Attribution Models: Which One Is Lying to You?

Attribution models are frameworks that assign credit for a conversion across the marketing touchpoints a customer encountered before converting. The model you choose determines which channels look effective, which get defunded, and, in many cases, which version of reality your business is making decisions from.

There are six common attribution models in widespread use: first-touch, last-touch, linear, time decay, position-based, and data-driven. Each tells a different story. The problem is that most marketing teams pick one, trust it, and forget to ask whether it’s the right story for their business.

I spent years running paid search and performance campaigns at scale, managing hundreds of millions in ad spend across 30-odd industries. The attribution debate was constant. Every agency pitch involved someone claiming their model proved their channel deserved more budget. Every client conversation involved someone asking why the numbers from their ad platform didn’t match their analytics. Attribution sits at the centre of almost every commercial disagreement in marketing, and most of the time, neither side is wrong. They’re just using different models.

What Is an Attribution Model and Why Does It Matter?

Before getting into the specific models, it’s worth being clear about what attribution is actually trying to do. A customer rarely converts after a single touchpoint. They might see a display ad on Monday, click a paid search ad on Wednesday, open an email on Friday, and then convert through organic search the following week. Attribution is the process of deciding how much credit each of those touchpoints deserves for the eventual conversion.

This matters because credit determines budget. If your attribution model says paid search drove the conversion, paid search gets more money. If it says email did, email does. The model is not a neutral observer. It is a political document. And if you’re not conscious of its biases, you’ll make budget decisions that systematically favour whichever channel happens to sit at the point the model rewards.

If you want a broader grounding in how analytics tools measure and misrepresent marketing performance, the Marketing Analytics hub covers the full landscape, from GA4 mechanics to measurement frameworks that actually connect to commercial outcomes.

Last-Touch Attribution: The Default That Favours Closers

Last-touch attribution gives 100% of the credit for a conversion to the final touchpoint before it happened. If someone clicked a branded paid search ad and then converted, paid search gets all the credit. Everything that happened before that click, the awareness campaign, the social post, the comparison site listing, gets nothing.

This was the default model for most platforms for a long time, and it still is in many environments. The reason it persisted is partly historical (it was easy to implement technically) and partly because it flatters bottom-funnel channels, which tend to be the ones with the most direct line to revenue. Paid search teams loved it. Brand teams hated it.

I saw this play out repeatedly when I was at iProspect. We were growing the agency from around 20 people to over 100, and paid search was the engine. Last-touch attribution made paid search look extraordinary. The problem was that a significant chunk of what we were capturing was branded search, people who had already decided to buy and were just using Google to handle to the site. We were taking credit for conversions that would have happened anyway. The attribution model was making us look better than we were, and it was making it very hard for clients to justify investment in anything that sat earlier in the funnel.

Last-touch attribution is not useless. For very short purchase cycles with a single touchpoint, it can be a reasonable approximation. But for any business with a considered purchase experience, it will systematically undervalue brand, content, social, and upper-funnel activity. It rewards channels that fish where the fish already are, not the ones that put the fish there.

First-Touch Attribution: The Opposite Problem

First-touch attribution is the mirror image. It gives 100% of the credit to the first touchpoint in the customer experience. If someone first encountered the brand through an organic search result and then converted three weeks later via a retargeting ad, organic search gets all the credit.

This model tends to be championed by content teams and SEO practitioners, for obvious reasons. It makes awareness channels look powerful. And to be fair, there is a legitimate argument that without the first touchpoint, the experience never starts. First-touch is asking a reasonable question: what introduced this customer to us?

The weakness is the same as last-touch but in reverse. It ignores everything that happened after the introduction. A customer who first saw a brand through a display ad and then spent six months reading blog posts, attending webinars, and engaging with retargeting before converting is crediting the display ad for all of that work. The nurture activity that kept them engaged and moved them towards a decision gets nothing.

First-touch is useful if your primary business question is “where are our best customers coming from originally?” It’s a reasonable input into acquisition strategy. It’s a poor basis for evaluating the full performance of your marketing mix.

Linear Attribution: Splitting the Difference

Linear attribution distributes credit equally across every touchpoint in the customer experience. If there were five touchpoints, each gets 20%. It sounds fair, and compared to the binary extremes of first-touch and last-touch, it is more balanced.

The problem is that equal credit is not the same as accurate credit. Not all touchpoints are created equal. A single highly targeted email at the moment of peak consideration is probably doing more work than the third impression of a display ad that the customer barely registered. Linear attribution treats them identically.

Linear attribution also has a volume problem. Channels that generate a lot of impressions or touchpoints will accumulate more total credit simply because they appear more frequently in the experience, not because they are more effective. High-frequency display campaigns can look deceptively strong under a linear model purely because they show up everywhere.

That said, linear attribution is a reasonable starting point for teams that are moving away from single-touch models and want a more distributed view without the complexity of building something more sophisticated. It’s a step in the right direction, even if it’s not the destination.

Time Decay Attribution: Rewarding Recency

Time decay attribution gives more credit to touchpoints that happened closer to the conversion. The logic is intuitive: the interaction that happened just before someone converted was probably more influential than something they saw three months ago.

This model tends to suit businesses with shorter consideration cycles and high-intent purchase behaviour. If someone is actively researching and comparing options in the days before they buy, the touchpoints during that window are plausibly doing more work than the ones that introduced them to the brand months earlier.

The limitation is that time decay can undervalue the touchpoints that built the brand preference that made the customer receptive to those late-stage interactions in the first place. A customer who bought because they trusted the brand, because they had seen consistent, quality communication over months, will have that long-term brand-building activity discounted heavily in a time decay model. You end up rewarding the last mile and ignoring the infrastructure that made it possible.

For businesses with a longer sales cycle, B2B companies in particular, time decay attribution can produce a distorted picture. A deal that took nine months to close and involved multiple stakeholders, content consumption, and relationship-building touchpoints is not well served by a model that concentrates credit on the final two weeks.

Position-Based Attribution: The Compromise Model

Position-based attribution, sometimes called U-shaped attribution, splits the credit by giving a larger share to the first and last touchpoints and distributing the remainder across the middle. A common configuration is 40% to first touch, 40% to last touch, and 20% spread across everything in between.

This reflects a reasonable intuition: the touchpoint that introduced the customer to the brand matters, and the touchpoint that closed the conversion matters, and the stuff in the middle matters somewhat. It’s a more nuanced position than single-touch models and acknowledges that both acquisition and conversion deserve recognition.

The honest limitation is that the 40/40/20 split is arbitrary. There is no empirical basis for those numbers. They feel balanced, which is not the same as being accurate. Position-based attribution is a better political compromise than a scientific measurement. It tends to reduce the most extreme distortions of single-touch models while being simpler to explain than data-driven approaches.

For teams that need a model they can explain to stakeholders without a statistics degree, position-based attribution is often the most practical choice. It’s directionally sensible even if it’s not precise.

Data-Driven Attribution: The Sophisticated Option With Caveats

Data-driven attribution uses machine learning to assign credit based on the actual patterns in your conversion data. Rather than applying a fixed rule, it analyses which combinations of touchpoints are most associated with conversion and weights credit accordingly. Google’s data-driven attribution model in GA4 works on this basis, as does the equivalent in Google Ads.

In theory, this is the most accurate approach. It doesn’t apply a predetermined formula. It learns from your specific data. For large accounts with high conversion volumes, it can produce genuinely useful insights about which channel combinations drive the best outcomes.

In practice, there are significant caveats. Data-driven attribution requires substantial conversion data to produce reliable outputs. Accounts with thin conversion volumes will get unstable, noisy results that shift significantly from month to month. The model is also a black box. You cannot fully audit why it assigned credit the way it did, which makes it difficult to challenge or interrogate. And because it’s built on observed patterns in your existing data, it can entrench existing biases rather than revealing new truths.

I’ve seen data-driven attribution used as a conversation-stopper in client meetings. “The model says so” becomes a way of avoiding the harder question of whether the model is actually capturing what’s happening. If you want to understand how GA4 handles engagement and attribution signals, Semrush’s breakdown of GA4 engagement rate is a useful reference for understanding how the platform is recording user behaviour before attribution even starts.

Data-driven attribution is worth using if you have the data volume to support it. But treat it as one input, not as the answer.

The Attribution Model You Choose Is a Commercial Decision

One of the things I took away from judging the Effie Awards was how rarely attribution came up in the entries that won. The most effective campaigns, the ones that could demonstrate real business impact, tended to be evaluated against harder measures: revenue growth, market share, category penetration. Attribution model debates are, in part, a symptom of not having those harder measures in place. When you don’t have a clear commercial anchor, you argue about who gets credit instead.

The choice of attribution model is not a technical decision. It is a commercial one. It reflects what your business believes about how customers make decisions, which parts of the funnel you think matter most, and how you want to allocate budget. Those are strategic questions, not analytics questions.

A useful discipline is to run multiple models simultaneously and look at where they agree and where they diverge. The channels that look strong under every model are probably genuinely strong. The channels that only look good under one specific model deserve more scrutiny. Semrush’s overview of data-driven marketing covers how to build a measurement culture that goes beyond single-model reliance.

The deeper issue is that all rule-based attribution models share a fundamental limitation: they can only attribute credit across touchpoints they can see. Cross-device journeys, offline interactions, word-of-mouth, and the cumulative effect of brand exposure over time are largely invisible to them. Attribution models measure what they can track, not what actually drives decisions. That gap matters more than which specific model you choose.

How to Choose the Right Attribution Model for Your Business

There is no universally correct attribution model. The right choice depends on your purchase cycle, your data volume, your channel mix, and what question you’re actually trying to answer.

For short purchase cycles with limited touchpoints, last-touch or time decay are defensible starting points. For longer consideration cycles, position-based or linear models will give a more balanced view. For high-volume accounts with strong conversion data, data-driven attribution is worth testing. For businesses that want to understand acquisition efficiency specifically, first-touch is a useful supplementary lens.

The more important discipline is to be explicit about what your chosen model is biased towards and to build that awareness into how you interpret its outputs. Every model has a thumb on the scale. Knowing which way it’s pressing is more valuable than pretending the scale is neutral.

It’s also worth being clear about what attribution models cannot do. They cannot tell you whether a touchpoint caused a conversion or merely correlated with it. They cannot measure incrementality. They cannot account for the channels you’re not tracking. HubSpot’s argument for marketing analytics over web analytics gets at this distinction well: the goal is business insight, not just click accounting.

Early in my career, I was running a paid search campaign for a music festival at lastminute.com. We saw six figures of revenue come in within roughly a day from what was, by today’s standards, a relatively simple campaign. It was exhilarating. But even then, I knew that paid search was capturing intent that had been built by other means: PR, word of mouth, the brand equity of the festival itself. The attribution model was giving us credit for work that had been done elsewhere. We were the last mile, not the whole experience.

That experience has stayed with me. Attribution models are useful tools for approximation. They are not a substitute for understanding how your customers actually make decisions.

For a fuller picture of how attribution fits within a broader measurement approach, including how to connect channel performance to commercial outcomes rather than just conversion counts, the Marketing Analytics hub covers the frameworks and tools that make that possible.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions