Understanding Attribution Models

A customer rarely discovers a brand and purchases in a single interaction. They might first come across you on TikTok, return through Google a few days later, click an email, and eventually place an order. Every one of those touchpoints was part of the journey — but they did not necessarily play the same role.

Attribution is the attempt to understand those roles and decide how much credit each channel should receive for the purchase.

There is no single way to assign credit across a customer journey. One model may focus on where the journey started, another on what closed the sale, while others look at timing or how much a channel actually influenced the likelihood of conversion.

That is why RetentionX calculates the most common attribution models in the Attribution analysis; but without requiring you to choose between them. Our blended RetentionX model brings the strongest signals from these approaches together in a single attribution model. This gives you a more complete understanding of how each channel contributes to performance.

Once enough historical data is available, Retention AI takes this one step further. It trains machine-learning models on your own marketing and revenue history to estimate the incremental impact of each channel, adding a broader view of what is actually driving business performance.

Starting with the Simplest Views of a Journey

The most familiar attribution models look at where a channel appeared in the journey.

First Touch

Full credit to the very first hello.

First Touch gives all credit to the channel that first brought the customer to your store. Imagine a customer journey that looks like this:

With First Touch attribution, Google Ads receives 100% of the credit. It does not matter that Klaviyo brought the customer back or that TikTok was the final interaction before purchase; First Touch is interested in one thing: where did the relationship begin?

That makes it a particularly helpful perspective when you want to understand discovery, awareness, and prospecting, in other words, which channels are successfully introducing new customers to your brand. What it cannot tell you is what happened after that introduction.

Last Touch

The classic: full credit to the final step.

Last Touch looks at the same journey from the opposite end. In our example journey, TikTok receives 100% of the credit because it was the final marketing interaction before the order.

It answers a different question: what brought the customer back when they were ready to buy?

That perspective is especially helpful when you want to understand which channels tend to close purchases. Because Last Touch remains common across analytics and advertising platforms, it is also a useful reference point when comparing RetentionX with third-party platform data.

At the same time, it can make channels such as branded search or email appear particularly strong simply because they often sit close to the purchase, even when another channel created the original demand.

Linear

Every step counts the same.

First Touch gives everything to the beginning. Last Touch gives everything to the end. Linear attribution takes the middle ground and gives every touchpoint the same share.

If Google Ads, Klaviyo, and TikTok were all part of the journey, each would receive one third of the credit.

There is something appealingly simple about that: every interaction that helped move the customer toward the purchase is recognized. It gives you a balanced view of the full journey without making assumptions about which stage mattered most.

The trade-off is that customer journeys are rarely that equal. An interaction from three weeks ago may not have had the same influence as the ad clicked five minutes before purchase. That is where timing becomes important.

Weibull

Recent touches count more — and your own customers decide how much more.

Weibull adds that timing to the picture. Rather than treating every interaction equally, touchpoints closer to the purchase generally receive more credit, while older interactions gradually receive less.

But how quickly should that influence fade? A fixed rule would assume the answer is the same for every business. RetentionX instead learns the timing curve from your own customers' purchase behavior.

If shoppers typically discover your products and purchase almost immediately, the model concentrates more credit around recent interactions. If customers usually consider a purchase for several days or weeks, earlier touches continue to carry meaningful weight for longer.

That makes Weibull especially valuable for brands where the buying journey stretches across days or weeks and the timing of each interaction is an important part of understanding what moved the customer toward purchase.

 

Any Touch

How often was this channel part of the story at all?

Sometimes you do not want to divide the credit at all. You simply want to know: was this channel part of a journey that converted? That is what Any Touch shows.

Take an order that followed our Google Ads → Klaviyo → TikTok journey. TikTok would claim the conversion and it's associated revenue. So would Google Ads. And so would Klaviyo. 

With Any Touch, the same order is intentionally counted for every channel that participated, which means the revenue across your channels will add up to more than your actual revenue.

That is not a problem with the calculation. Any Touch is not trying to divide revenue fairly between channels. It is showing you where channels are involved.

This is especially helpful when you want to uncover the quiet team players: channels that may rarely open or close the journey but appear repeatedly in journeys that successfully convert.

 

 

Moving from Position to Actual Influence

All of the above models make some assumption about how credit should be assigned. Data-driven attribution takes a different approach: it lets the patterns in your customer journeys tell us how important a channel appears to be.

Markov 

What would happen if a channel disappeared?

Being present in a successful journey does not necessarily mean a channel caused the purchase. Markov tries to get closer to that distinction.

Instead of deciding beforehand that the first, last, or most recent interaction matters most, the model learns how customers actually move between channels — including journeys that end in a purchase and journeys that end without one.

It then asks a powerful question: What happens if we remove one channel?

First, Markov looks at all observed customer journeys and calculates how likely they are to end in a purchase. In this example, the probability is 64%.

Next, TikTok is removed from the model. Markov recalculates the remaining journeys, and the purchase probability falls to 35%. That drop tells us that TikTok was playing an important role in helping customers convert.

That drop tells us that TikTok was playing an important role in moving customers toward purchase, even if TikTok was rarely the final interaction before the order.

RetentionX repeats that process for each channel. The more the probability of purchase falls when a channel disappears, the more influence that channel has demonstrated.

This gives you a much stronger view when you want to understand which channels are genuinely driving conversion rather than simply appearing somewhere in successful journeys.

Shapley

Game theory’s answer to “who contributed what?”

Shapley looks at contribution from the other direction. Markov asks what happens when a channel disappears. Shapley asks what does that channel add when it is part of the mix?

Think about Klaviyo appearing across many different journeys. Sometimes it may appear on its own. Sometimes alongside Google Ads. Sometimes with TikTok. And sometimes all three channels are involved.

Shapley compares how the outcome changes across those different combinations. By looking at the value a channel adds whenever it joins a different mix of channels, the model can calculate its fair share of the result.

That gives you another data-driven perspective on contribution and a useful cross-check against Markov — particularly when you want to understand how channels contribute as part of a wider marketing mix rather than in isolation.

Synergy

How do your channels work together?

Attribution models typically ask how much credit each channel should receive. Synergy looks at something different: how channels influence one another across the customer journey.

For every channel, RetentionX looks in two directions:

Influence given looks at which channels customers tend to interact with after this channel. This captures how a channel can create demand that another channel picks up later.


Influence received looks at which channels customers tend to interact with before reaching this channel. This captures how a channel can benefit from demand created elsewhere.

This adds another perspective to attribution by considering how your marketing channels influence one another, rather than evaluating each channel in isolation.

Because Synergy is based on your full observed journey history, its values evolve gradually rather than reacting to short-term changes.

Bringing Those Signals Together

By now, you can see why choosing a single attribution rule is difficult. Where an interaction happened in the journey matters. Its timing matters. The influence a channel has demonstrated across thousands of journeys matters. And there are parts of the customer relationship that cannot be captured by looking at clicks alone.

RetentionX 

The best of every model, combined into one number.

Our proprietary RetentionX attribution model combines the strongest signals from the different attribution approaches into one blended model, rather than forcing every purchase into a single rule. It brings together what the individual models are best at: channel influence, timing, and the customer relationship behind the journey. It also differentiates between new and repeat customers, because acquiring someone for the first time is very different from bringing an existing customer back.

Email is a good example of why that matters. A customer may have been subscribed to your newsletter for months and regularly engaged with your emails before eventually returning through Google Ads and purchasing. Giving Google Ads 100% of the credit would ignore that relationship. But giving Email the full credit simply because the customer received or opened a message would overstate its contribution.

Instead of choosing one of those perspectives, the RetentionX model blends the most relevant signals across them so credit can be shared in a way that better reflects the complete customer journey.

RetentionX AI

Connecting attribution to the big picture.

Once enough historical data is available, RetentionX trains machine-learning models on your own marketing and revenue history. Rather than asking which touchpoint should receive credit for an individual order, the model learns how different factors shape overall business performance; including the impact of marketing spend, promotions, and broader market dynamics.

This helps separate the effect of a channel from everything else happening at the same time. For example, a revenue increase during a promotional period should not automatically be attributed to higher ad spend if the promotion itself was a major driver. That broader perspective becomes particularly important when you are making budget and channel investment decisions and want to understand how much each channel is actually contributing — and what could happen if you invest more or less in it.

Not using RetentionX yet?