Every time a customer buys something from a brand that runs digital advertising, something called an attribution model decides which ad gets the credit for that sale. This sounds like a technical detail. It is actually one of the most consequential decisions in marketing, because it determines which channels receive more budget, which campaigns get celebrated in the next quarterly review, and which ones get quietly cut.

Most brands are using a model that consistently gets this wrong.

Last-click attribution — the system that gives 100% of the credit for a sale to the last ad a customer clicked before purchasing — is still the default setting in most advertising platforms and most analytics dashboards. It is simple to understand and easy to implement. It is also, for most businesses, a distorted picture of how customers actually make purchase decisions. And when you make budget decisions based on distorted data, you end up systematically underfunding the channels that build your business while overfunding the channels that just show up at the end.

Understanding why this happens — and what to do instead — is one of the highest-leverage improvements any marketer can make to how they manage advertising.

Why last-click attribution is wrong

Think about how you actually decide to buy something. You do not click one ad and immediately purchase. You see a brand on Instagram. A week later you hear it mentioned in a podcast. You Google it, read a review, and click on one article. Another week passes. You see a retargeting ad on a website you were reading. You click that ad, visit the product page, and buy.

In last-click attribution, the retargeting ad gets 100% of the credit. The Instagram post that introduced you to the brand gets zero credit. The branded search that showed you Google cares about this brand gets zero credit. The content article you read gets zero credit.

Now imagine that the brand’s marketing team looks at this data and makes budget decisions from it. They cut Instagram spending because the data shows poor return. They cut content marketing because it never shows conversions. They scale up retargeting because the ROAS looks great. The retargeting campaign then performs terribly because there are fewer new customers entering the funnel from Instagram and content. The outcome looks like a mystery — retargeting stopped working — but the cause was following the incentives that last-click attribution created.

This pattern plays out across brands of all sizes, all the time. The model creates a feedback loop that gradually starves the top of the funnel while overinvesting in the bottom.

The alternatives that actually exist

There is no single perfect replacement for last-click attribution. There are several imperfect tools that, used together, give a much more realistic picture than last-click used alone.

Multi-touch attribution (MTA) distributes credit across multiple touchpoints in the customer journey. There are several variations — linear models spread credit equally across all touchpoints, U-shaped models give more credit to the first and last touch and less to the middle, time-decay models give more credit to touchpoints closer to the purchase. Modern versions use machine learning to weight touchpoints based on statistical analysis of which combinations of ads actually correlate with conversion.

MTA is better than last-click for understanding which channels are contributing to the funnel. Its limitation is that it requires tracking the full customer journey at an individual level — which, post-iOS privacy changes and cookie deprecation, is increasingly difficult for a significant portion of your audience. As we covered in our piece on what replaced third-party cookies in 2026, the data trail that connects a social impression to a downstream purchase has gotten significantly shorter for a large share of any audience.

Marketing Mix Modelling (MMM) takes a completely different approach. Instead of tracking individual user journeys, it analyses the relationship between your total advertising spend across all channels and your total sales outcomes over time, using statistical regression. It does not need to track individual users at all — it works from aggregated data. This makes it immune to privacy changes, iOS opt-outs, and cookie blocking.

MMM’s limitation is that it requires significant historical data — typically six to twelve months of spending and sales data — to produce reliable results, and it tells you about your past decisions rather than giving you real-time campaign-level guidance. It is a strategic planning tool, not a dashboard you refresh daily.

Incrementality testing is the most rigorous approach for validating whether a specific channel or campaign is actually causing additional sales, rather than just correlating with sales that would have happened anyway. As we explored in the context of influencer marketing measurement, the difference between correlation and causality is everything when you are deciding whether to scale a channel. A geo-based incrementality test — running a campaign in some regions while holding matched regions as controls, then comparing sales — directly answers the question “would these customers have bought from us even without seeing this ad?”

What most brands should actually do

The answer is not to immediately build a sophisticated multi-model attribution infrastructure. Most brands do not have the data science resources or the historical data required for that. The answer is to stop treating last-click as the source of truth for budget decisions, and start using a simple combination of inputs.

For day-to-day campaign management, use the attribution data from your platforms but apply healthy scepticism to any channel that consistently appears at the end of the funnel. Retargeting and branded search will always look great in last-click models because they appear just before purchase. That does not mean they are your most valuable channels — it may just mean they are the channels that capture intent that other channels created.

For quarterly budget decisions, look at your total spend by channel alongside your total sales, and look at how those numbers move together over time. If you cut Instagram spending by 40% and branded search volume dropped three weeks later, that is a signal that Instagram was driving demand that was then captured by search — not that search drove it independently.

For validating a channel you are considering scaling significantly, run a geographic incrementality test before you commit major budget. It is the most direct way to answer whether the channel is actually generating new customers or just claiming credit for customers who were coming anyway.

The friction you need to prepare for: When you start looking at attribution data more critically, your most efficient-looking channels will become less defensible and your upper-funnel spending will not have obvious numbers to point to. This creates internal political problems — brand managers whose channels suddenly look less productive, and platform reps who will tell you their attribution shows excellent results. The honest answer to both is that better measurement makes every channel look more complicated, not simpler. That discomfort is a sign the measurement is improving, not that something is going wrong.

Attribution is not a technology problem. It is a thinking problem. The technology can be improved, but the most important change is simply stopping the habit of accepting the number in the dashboard as the truth, and starting to ask what question that number is actually answering.