You are probably familiar with the term “attribution”. In this article we will be explaining how an attribution model works and why you should consider it as an integral part of your marketing activities. A data-driven attribution model gives you much better insight into how your marketing kroner actually perform and provides a detailed picture of which particular investment eventually led to a conversion – the actual sale.
The model that tells you exactly ‘what resulted in a conversion’
By Linus Larsson, Head of Digital Marketing and Analytics, Knowit Experience, Gothenburg
Our general advice is to distribute your marketing budget based on data-driven decisions and increased returns!
How most businesses currently evaluate revenues and channels
The most common way of assessing how an online campaign has done is probably to visit Google Analytics or a similar tool and generate a report with traffic sources. There you can deduct your revenues from the respective traffic source and compare the figures with the cost per individual you have reached. Next you can decide whether the returns were above or below target.
In some cases you can also connect transaction data from Google Analytics directly to your promotional tool and read the results there instead. This will often make it easier to adjust your promotions directly in the tool and optimise it based on your returns. The problem with this kind of analysis is that it does not tell you whether the promotion influenced a visitor at an earlier stage than when the actual transaction was carried out.
Let’s look at an example
A company invests in a Google promotion to reach people looking for hotels in Stockholm, say. Somebody in the marketing department decides to activate a remarketing campaign aimed at visitors to the website. Having completed the campaign and analysed the results in Google Analytics, they discover that the Google campaign generated returns of 220%, while the remarketing campaign generated up to 1,800%. The marketing department concludes that the Google campaign failed to yield adequate results to justify continued advertising, while the remarketing campaign is a goldmine that should be given even more investment. Next month you invest twice in remarketing and nothing in the Google campaign. When you analyse the campaign at the end of the month you realise that returns have fallen to 400%. For the sake of this argument, I have exaggerated these figures; the differences are normally not that great, but perhaps you have seen similar trends with your advertising efforts in the past? What did you conclude? Was it the change in marketing tactics in other channels that had an effect, or did you choose to interpret it in light of the bigger budget? Or was it just a coincidence? Now is the time to talk about attribution!
What is attribution?
Attribution defines a value for a touchpoint based on the effect it has on a target that has been set for the activity, e.g. a completed transaction. For example, if you visit a website four times from four different sources and shopped for NOK 800, you can split the value between all four sources and allocate them each a value of NOK 200. This kind of model is called linear attribution, which means that you assign each touchpoint the value of the transaction divided by the number of touchpoints that the visitor used to access the website before making the purchase. There are numerous attribution models, and different models are suitable for different activities.
The most common attribution model is the “last click” model, which is the standard model in Google Analytics. This means that the entire value of a transaction is attributed to the channel that the visitor used to access the website when completing the transaction.
You should now involve one of your marketing people before reading on
The opposite model is the “first click” model which, as the name suggests, ascribes the entire value to the channel that the visitor used to access the site on their first visit. In some cases it may be more important to measure the effectiveness of promotions designed to identify new customers. This model is a good fit for such campaigns. One problem with this model, however, is that a visitor may have accessed the site on multiple devices. Because analysis tools use cookies to track users, a person who uses both their mobile phone and a computer will be registered as two unique users. This means that the first channel that the user emerged from upon use of attribution models may not be the first, only the first on that particular device. The user may see the promotion on their mobile phone and then type in the website URL on their computer to make their purchase. However, there are ways of improving this kind of tracking, including through the use of the Google Analytics user ID.
Two other common models are the position-based model and the theta model. A position-based model attributes 40% of the value of a transaction to both the first and last touchpoint. The remaining 20% is equally distributed between the touchpoints in between. A theta model assigns the lowest value to the first touchpoint before increasing the value for each subsequent touchpoint.
Data-driven decisions trump everything else
There are a number of attribution models available, and Google Analytics permits users to create models based on what is most effective for the company. To complicate things even further, I also though I would explain the data-driven model. This is an incredibly powerful model that analyses every step of the different conversion paths and assigns a value to each touchpoint based on the effect of that touchpoint in the conversion process. This model is not available in the standard version of Google Analytics but is used in Google Analytics 360. Attribution modelling can also be carried out manually but it is a time-consuming and highly complex process. Many people who work with data-driven attribution modelling base themselves on co-operative game theory, which systematically identifies the conditions that generate the best results and determines how the different conversion paths are affected if a touchpoint is deleted. This way you can identify how much value an individual touchpoint should be assigned and can therefore measure the actual effect of a promotion more accurately.
Why attribution should be a given
By adopting an attribution model adapted for your business, or a data-driven attribution model, you will be able to gain a far better insight into your marketing performance. You can see the impact of your different promotions on overall sales rather than just the last touchpoint. Think about it: how many times have you clicked on an ad or a link, arrived at a website and thought about buying something, only to not complete the transaction? Perhaps you will revisit the same website the following day? Would you then type the URL into your browser, or would you google the retailer to access the website and complete the purchase? In such cases – with no attribution model – the business is unable to ascribe the actual transaction to the very advert that aroused your interest in the first place. When you implement a suitable attribution model you can therefore distribute your marketing budget to those channels that actually contribute to the highest sales, thereby significantly increasing your returns.