Multi-channel attribution: how to truly know which channel drives the most sales

At InnovaDeluxe, we have spent more than fifteen years analyzing ecommerce data, and there is one conversation that repeats itself surprisingly often: a client tells us that 80% of their sales come from Google Ads, and when we dig deeper into how they are measuring that, we discover they are making investment decisions based on an attribution model that completely distorts reality. They have paused email marketing campaigns that “weren’t converting”, reduced their SEO budget because “it didn’t appear in reports”, and are overinvesting in the last click before the purchase without understanding what happened beforehand.

That mistake, multiplied by thousands of euros in monthly investment, is exactly the problem we want to solve in this article.

What multi-channel attribution is and why it matters

When a customer buys from an online store, they rarely do so after a single interaction with the brand. The real customer journey is usually far more complex: they discover the product or service through a blog article ranking on Google, days later they see an Instagram ad, later they receive an email with an offer, and finally convert by clicking on a branded search result.

Who should we attribute that sale to? The SEO strategy that generated the first interaction? The Instagram ad that maintained interest? The email that created urgency? Or the branded SEM campaign that closed the conversion?

Multi-channel attribution is precisely the set of methodologies that allow us to answer that question in a reasoned way. And the answer has direct consequences on how we distribute our marketing budget.

The attribution models that exist and their limitations

Before talking about solutions, it is important to understand the problem in depth. There are several attribution models, and each one tells a different story using the same data.

Last click. This is the default model in most platforms and the most misleading one. It assigns 100% of the credit for the conversion to the last channel the user interacted with before purchasing. It systematically favors closing channels and completely hides discovery and consideration channels. In our experience, it is the model that generates the most wrong decisions.

First click. The opposite extreme. It attributes all the value to the first touchpoint. Useful for understanding which channels generate discovery, but just as partial as the previous one because it ignores everything that happens afterward.

Linear. It distributes credit equally across all touchpoints in the journey. Conceptually fairer, but it assumes that all channels contribute equally, which is also unrealistic.

Time decay. It assigns more weight to touchpoints closer to the conversion. It has some commercial logic, but it still penalizes awareness channels that are essential for feeding the funnel.

Position-based. It gives more weight to the first and last interaction while distributing the rest among intermediate touchpoints. It is one of the most balanced traditional models.

Data-driven. This is the most sophisticated model and the one we recommend whenever conversion volume allows it. It uses machine learning algorithms to analyze all real conversion paths and assign weight to each channel according to its actual statistical contribution. Google Analytics 4 implements it natively and, today, it is the most accessible starting point for most ecommerce businesses.

How attribution works in GA4 and what we should configure

Google Analytics 4 represented a major paradigm shift compared to Universal Analytics, especially in this area. The default model in GA4 is data-driven attribution, although it requires a minimum number of monthly conversions to activate. For stores with lower volumes, it uses the last non-direct click model as an alternative.

The first thing we should do in GA4 is verify that all traffic sources are configured correctly. This means consistently tagging all campaigns with UTM parameters: source, medium, campaign, content, and term. If we do not tag our emails, social media posts, or influencer campaigns, that traffic will appear as direct or unclassified referral traffic, and the attribution model will work with incomplete data.

We must also properly configure the conversions we want to measure. Not only the final transaction, but also the micro-conversions that help us understand the funnel: newsletter signups, add-to-cart actions, checkout starts. The more correctly configured events we have, the richer the information the attribution model can process.

The conversion paths report in GA4 is one of the most powerful and least visited tools among our clients. It shows exactly which channel combinations appear most frequently in journeys that end in purchases, how many touchpoints those journeys have on average, and how much time passes between the first interaction and the conversion.

The problem with direct traffic and how to interpret it

One of the biggest headaches in any attribution analysis is direct traffic. In theory, it represents users typing the store URL directly into their browser. In practice, it is also the catch-all bucket where all visits we failed to tag properly end up.

Emails without UTM parameters, links shared in messaging apps, traffic coming from downloaded PDFs, and some clicks from social media mobile apps: all of this can end up classified as direct traffic. If your store has more than 20–25% direct traffic, it is worth performing a tagging audit before drawing conclusions about channel performance.

How to integrate data from external platforms

The major limitation of GA4 is that it only sees what happens inside the ecosystem of our store. Advertising platforms have their own attribution models, which are almost always more favorable to themselves. It is very common for the total conversions reported by all platforms combined to greatly exceed the store’s actual sales. We call this attribution overlap, and it is completely normal.

To get a more honest view, we recommend working with data from three sources simultaneously: GA4 as the central reference point, each platform’s native data as a relative performance indicator within that platform, and the store’s real sales data as the ultimate source of truth.

Independent attribution tools such as Northbeam, Triple Whale, or Rockerbox allow businesses to aggregate data from all sources and apply their own attribution models. They become a worthwhile investment once advertising spend reaches certain levels, generally above five thousand euros per month in combined ad spend.

What concrete decisions good attribution allows us to make

All this technical work has a single purpose: making better investment decisions. When we have an honest view of each channel’s contribution, we can identify discovery channels we were undervaluing, detect remarketing campaigns we thought converted strongly but were actually just closing purchases that would have happened anyway, understand how long customers need to make decisions and adapt touch frequency accordingly, and justify investments in SEO or content whose results are difficult to attribute in the short term but feed the entire funnel.

Perfect attribution does not exist. There will always be offline conversations, recommendations between friends, or advertising impacts we can never track. The goal is not absolute precision, but having a model good enough to make better decisions than we would make using last-click attribution alone.

At InnovaDeluxe, we always tell our clients the same thing: it is not about finding the channel that sells, but understanding how channels work together. Because in a mature ecommerce business, no channel sells on its own. They all depend on each other. If you would like help with a digital marketing strategy that helps you sell more, contact us.

Related Posts