How does the Data Driven attribution model work?

Attribution models, a tough subject when it comes to understanding and applying them. And with the introduction of more online platforms, the customer journey is becoming increasingly complex. It has become standard for the customer journey to consist of multiple touchpoints on different platforms. Therefore, the challenge in these extended customer journeys is to determine how much each channel has contributed to the conversion.

Many businesses use Google Analytics for data analysis and to view the performance of each marketing channel. However, what many business owners fail to realize is that Google Analytics attributes according to the "Last click attribution model". In this model, all value is attributed to the last click in the Customer Journey. And actually, this is not completely fair. To explain this properly, we have created an example:

"In a football match, a striker cannot score without a good cross from another player. So people have several touch points to reach the final conversion."

Keywords, campaigns or marketing channels that are used earlier in the orientation process, receive far less conversions and therefore have a higher cost per conversion. As a result, many businesses are often tempted not to include them in their marketing strategy. And that is a shame, because these are often the keywords, campaigns or channels that bring the consumer or prospect in touch with your brand or product.

So how can you prevent this from happening and how do you create a fair attribution to the different marketing channels? For this, you can choose to use a Data-Driven attribution model.

What is attribution?

To get a good picture of this topic, we need to start with the basics: "What is attribution?". Attribution is a method by which online marketers assign value to each touchpoint that has contributed to an eventual conversion. So, conversion attribution provides an overview of the influence that various touchpoints and interactions with online channels have had in the customer journey of the consumer. Different attribution models for online marketing There are various models within online marketing that determine how value is distributed. Some of them are shown in the image below.

The choice of an attribution model depends on, for example, the activities of your company, or on the channels you use. Within Google Analytics, the 'last non-direct click' attribution model is usually applied. This model assigns full value to the last channel clicked prior to conversion.

So what is the Data Driven attribution model?

The Data Driven Attribution model is an algorithmic model that makes sure that every click before conversion is taken into account. By using Machine Learning. The resulting data-driven model learns how different touchpoints affect conversion results. It predicts how much impact a channel could have on a conversion, based on the presence, timing, device type, amount of interactions with the ad, order of ad display and type of ad files.

Using a counterfactual approach, the data-driven model contrasts what happened and what could have happened to determine which touchpoints are most likely to generate the most conversions. Based on this likeliness, the model attributes conversions to these touchpoints.

Example: Clothing, the customer journey and the data-driven attribution model

Imagine, while scrolling through your Facebook timeline, you come across an advertisement with a new clothing collection. A few days later, you read a blog with this season's latest trends, using clothes from this brand.

At night, sitting on the couch, you start looking for nice new clothes for the coming season, and you search for different clothing brands. As a result, you end up on many different clothing sites. The many choices make you doubt about what you want to buy. You decide to clear out your wardrobe first to get a better idea of what clothes you need.

A few days and a cleaned-out wardrobe later, you decide to continue your search for new clothes. This time, you search by specific terms, because now you know exactly which colors and items of clothing are still missing in your wardrobe.

You go to the website of a brand that appeals to you, but this brand has clothes with a slightly more expensive price tag. This makes you hesitate and postpone your purchase for a while. On the website, you see that if you sign up for the newsletter, you get a 10% discount. So you sign up.. You receive the email with the discount code, but are now busy with dinner. You decide that you will order the clothes with the discount code later that evening.

Later that evening, when you are sitting on the couch and are scrolling through your Facebook or Instagram timeline, you suddenly see advertisements of the clothes you viewed earlier that day. This reminds you of the fact that you have a discount code in your mail. You use the discount code from the email and order some new clothes. Purchase done!

Based on this example, there are several questions that you ask when you're doing online marketing for the clothing brand. How valuable is each touchpoint in this customer journey? And what impact would it have if one of these touchpoints did not occur? How much value does the blog article add? To answer this question, the conversion factor per touchpoint is very important.

The value of data-driven attribution

The question that probably comes to mind is: 'how is a data-driven attribution model (and thereby the value assignment) defined?

With the data-driven attribution model, the value is determined proportionally, by excluding the contact point in the conversion path. The model looks at what would change in the conversion ratio if the visitor did not come into contact with a certain marketing channel (display, facebook ads, instagram ads, search ads, blogs etc.). This is also called a counterfactual analysis. When putting together the entire data-driven attribution model, there are four layers:

Exposure
In the first layer, value is assigned based on the change in conversion rate when the channel does not occur in the conversion path.

Frequency
In this layer, the value is based on how the opportunity for a conversion changes when a contact point of the same type of channel occurs a certain number of times.

Recency
In the third layer, value is assigned based on the occurrence of the touchpoint in a specific time frame.

Frequency at each value recency
In the last layer, a value is assigned based on how often a channel contact point occurs within a specific time frame.

What are the main advantages and disadvantages of working with data-driven attribution?

Benefits

1. Data-driven attribution provides better insight into which generic campaigns contribute to the final conversion.
2. A large part of the traffic is mobile, and the final conversion does not take place on mobile in many cases. With data-driven attribution, there is better insight into what contribution mobile has made to the final conversion.
3. A shift to the use of more campaigns, marketing channels and search terms at the beginning of the customer journey, because you get a better understanding of their role in the entire customer journey.
4. It offers the possibility of deploying, for example, blogs from a website or more informative content.
5. Finally, the contribution of these campaigns becomes more transparent thanks to the data-driven model.

Disadvantages

1. Campaigns are more difficult to manage because a touchpoint can still be assigned (part of) the conversion up to 90 days back.
2. It cannot be applied to historical data.

Application and impact of the data-driven attribution model

Choosing a different attribution model is not simply clicking on another model. The choice for the data-driven attribution model requires a different view on the conversion path and therefore a consistent and well thought-out implementation.

This blog mainly describes how the data-driven attribution model works, but it is also important to be aware of the impact on other aspects. You can think of changes in the data, the way of analysis and the cooperation between different marketing channels within the customer journey.

If you apply this model, the expectation is that you will see improvements in the upper funnel marketing activities and a reduction in the lower funnel marketing activities.

Would you like to set up Data Driven Attribution for your marketing campaigns or do you have questions about Data Driven Attribution? Then get in contact with us!