In this exciting digital age, everything is measurable, from consumer behaviour to advertising performance. Transparency is paramount, and as
advertising spend increases, so comes an unprecedented expectation of accountability and proving ROI.
Unfortunately, there is no one “path to purchase”. Consumers are exposed to numerous brand
touch points in a variety of orders. It gets tricky when the purchase decision was influenced by a number of these touch points, (which is generally the case). There are many, many paths to purchase that it make it almost impossible to measure all of them.
Enter Attribution Modelling
The concept of attribution modelling is not new. Our ability to measure digital channels is improving, and data driven decision making is fast becoming a critical component of the marketing mix whilst still keeping the consumer in mind. At least that how it should be!
Attribution modelling is fundamentally market mix modelling at an individual level. By tracking the consumer journey through online and offline touch points, it can map the exposure to conversions.
I like this concept; it’s more user-centric and provides a more holistic outlook. At the end of the day, we are measuring behaviour; the consumer’s journey through the purchase funnel.
There are several types of attribution models: here is a brief overview of the main ones, along with their associated risks and benefits.
James Green, the chief executive officer of Magnetic, sums it up rather nicely in his post for Mashable :
1. Full Funnel Attribution
Full funnel attribution is more of a theory than a model solution. In an ideal world, everyone would choose full funnel attribution because it assigns values throughout different stages of a consumer’s experience, providing deep insights into the role that each ad plays.
Consumers move from awareness, to interest, to consideration, to preference and eventually, to purchase. Some people would argue that the first ad and last ad deserve the greatest credit because they’re responsible for initiating awareness and closing the sale.
Full-funnel holds that not all ads are created equal, and that brands should understand the impact that each ad has on creating awareness, influencing brand preference, and driving the desired outcome.
2. Post-Click or Last-Click Attribution
Post-click or last-click attribution is based on the general notion that the last advertising medium to persuade a consumer to click on an ad will receive credit for the entire sale. At first glance, this model appears to be the most logical. After all, why should others receive credit if they were unable to generate a sale after the click? When you dig deeper, this model does not take into account the possibility that a consumer might have been motivated by other ads for the same product in advance of the last click. Specifically, it doesn’t consider a product’s longer-term building of awareness and interest, or the evaluation process that a consumer goes through. Unlike the full-funnel theory that looks at multiple data points, post-click or last-click only considers one data point associated with the last action taken by the consumer just before a purchase is made.
Search marketers tend to favour this model, as it guarantees that there’s no confusion if a consumer clicks on an ad and then makes a purchase. However, this model does not take into account the possibility that other forms of advertising might have piqued a consumer’s interest first.
3. Post-View Attribution
According to the post-view model, the last channel to show a person an ad is the channel that receives credit for it. However, this model is even less accurate than the post-click model, as it encourages media partners to plaster ads as widespread as possible in order to take credit for the conversion, even if a consumer doesn’t actually see the ad. The benefit to post-view is that it enables marketers to measure if the viewing of the ad actually resulted in a positive outcome.
However, the risk associated with this model is that not every ad that is shown to a consumer is actually seen. For example, an advertisement might be posted on a window of a store, but that doesn’t guarantee that a consumer walking by will see the ad. Similarly, a banner ad on Facebook or AIM might be present on one’s computer screen, but might not be noticed. Not all advertisements are created equal.
4. Equal Attribution
A step in the right direction is equal attribution. This is a form of post-view where equal value is assigned to every single ad placement. For example, prior to purchasing a product, a consumer viewed four advertisements from four different vendors. Each ad is then assigned credit for 25% of the sale. However, the risk is that this model assumes that all ads are created equal. Branding campaigns are typically more likely to utilise an equal attribution model, as this model focuses on reach and frequency, over specific types of metrics.
5. Fractional Attribution
In James Green’s opinion, fractional attribution is probably the best solution and the most popular. However, unlike the other options listed above, it’s necessary to work with an attribution vendor in order to effectively measure fractional attribution.
Fractional attribution is the most accurate model available today, and is as close as we can get to practicing full funnel attribution. The trickiness lies in the fact that all companies have different goals when it comes to advertising and whom they are trying to reach. As a result, all attribution funnels are measured differently.
Attribution is only as good as the mind of the marketer who creates the marketing plan, which is another truth of our digital world. The more sophisticated the technology, the more reliant we are on truly smart individuals to plan, implement and interpret.
Attribution vs Marketing Mix Modelling
I. It measures exposure at a very granular level—e.g., individual search keywords, unique display ads, etc.
II. Second, it measures the combinatorial impact of different sets of touchpoints. That is, knowing the impact of A and B, what’s the impact of A+B?—and all other possible combinations.
III. Third, it can measure the sequential impact of advertising exposure. And lastly, it’s fast. The models can be updated daily.”
It makes perfect sense and provides us with great insight into the real path of conversion across the whole purchase funnel. The modelling mix model tells us which touch points are providing the most impact and which combinations have the greatest impact. This is, in essence, the “path to purchase.”
The model also shows where each touch point had the most impact in the exposure. Understanding the order that achieved the highest impact can help improve overall media planning. This enables us to understand whether the impact of media touch point exposure differs by top and middle of funnel brand metrics vs. sales metrics, and, if so, how to optimise one or both.
Attribution modelling is moving the path to purchase and purchase funnel concepts to real world tools, to improve advertising and media performance. This is an invaluable toolset that all marketers should be using to gain enhanced insights into how consumers behave and how advertising translates into conversions.