Attribution Models: Understanding the Value of Paid Traffic

Attribution Models: Understanding the Value of Paid Traffic

Attribution Models - Understanding the Value of Paid Traffic

With the level of detailed measurements available in digital advertising, it’s sometimes hard to know where to start analyzing your marketing campaigns’ performance.

That’s where attribution models come in. Performance can be segregated into different channels or campaigns and help the advertiser understand where conversion value should be distributed.

So, What Is It?

When a conversion is carried out, that value will be assigned to a channel/ campaign. The attribution model you use will determine how this is assessed.

So, for example, the traditional model was the Last Click. This means that the last interaction a user has with a website, the channel or campaign that facilitated that visit will be awarded 100% of the conversion value.

Another commonly used attribution model is Linear. Each interaction a user has with the website leading to the conversion will be awarded equal value.

If a user arrives at a website via a paid ad, then later comes by an organic listing before converting, both channels will receive 50% of the value. With the Last Click, organic traffic would have received 100%.

The multi-touch models take into account every interaction on the conversion path:

Linear Model – Attributes the same share in conversion to all interactions;

Time Decay Model – Attributes more value to interactions that take place in the shorter time before conversion;

Position-Based Model – Assigns 40 percent of the conversion value to both first and last interaction, and the remaining 20 percent is divided equally between the middle interactions;

Data-Driven Model – Algorithmic model analyses all conversion paths in the account and assigns relevant values to interactions depending on their importance to the conversion process (available in Analytics 360 and simplified version) in Google Ads DoubleClick Search).

Why Is It Important?

With an increasing number of touchpoints in a user’s conversion journey, you must understand which ones have been most influential in getting that conversion. By getting this right, you will be able to make decisions based on accurate data.

When a conversion is carried out, that value will be assigned to a channel/ campaign. The attribution model you use will determine how this is assessed.

With an increasing number of touchpoints in a user’s conversion journey, you must understand which ones have been most influential in getting that conversion. By getting this right, you will be able to make decisions based on accurate data.

However, get it wrong, and it will often mean under-reporting, resulting in poor judgment about what is and isn’t working. An advertiser’s work and effort into a campaign can be wasted if the conversion value is not fairly reported back to the most influential interactions — often paid channels.

An analogy often used is a football team. A striker will not score any goals without their teammates assisting in the process. If three players pass the ball before passing to the striker who scores, are they all worthy of a degree of credit? This, effectively, is attribution modeling.

Challenges of Attribution Models

1. Knowing which attribution to model

There is no ‘right’ answer when selecting the attribution model to use, just the one that best represents the fairest distribution of value.

Many factors need to be considered: the business type, industry, target market, and the type of conversion you’re looking to achieve.

With eCommerce businesses, many operate in competitive industries where other retailers sell the same or similar products.

The fitness and health industry, for example, is highly competitive with up to 300K monthly searches for a single keyword (“gym”) in the UK alone, and the highest CPC on a keyword is clocking in at £78.31 for the keyword “box CrossFit.”. And according to our research, the competition is likely to get more complex over time.

Source: SEMrush CPC Map for the UK

Those high CPCs won’t be changing any time soon, at least not in the direction that we’d like. The competition is getting even more challenging, and we will continue to see increased CPC over time.

With this in mind, eCommerce businesses must compete against multiple rivals, using several channels and platforms to connect with customers.

The customer’s buying process is often much more complicated to accommodate the research phase. Therefore, the right decision must be made when selecting an attribution model(s).

But with more touchpoints to consider, it can be a difficult choice to make.

2 Complexity of the customer journey

With more information and purchasing choices available, understandably, the buying process is continuing to get longer and more complex.

You can see some of the journeys carried out by users before

a purchase is made in the image below. You can also see the number of conversions and the total revenue each path generated. With many touchpoints, it’s almost impossible to single out one attribution model that fairly represents all conversion path variations.

The traditional default attribution model

As mentioned earlier, a commonly used attribution model is Last Click or a similar version of it.

The reason for this is its simple design, which allows people to understand how conversion value is distributed clearly. As a result, it’s used across many mainstream reporting platforms as the default model.

However, despite its clarity, the Last Click is also heavily biased towards the last interaction and offers no fair representation of other essential interactions that assisted in a conversion.

Over time, different attribution models have been, and continue to be, introduced to try and remedy it.

Comparing different attribution models

To demonstrate how different attribution models represent conversion value in various ways, we have provided examples. The first case shown is an eCommerce fashion retailer; the image below shows results over three months in 2017. In it, we compare the traditional Last Click model and the Linear model (mentioned above), and the metrics used are conversion value and ROAS.

As you can see, the direct channel shows up very favorably when using the Last Click, whereas other channels such as paid and email are less impressive.

However, when using a Linear model, paid search increases 23% conversion value, receiving more credit for those previous interactions.

The consequences of only using the Last Click could be severe, mainly if you were to make key decisions — such as increasing or reducing ad spend — based on its results.

Looking at the same business, on the image below, you can see that the second most common conversion path is paid > direct. If using the Last Click model alone, the paid search would get no credit whatsoever.

This highlights the potential risks and misrepresentation that the wrong attribution model can display.

In the image below, we show different eCommerce business data. Using the same metrics, we now highlight the difference between the Last Click model and Time Decay.

Time Decay distributes most of the conversion value to the interaction closest to the point of sale and gradually decreases the value of the interactions the further away they are.

Again, you can see that credit is equitably distributed across specific channels with Time Decay, although it’s not as evenly spread as Linear.

So, what would the result look like if you used that model too?

Unsurprisingly, paid search, which provides many of the early interactions, is assigned much more value with a Linear model than with Time Decay.

So, when we use a First Click model, which operates opposite to the Last Click, assigning all values to the first interaction, you can see that paid’s conversion value increases even further.

Attribution models interpret the same data set in various ways, painting different pictures for each channel or campaign performance.

Therefore, this highlights the importance of understanding each model and which one(s) best represents your data.

What It Means to You as a Marketer

When considering how the paid channel acquires users and what stage they come in the conversion funnel, we can see that using the Last Click model is not the fairest when reviewing performance. It also overweighs credit to channels that typically acquire returning visitors, such as direct.

There isn’t a ‘correct answer when assigning an attribution model to analyze paid traffic performance. However, you need to select one that supports early interactions as this is where the main paid touchpoints will be found.

Our recommendation is to use an attribution model that supports multiple touchpoints rather than one singular. So, Linear or Position-Based could be appropriate choices. Those who use Google Ads may access a Data-Driven Attribution Model (DDAM), which is fluid, segmenting each conversion’s value in a bespoke manner to different interactions.

According to SEMrush, Google has reported that the DDAM can improve the accuracy of assigning conversion value by up to 15%.

If you want to assess best what marketing efforts and PPC campaigns are working most effectively for you, you should:

1 Choose the suitable attribution model. – For most businesses, this will mean ditching the Last Click attribution model, as it fails to account for all events leading up to the conversion and thus doesn’t give an accurate portrayal. Using Linear or Position Based models will take multiple touchpoints into account.

2 If available to you, use the Data-Driven Attribution Model for your Google Ads. It’s more fluid and, thus, often more accurate.

3 When deciding on an attribution model, look through the current analytics that you have set up. If you have the default, the Last Click enabled, for example, and it’s telling you that only search traffic is getting you conversions, even though you see significant clicks from Google Ads to those product pages, you’ll likely realize that the attribution model isn’t accurate.

To find the best one for you, you may have to test out different models to see which one helps you make sense of conflicting data.

Conclusion on Attribution models

Using attribution models, you can see which aspects of your digital marketing are working and which aren’t, so you know what you need to optimize and where you can stay the course.

Relevant case studies of Todays.Agency

What do you think? Which attribution model makes the most sense for your business?

Please share your thoughts in the comment section below.

Looking to improve your business and increase your sales?

Get a free quote