piwik PRO, Matomo). The choice depends on the business model, purchase cycle and the organisation's measurement maturity.

Model	How it assigns weight	When it fits
Last-click	100% to the last touchpoint	Impulse purchase, <24h cycle
First-click	100% to the first touchpoint	Awareness acquisition phase
Linear	Equal weight to all	No clear prior hypothesis
Time decay	More weight to most recent	Promotions and expiry campaigns
Position-based	40-20-40 across first, middle, last	Medium cycle (considered B2C, agile B2B)
Data-driven	Statistical model on real data	Sufficient volume (>600 conv./month)
Markov / Shapley	Advanced mathematical models	Organisations with in-house data scientist

Google Analytics 4's data-driven option uses a combination of Shapley value and machine learning over the advertiser's actual data. It is the option recommended by Google Analytics official documentation since the introduction of GA4 in 2020, and replaced last-click as the default model in 2023. To activate it, a minimum volume of 600 conversions in 30 days within the property is needed — below that threshold, GA4 forces the switch to an account-level data-based model or, failing that, to last-click.

What evidence is there that changing the attribution model moves the budget?

The study Multi-channel attribution analysis published by Google in 2019 (with data from 800 advertisers) quantified the typical split before and after moving from last-click to data-driven. Summary of findings:

• Discovery channels (display, YouTube, paid awareness social) saw their contribution increase between +22% and +41% after the change.
• Brand search saw its contribution reduced between -15% and -28%.
• Email marketing and retargeting tended to stabilise or slightly reduce.
• Non-brand organic SEO maintained or slightly increased its contribution.

Translated into euros: an advertiser with an annual digital budget of €600,000 that moves from last-click to data-driven and reassigns accordingly usually improves aggregate ROI between 12% and 19% in the following 6-9 months (Bain & Company study 2022 on 320 European mid-sized companies). The reason is simple: better evidence, better decision, better budget distribution.

Why do so many companies still use last-click if it does not work?

Four reasons explain the persistence of the last-click model despite its limitations:

1. It is the default in legacy tools. Google Analytics Universal Analytics (replaced by GA4 in July 2023) had last-click as the default model for a decade. Many organisations still report with inherited dashboards carrying this criterion.

2. It is the simplest model to explain. "The last channel that touched the customer before buying" is understandable for any executive. Any alternative requires additional explanation.

3. Lack of data volume for data-driven models. Below 600 monthly conversions, GA4 cannot activate the data-driven model. Many Spanish SMEs do not reach that threshold and stay in last-click by tool technical decision, not by conscious choice.

4. Internal resistance from the closing channel manager. The manager of Google Ads or retargeting (the closing channels) usually resists the change because last-click gives them more metric than a distributed model. This resistance is real and needs political, not just technical, management.

How to choose the right model for your company?

The criterion I apply in consulting is based on three sequential questions:

1. How long does your purchase cycle last from first touchpoint to sale? If under 48 hours (impulse purchase, restaurants, fast-moving retail), last-click is defensible. If between 5 and 30 days (considered B2C, large appliances, travel), position or time decay is recommended. If over 30 days (B2B, real estate, high-ticket training), data-driven or a custom model is needed.
2. How many monthly conversions do you have? Above 600, activate data-driven in GA4 directly. Between 200 and 600, use position (40-20-40) as a reasonable approximation. Below 200, data-driven models give statistical noise; stay at position or linear.
3. How much does your company invest in early-funnel channels (display, branded content, organic social, video)? If more than 25% of the budget, last-click is unacceptable because it is systematically underrating that investment. If less than 10%, last-click reasonably approximates reality.

The most frequent conclusion in real projects: the average Spanish SME with 14-30 day sales cycle and volume of 100-400 monthly conversions should be on position-based model (40-20-40), not last-click. The change is made in GA4 in less than 10 minutes and improves the quality of budget decisions from the next month onwards.

How do privacy restrictions and the end of cookies affect the attribution model?

The measurement context has changed radically between 2022 and 2026. Three relevant movements:

• End of third-party cookies in Chrome. Google progressively removed them through 2024 for a part of users. This broke the cross-domain tracking that underpinned traditional multi-channel last-click.
• Apple's App Tracking Transparency (ATT). Since iOS 14.5 (April 2021), Apple requires apps to ask for explicit consent for cross-app tracking. The opt-in rate in Spain is around 27% (Branch data 2024). The remaining 73% falls outside the classic attribution model.
• GDPR and consent mode v2. The European data protection regulation, complemented by the consent mode v2 mandatory in Google since March 2024, restricts data collection without explicit consent.

The consequence: the percentage of the customer journey directly observable by the analytics tool has fallen from 85-95% in 2019 to 55-70% in 2024 according to IAB Spain estimates. The rest is modelled. This explains why data-driven models (which fill the gaps with statistical inference) are increasingly necessary and why last-click is increasingly imprecise — it simply does not see most of the journey.

"In 2026 the attribution model is not just another technical option: it is the difference between making budget decisions based on real data or making them based on a small biased portion of the data."

— Fernando Maciá, Marketing Online 4.0 (Anaya Multimedia, 2023)

What role do media mix modelling and incrementality testing play?

The multi-touch attribution model calculated from Google Analytics or similar tool is the operational, accessible and fast way to do attribution. It has two limitations worth knowing:

1. It only sees online channels observable by the tool. TV, radio, offline media, branded content not generating direct click, are left out.
2. It assumes all touchpoints are causal. In reality, some are merely correlative (the user would have converted anyway without that touchpoint).

To address both limitations, two complementary techniques have become popular:

Media Mix Modelling (MMM). Statistical model that correlates investment per channel with sales over time, including online and offline channels. Does not need cookies. It is the favourite technique of large advertisers since privacy broke multi-touch attribution. Google released an open-source version called Meridian in 2022 that has partially democratised the discipline. Implementation cost: €30,000-100,000 in specialised consultancy the first year.

Incrementality testing. Geographic or audience test in which a channel is switched off in a control group and kept active in an experimental group. The sales difference measures the real contribution of the channel, not the attributed one. It is the gold standard for validating investment in branded content and other channels with hard click-by-click measurement. Platforms like Meta and Google offer integrated tools (Conversion Lift Studies, Geographic Experiments).

The mature organisation combines the three layers: multi-touch attribution for daily tactical optimisation, MMM for annual strategic split between online and offline, incrementality testing to validate specific hypotheses. Below €1.5 M annual digital budgets, the latter two are usually unattainable and only enhanced multi-touch attribution is used.

How to implement a good attribution model step by step?

Implementation protocol I recommend in consulting:

1. Audit of current measurement. Identify the active model in GA4, the UTMs used, the configured conversion events and consent mode coverage. Usually takes 1-2 weeks. This phase surfaces tagging errors that invalidate any later analysis.
2. Definition of target model. Decide which model is reasonable given the volume and the purchase cycle. Document it in one page with justification. This page is the reference for any future discussion.
3. Configuration in GA4. Apply the model in GA4 (Configuration → Attribution settings → Reporting attribution model). The change applies to new reports, not retroactively.
4. Reconciliation with advertising platforms. Google Ads, Meta Ads and other platforms have their own attribution model. Align conversion windows and model between the platform and GA4 to reduce discrepancies.
5. Budget reassignment. After 4-8 weeks with the new model, recalculate ROI per channel and reassign the budget according to the new evidence. Document the decision with dashboard screenshots, in case of need to revert.
6. Quarterly review. The model is not set-and-forget. Each quarter, review the split with the leadership team and adjust.

What tools does a Spanish SME need to do attribution well?

Minimum viable tools, with a reasonable SME budget:

• Google Analytics 4. Free. Covers 80% of an SME's multi-touch attribution needs.
• Google Tag Manager. Free. Manages the consistent tagging that sustains any model.
• Google Ads + Meta Ads Reporting. Native metrics inside the platforms. Indispensable.
• Looker Studio (formerly Data Studio). Free. Single dashboard combining data from multiple platforms.
• Server-side tagging (recommended above a certain volume). Server-side GTM allows recovering part of the information lost with cookie blockers. Cost: €50-200 per month on Cloud Run.

Above annual digital budgets of €600,000, consider additional tools: AppsFlyer or Adjust for mobile attribution, Singular or Segment for identity resolution, MMM tools like Meridian or Robyn for advertisers with relevant offline investment.

What frequent mistakes should be avoided when working with attribution models?

Five mistakes that appear again and again in projects:

1. Changing the model without informing the team. The split across channels will change and the affected team will interpret the change as a political decision against them. Communicate it in advance.
2. Confusing model with reality. The model is a reasonable hypothesis of how merit is distributed for the conversion, not the absolute truth. Changing the model does not change sales, only how they are reported.
3. Not reconciling GA4 with advertising platforms. If Google Ads says 320 conversions and GA4 says 245, the discrepancy must be investigated. It is usually a mix of different conversion windows, different deduplication and different cross-domain attribution.
4. Optimising on wrong metrics. ROAS without taking into account the active attribution model is misleading. A high ROAS in last-click can be a mediocre ROAS in data-driven.
5. Forgetting the time dimension. Conversion windows should be reasonable. A sale attributed 90 days after the first touchpoint needs explanation, not automatic attribution.

How to integrate attribution with the overall marketing strategy?

The attribution model is a tactical tool, not a strategy. It does not replace the question of what need your company satisfies nor the decision of what kind of relationship you want to build with your customer. The big strategic bets — investing in branded content long term, dedicating budget to utility marketing, betting on programmatic or SEO — are taken before and above the attribution model.

What the model contributes is the tool to optimise the budget split within that strategic decision. Attribution without strategy is empty optimisation; strategy without attribution is budgetary blindness. The two need each other.

If you want to review how attribution is configured at your company and what quick change would bring the most value in the next 90 days, book a first session at no cost. In 45 minutes we review your GA4, identify the three highest-impact adjustments and chart a realistic implementation plan.

Frequently asked questions

What is a multichannel attribution model?

It is the rule that assigns credit for the conversion to the different touchpoints the user travelled through before buying.

What are the most used attribution models?

Six classics: last-click, first-click, linear, time decay, position-based (40-20-40), and data-driven (GA4 algorithm).

What is the best attribution model?

It depends on the business. For short cycles (impulse B2C) last-click works. For long B2B cycles with many touchpoints, linear or data-driven.

How do I choose the right attribution model for my SME?

Three questions: how long is your funnel?, do you invest in branding?, do you have enough volume for data-driven in GA4?

How does cookieless affect attribution?

It forces a shift from third-party cookies to contextual audiences, first-party data, universal identifiers and predictive models.