Earlier today, Google announced that they would be rolling out a free data-driven attribution modelling tool to their customers. Here, I take a look at what Google Attribution says about Google’s vision and the future of their advertising business.

Attribution, the method by which you understand the relationship between your marketing spend and the revenue it generates, is extremely important. For many companies, it dictates how their marketing budget is spent. There are numerous models which can be used to model the relationship between your marketing and your customer’s activity with your brand but the most common ones are variations of last-click; the last piece of marketing that a user interacts with has the full value of the sale attributed to it. It’s a very simple model, it’s a very basic model and one of it’s variations is the default model in Google Analytics.

Image demonstrating last click attribution
Image Source: raak.be

By making Google Analytics free and easily accessible, it gained extremely deep market penetration – up to 50 million websites are using GA to give website owners data about their visitors. Thanks to this penetration, it has also allowed Google to define the paradigm of web analytics and attribution, often in ways which ultimately benefit Google more than it benefits you.

By default, Google’s variation of last-click attribution is last non-direct click attribution. This means that the last piece of marketing that a user interacts with receives the full credit for the sale and if the user’s final interaction is to visit your website directly, Google Analytics will fall back to the second-last marketing interaction. Last-click and last-non-direct-click models will generally favour lower-funnel marketing activity such as the search advertising that makes up so much of its revenue. By holding such a dominant position in lower-funnel marketing and analytics; Google can effectively mark their own homework.

Image Demonstrating Last Non-Direct Click Attribution
Image Source: raak.be

But over time, the established paradigm – that Google itself helped to establish – became troublesome. Google’s advertising moved into display and YouTube whilst non-traditional, digital-first advertisers stayed wedded to last-click models and the lower-funnel marketing they encourage. Google Analytics gives little incentive for it’s smaller clients to move up the funnel with them. For Google to convince the masses that YouTube advertising is the way forward, it needed to reinvent the wheel of web analytics. Today, it’s taken a huge step forwards with Google Attribution; a free and straightforward tool designed to move the majority of companies from last-click attribution models to multi-channel models based on machine learning.

Now, I’ve yet to see anything of Google Attribution beyond what was announced in the keynote speech but the aim appears clear. As it did with the launch of Analytics, Google intends to use its significant resources to define the way that people view web analytics and attribution with the aim of proving the value of Google’s advertising products, specifically those beyond search.

In the keynote, there was a brief explanation of how straightforward integrating Google Attribution is. The tool can take in data about Google’s various products and then push data back out. Integrating Adwords with Attribution will no doubt be a very straightforward process. There were no specific details given, however, about integrating with other channels. This is vital because the specific data that you collect will dramatically skew your results. Now, it will be possible to collect data about the clicks that drive each channel to your site, probably using UTM parameters but there is more to online advertising than just clicks. For display advertising and YouTube, impression-level data is just as important. If Google offers impression-level integration of its own products, but not of any competitors; then advertisers may find that Google Attribution unfairly overvalues these platforms over their competition.

Demonstrating how Google Attribution works
Image Source: Google Adwords Blog

The above is just one example, but it represents a big problem: ultimately, Google has a trust issue to overcome. They are asking us as a community to let them be the arbiter of what types of advertising provide value whilst at the same time being primarily funded by advertising revenues. Yes, in the past, they have managed to make last non-direct click work to their advantage, but last-click models are, at least, simple. Data-driven attribution is a black box and we are at real risk of putting a huge amount of control and power in the hands of a company who really shouldn’t have it.

The parties who should most be concerned are Microsoft and – especially – Facebook, Google’s main competitors in the online advertising space. Facebook, since acquiring Atlas, have dabbled with attribution modelling through a variety of beta tests and can legitimately claim to have the greatest ability to identify the same user across multiple devices – a key tenet of doing data-driven attribution well. But, unlike Google, Facebook have no pedigree in this space. Google Analytics is a source trusted by so many that it’s easy to ignore the stark conflict of interests involved. For Facebook, they don’t have the legacy of skilled PR that Google have. Any attempt to move into this space could look self-serving and disingenuous.

Attribution as a concept is still in its infancy and the data and modelling behind it will take huge strides forward in the next five years. The fact that Google is attempting to bring it to the masses for free is only going to help the cause but we need to be really careful. Good marketers are platform agnostic, they recognise the risk of losing out by focusing their efforts on only one or a few channels. They would never believe that the data that they receive from Adwords, Bing or Facebook is a single source of truth. Maybe we need to take a similar approach to our attribution.

Leave a comment

Your e-mail address will not be published. Required fields are marked *