Big data attribution: why are current models falling short?
In my last blog, I talked about the subject of big data, its meaning and the relevance in the business world.
This blog explores how to choose the best attribution model for analysing big data, and asks – why are so many of the current models are falling short?
What is attribution?
For in-house marketers and analysts in a digital marketing agency, big data can deliver huge volumes of information regarding their interactions with prospective customers, yielding insights that help create strategies to optimise marketing performance. But, the big question is – how do you decide set of data is most important?
Attribution relates to the importance we “attribute” to different datasets when we we analyse them, eg for customer behaviour prediction and segmentation purposes.
There are many new types of data now readily available and many options for database website development. Figuring out an attribution model that works across all media data is an epic task beyond this blog’s scope. So with digital marketing being my forte, I’m going to focus here only on digital media attribution.
As the humorous cartoon by Thierry Gregorius below shows, there can be problems when data is simply taken at face value:
Why are current attribution models falling short?
1. Post-click or last touch attribution
This model is based on the idea that the last channel to persuade someone to click gets the credit for the entire sale. This does seem the most logical because why should others get credit if they weren’t able to generate a sale after the click?
However, in this case, there is no consideration given to the channel that influenced the consumer in the beginning, or any time other than the last click. It might have been a build-up of numerous touch points that encouraged the eventual sale and surely these can’t be ignored.
Post-click attribution has effectively been killed by the rise of site retargeting (the ability to know who has visited your site and then target ads based on what they saw on the site). Site retargeting generates fantastic results — especially when post-click attribution is being used. To illustrate why, consider the following:
A consumer watches a TV ad about a promotion at Tesco and then visits their local Tesco store. In the store, they are handed a flyer about the discount. The consumer goes to the cashier with the discount flyer and then makes a purchase. If the retailer were using last touch attribution, then the conversion is attributed in full to the flyer — not the TV ad that actually drove the consumer to the store in the first place.
However, the way to prove this exists in the online world is actually rather easy. When running a site retargeting campaign, cancel all other traffic and media. Obviously, clicks and purchases will decline. However, there will also be a decline in the click through rate (CTR) of the site retargeting campaign, which proves that something other than the site retargeting campaign is making the campaign perform better and thus should also be getting credit for purchases.
2. Post-view attribution:
This model is based on the idea that the last channel to show a person an ad is the channel that receives credit for it. This model is even more inaccurate than the post-click model mentioned above because 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 an ad. This will still count as an impression and the media partner, typically the one with the largest reach, obtains credit for it.
An example of this model is AOL Instant Messenger (AIM) ads. AIM is typically open on a consumer’s computer screen, so ads are constantly being shown whether or not you’re looking at the AIM screen at that moment. Even when the consumer is on a retail site making a purchase, AIM can be showing the ad and thereby getting credit for the conversion.
3. Having no model at all
This common model allows for all of the media channels to show their digital marketing campaign contributing to the purchase, which then results in claims that several hundred percent more goods were sold than in reality.
So, what’s the answer to this big data attribution problem?
Attribution is not an art or a guessing game; it’s complicated and it’s a big data problem. There are companies that specialise in this, for example Adometry and C3 Metrics. And, some advertising agencies have even developed their own tools to show clients the worth of their campaigns.
The goal of these companies is to provide marketers with a complete picture of consumer touch points and assign weighted values for various levels of engagement throughout the conversion process, with the hope it ultimately ends in a sale or take up of services.
While I am a big advocate of the value of organic search, I – along with many others in the industry – imagine that the demand for digital advertising effectiveness will not die down anytime soon.
Pressure will continue to rise from both sides. First, major companies/advertisers such as Unilever and P&G will push for new attribution models as they shift more pounds to digital. Second, ad technology companies and publishers will advocate for more accurate attribution, as they don’t want to miss out the credit they might deserve or showing off about achievements on behalf of their clients.
An ideal attribution system should ideally measure individual and combined channel performance while providing real-time insights to arrive at optimal scenario and knowledge about points of diminishing returns.
And this is just the beginning because lack of effective measurement and attribution is the number one reason brands indicate they are limiting their digital advertising spend. In conclusion – it’s time to embrace the attribution revolution.
Is your a company curious about attribution and how best to monitor and evaluate campaigns? Why not get in touch and see how I can help.
Jonny Ross is a digital marketing agency specialising in effective website design for both SEO and conversions. I can talk you through the kinds of digital data you should be collecting and how you could use it to inform your marketing activities.