You need to understand which marketing activities make a dent in your bottom line. Marketing attribution models are one way to do this. They’re powerful tools for measuring marketing performance, and you can determine the level of nuance in your measurement based on which model you choose.
However, they shouldn’t be the only way you measure.
For a while, attribution has been the shining star of measurement, and most marketers still rely on it. But they’re doing so while also exploring new approaches that don’t rely so much on first-party data, which is all the rage but hard to collect.
This article dives into the different attribution models and what they’re used for. You’ll also learn how to view those models as part of a larger measurement framework that will fill the gaps left by using attribution alone.
What are marketing attribution models?
Marketing attribution models give each consumer touchpoint a certain amount of credit for closing a sale. When you choose an attribution model, you decide how influential each touchpoint is — including both offline and digital attribution.
Some attribution models focus on one touchpoint, and others incorporate multiple touchpoints to get a clearer picture.
If you were trying to decide who was responsible for the fact that you didn’t get any pizza at a party, you might want to blame the person who took the final slice. In that way, you’re attributing the end result — your hunger — to whoever took that last slice. It’s a simple strategy. However, it’s probably not that accurate.
When you glare at that one individual, they shrug their shoulders, knowing there were other people who contributed to your empty stomach. So you decide to assign equal amounts of blame to everyone who ate some of the pizza. Or maybe, knowing one of the party-goers tends to have a big appetite, you assign 30% of the blame to them, and 10% each to everyone else.
Ultimately, you use your knowledge of past eating habits, the number of people who took a slice, and other factors to determine how to feel about the other party-goers. Likewise, marketing attribution models look at data for things like past channel performance and the likely impact of each channel to determine how to assign attribution in a meaningful way.
The benefits of marketing attribution models
Attribution models have been a staple of marketing measurement for a reason. They’re beneficial when you have enough data to power them. They ultimately can:
- Prove your worth outside sales touchpoints by finding the interactions that have the biggest impact on conversions.
- Find which creatives are underperforming to optimize your campaign. This will allow you to justify adjusting messaging, targeting and channels.
- Shed light on the customer journey by forcing you to map out touchpoints that lead to conversions, which helps you prevent biased decision-making.
- Help you secure stakeholder buy-in by providing clear data to justify your marketing decisions.
11 types of attribution models
Different attribution models work better for specific situations. Each has its pros and cons, and what’s right for your measurement journey depends on your goals, data and systems.
Simple attribution models
Simple attribution models give all the credit to one touchpoint, usually the first or last click before a purchase. They’re easy to set up and track, but they oversimplify things by ignoring other important interactions.
1. First-click attribution
As the name suggests, first-click or first-touch attribution assigns all credit for a sale to the very first interaction a customer has with a brand. This model is considered a simple attribution model and is easy to set up and track.
First-click attribution helps marketers quickly see which channels are effective for driving initial brand awareness. However, this approach has a big limitation: it overlooks the impact of any other interactions that may have influenced the customer’s decision to buy.
How a first-click attribution model works.
Imagine a car manufacturer is launching a new model. A potential customer learns about the car through a banner ad and clicks to read more on the website. In the following days, they see promotional emails, TV commercials and even visit a dealership.
Eventually, they decide to buy the car. With a first-touch model, all the credit for this sale would go to the banner ad, ignoring the impact of every interaction that came afterward.
2. Last-click attribution
Last-click or last-touch attribution assigns all credit for a sale to the final interaction a customer has with a brand. This single-touch model is also simple to set up and track.
Last-touch attribution helps marketers understand which channels are most effective at closing sales, but just like first-touch attribution, it ignores the influence of earlier interactions that played a role in guiding the customer toward purchase.
How a last-click attribution model works.
Say a retail store is promoting a new clothing line. A customer sees a social media ad, visits the store to browse and receives several promotional emails. Finally, they click on an email offer and make an online purchase. With a last-touch model, all the credit for this sale would go to that final email, overlooking the impact of every interaction that led up to it.
Weighted attribution models
The rest of the models we’ll cover are multi-touch attribution (MTA) models. They distribute the credit for a sale across several touchpoints to provide a better view of which channels contributed to a conversion. They’re harder to implement than single-touch models but are more accurate.
3. Linear attribution models
Linear attribution models give equal credit to every touchpoint with a customer. They’re relatively simple to set up and allow each channel to contribute. However, they don’t show which touchpoints had the most impact, so you won’t gain a granular understanding of channel impact.
How linear attribution models work.
A fitness brand launching new workout gear might run a Google Ads campaign. A customer clicks on an ad, checks the website and signs up for a fitness tips newsletter. Over the next few days, they get emails with workout tutorials, see a retargeting ad on Instagram and finally purchase after getting a limited-time discount email.
A linear attribution model would split the credit for the sale equally among all these steps, even if some of them influenced the decision more than others.
4. Rules-based attribution models
Rules-based attribution models are multi-touch, meaning they spread credit across multiple interactions. The difference is that you can set the rules on which touchpoints deserve more weight. It gives you more control over how credit is assigned, but it takes a bit more strategy to set up.
How rules-based attribution models work.
Picture a pet food company promoting new cat treats. A customer first discovers the treats through a Facebook ad, and then checks out the website after seeing a Google ad. They finally make a purchase after clicking a retargeting ad on Instagram.
With a rules-based model, the company might give 40% of the credit to the Facebook ad, 30% to the website and 30% to the Instagram ad.
This model is handy if you already know which touchpoints matter most. If you know that early interactions are important for awareness or that last clicks impact a sale, you can assign more weight to those points.
5. Position-based attribution models
A position-based attribution model, also known as a U-shaped attribution model or 40-20-40 attribution, is a multi-touch model that assigns 40% of the credit to the first interaction, 40% to the last and the remaining 20% to any middle touchpoints. This setup helps you see which channels get people interested and which ones actually close the sale while still giving some credit to those mid-funnel interactions.
How position-based attribution models work.
Say a financial services company is promoting a new credit card. A customer first learns about the card through a Facebook ad (40% credit), and then checks out the website to read FAQs and uses a calculator to compare benefits (20% credit). Finally, they apply for the card after clicking a retargeting Google ad (40% credit).
This model works well if you know the first and last interactions tend to be the most important but still want to account for the middle steps.
6. Time decay attribution models
Time decay attribution gives more credit to interactions that happen closer to the actual sale, but it still counts those early steps. It’s great for businesses with long sales cycles — think business-to-business (B2B) or expensive products — because it focuses on the touchpoints that have the most significant impact on closing the deal.
How time decay attribution models work.
For example, say a construction company is promoting heavy machinery rentals. A customer first clicks on a link in an industry newsletter and then visits the company’s website. Over the next few weeks, they see social media ads, attend a webinar and convert after a direct sales call. With time decay, the sales call would get the most credit, followed by the webinar, the ads and the newsletter.
This approach works well for companies where the last few interactions, like a sales call, have a big impact on the sale. It highlights which steps directly lead to conversions in a longer sales journey.
7. W-shaped attribution
W-shaped attribution splits credit between three main steps in a customer’s journey: 30% goes to the first touch, 30% to the lead creation and 30% to the final sale. The last 10% is spread across any other interactions in between. This model helps you see what’s working at each stage.
How W-shaped attribution models work.
Imagine a beauty brand launching a new skincare line. A customer first discovers the product through an Instagram ad (which gets 30% credit). Next, they sign up for the brand’s newsletter after visiting the website (another 30% credit). Later, they get a final discount email and decide to buy (another 30% credit). The remaining 10% covers any other steps in the middle, like reading blog posts or checking out product reviews.
This approach is useful if you want to understand which steps pull in leads while also moving customers toward a sale.
8. Z-Shaped Attribution
Z-shaped attribution splits credit across four major stages in a customer’s journey: the first interaction, when they become a lead, when they show serious interest (like booking a meeting) and the final sale. Each stage gets 22.5% of the credit. There are also touchpoints with minimal impact in between. This type of model works well for industries where customers take longer to make decisions.
How z-shaped attribution models work.
Think about a legal services firm promoting corporate law consultations. A client first learns about the firm through a LinkedIn ad (25% credit), then fills out a contact form on the website after checking out some legal resources (another 25% credit). Later, a rep follows up with a consultation offer (another 25% credit), and finally, the client books a consultation after a call with a legal advisor (last 25% credit).
9. Heuristic attribution
Heuristic attribution is a simple, rules-based approach that assigns however much credit you choose to different steps in the customer journey. It uses set rules, which makes it a good fit for businesses that don’t have tons of data or require an easy-to-apply model.
Let’s say a SaaS company is promoting accounting software. A customer downloads a whitepaper (getting 50% credit as the first touch), then interacts with a chatbot (10% credit) and later attends a webinar (another 40%). In this case, heuristic attribution assigns credit based on rules that give more weight to the beginning and end of the journey.
Data-driven attribution
Data-driven attribution examines all the touchpoints that lead to a sale and assigns credit based on real data, not guesses or set rules. It gives a more nuanced picture of the customer journey by showing how much each step actually matters without having to pick weights randomly.
10. Data-driven attribution models
Data-driven attribution models use machine learning (ML) to assign credit to each step in the customer journey based on how much it influenced the final purchase. Unlike set rules, this model relies on historical data to find patterns and keeps improving as more data comes in, so it constantly adapts based on what’s working.
Think of an e-commerce retailer. A customer engages with a social media ad, visits the website, gets an email offer, clicks on a retargeting ad and finally makes a purchase. A data-driven model might give the most credit to the retargeting ad based on similar patterns it’s seen before.
This approach makes decisions using real-time data rather than fixed rules, giving a more accurate view of what’s driving conversions.
11. Algorithmic attribution
Algorithmic attribution is a multi-touch model that uses algorithms to decide how much credit each step in the customer journey deserves. It considers things like when each interaction happened, the order of steps and how strong each one was. You can customize it to focus on what matters most for your business goals, but it does need more advanced tools and setup.
Let’s say a subscription clothing service is running a campaign. A customer reads a blog post, clicks on an influencer’s video and finally signs up through a link in an email. Since the algorithm is set to prioritize influencer content, it figures out that the video had the biggest impact on the customer's decision to subscribe, so it gives more credit to that touchpoint in the final report.
How to choose an attribution model for your business
Think of picking a model like choosing a camera lens — each shows a different version of the picture.
Follow these steps to help you decide what model you want to use.
- Map your customer journey: Knowing where to stop, from awareness to conversion, helps marketers focus efforts where they’ll have the most impact.
- Collect and transform data: Data must be clean, relevant and organized to be useful. Raw data is messy, with incomplete entries, inconsistent formats and outliers. Once refined, it can be explored and modeled.
- Choose a tool for analysis: Pick one that fits your goals, is user-friendly, integrates with your current tech and can scale with you.
If you stop at just choosing an attribution model, your measurement strategy will fall short of your expectations. The industry is moving away from relying on attribution models after consumer pressure and privacy regulations made it harder to track individual interactions. Attribution models are deterministic — they need a high volume of consumer interactions to work well.
As a result, the industry (and our solution, Funnel Measurement) has moved from deterministic models to a probabilistic approach using anonymized data to spot trends and calculate channel-level contributions.
Funnel relies primarily on marketing mix modeling (MMM), which uses a statistical analysis of historical data to quantify the impact of a given channel on sales. Funnel uses MTA models to complement MMM and refine its findings. Together, they create a resilient, privacy-compliant measurement strategy.
Combining MMM and MTA to explain incremental sales.
This visualization of marketing performance measurement breaks down sales and incremental sales by media channel. Funnel assigns each a contribution to total sales, combining elements of MMM and MTA.
What's better than attribution modeling?
Attribution models — particularly MTA — have been the go-to tools for measuring marketing performance for a while. However, attribution modeling alone only shows part of the picture, and with third-party cookies on the decline, measurement is getting even trickier.
To really understand what’s working, businesses need a mix of approaches, or what’s called triangulation — relying mostly on marketing mix modeling (MMM) complemented by incrementality testing and attribution models.
When you combine MMM’s macro-level insights with MTA’s micro-level precision, you get:
- A clearer understanding of channel-level contributions
- Resilience to changes in data availability
- Compliance with increasingly stringent privacy regulations
With the addition of incrementality testing, you can measure the impact of your campaigns using experiments and control groups.
Triangulation combines MMM, incrementality testing and MTA to support your attribution models.
If you were baking a cake, multi-touch attribution would work to determine how responsible each ingredient (like sugar, flour or butter) is for the eventual delicious dessert.
MMM looks at the relationship between factors that impacted your recipe and the end result. For example, say your friend recommended using a special flour and you decide to mix the cake batter a little longer than usual — both of those factors would have an impact on how well your cake turns out.
Incrementality testing measures the additional value of a marketing campaign by comparing the results with experiment groups and control groups. Think of it as adding cinnamon to your cake recipe to see if people think it tastes better. You’d bake one cake with cinnamon and one without, feed it to different people and determine if the cinnamon made your cake tastier.
When complemented by multi-touch attribution and incrementality testing, MMM becomes the foundation for a trinity of holistic marketing measurement that lets you see your impact at a granular level, down to individual ads, but also zoom out by channel or overall strategy.
Probabilistic models are the future
Probabilistic models are an effective measurement solution in a world where you need to comply with privacy laws, avoid granular consumer-level tracking, scale even when first-party data is limited and be able to make changes quickly across channels.
Deterministic attribution models are still valuable in some contexts, but adopting a probabilistic mindset will future-proof your measurement strategies. By leaning into MMM and triangulation — the integration of MMM, incrementality testing and attribution models — marketers can build a comprehensive view of performance without over-relying on granular data.
Tools like Funnel make it easy to implement triangulation by seamlessly integrating MTA, MMM and incrementality into one platform. Download the Triangulation Tango e-book to learn more.