Ask the Expert – Attribution Modelling
Attribution Modelling can be a challenge. Return on Investment calculations within Marketing are nothing new. But the fact that we now have access to a sometimes overwhelming and ever-growing amount of data at our fingertips means that ROI calculations can be taken to an entirely new level. Especially within the digital space, we can now often gather data on exposure to all sorts of touchpoints.
For some practical insights and pointers, to get your setup right and to assist with your Marketing ROI and attribution calculations, we asked Jon Boon, Lead Analytics Consultant at SearchStar about ‘Attribution’, what it is, how it works and how to bake it into your marketing reporting.
ADLIB: Could you share your take and perspective on “What is Attribution”?
Jon Boon: Put simply, in the world of analytics attribution is the rule, or set of rules, that determines how you give channels credit for sales and conversions.
The problem is that there are multiple attribution models and no one can agree on which model is the best or should be the industry standard. Platforms like Facebook, LinkedIn and Google AdWords understandably want to give their own platforms the most credit.
This can lead to confusion, as adding together conversions from each platform may not equal the total number of conversions. Facebook, for example, will give itself 100% of the credit for a conversion if you click an ad and then convert up to 7 days later.
Furthermore, it will also give itself all the credit if you are shown an ad and then convert within 24 hours. This might sound a little unfair, but what we don’t have visibility of is the impact that a product impression had on the user’s decision to purchase.
Google Analytics, which many businesses use to analyse website performance and aggregate all their media performance, uses a last click model by default, which means that all touch points other than the last one are ignored.
This model is 16 years old and was devised at a time when smartphones and tablets didn’t exist. The issue we often encounter is that last-click is easy to understand and this can result in clients sometimes being averse to moving to a more accurate model which reflects the many touch points involved in a conversion.
ADLIB: Can you give some pointers on “How” it works and how readers can make it work for their own attribution reporting?
Jon Boon: The last-click model used by Google Analytics, while not a fair way to attribute, is the easiest to understand and report on. If you have 100 conversions, you have 100 last-click touch points.
Google Analytics allows you to compare different attribution models; ‘linear’ (all touch points get equal credit), ‘time decay’ (credit increases the close the channel is to the final conversion) and ‘first click’ (channel driving the first session gets 100% of the credit and all others get nothing).
The issue with all of these models is that they are position based, so they don’t help you understand the impact and role they played in users’ decision to convert.
AdWords have recently moved to a more dynamic ‘data-driven’ algorithmic attribution model and Google Analytics is soon to follow suit. A data-driven model uses machine learning to determine how much credit to give to each step in the user journey, from when they first engage with a brand during their research phase, all the way through to the final click of purchase.
It analyses your website’s conversion patterns and compares them to your visitors that do and don’t convert, so the results are unique to your business. As an example, you might find that organic is the best as the last touchpoint in converting your visitors, but social is the best at generating initial brand awareness. This level of insight will help reallocate marketing budgets to where they will be the most effective.
Furthermore (and this is the part where you can use the phrase ‘gamechanger’), the data-driven attribution model that AdWords and Google Analytics will use will include cross-device tracking, so you will be able to understand the role that each device plays in the user journey.
This cross-device tracking is different from the User-ID tracking previously offered by Google (which we believe is extremely limited) as it is dependent on the user being signed into their Google account on all their devices.
ADLIB: Now practically, do you have any recommendations for marketers, to bake attribution modelling into their marketing reporting, for digital channels and also to report on all their marketing ROI?
Jon Boon: You need to use the tools at your disposal. For example, Google Analytics has conversion reports that allow you to examine “assisted conversions” and “multi-channel funnels”. These reports allow you to interrogate the data to see how your channels perform alongside one another.
However, our preferred report is the “model comparison” report which enables users to see the performance of one model against another in terms of conversions and revenue. By using this report it reminds you that one model isn’t necessarily “right” and looking at them all helps you uncover more insight.
The key to being able to bake attribution into an organisation is education. Last-click is simple to understand and it makes reports easy for everyone to digest. An effective attribution model isn’t going to be as black and white, so you need to ensure that all levels of an organisation understand why moving away from last-click is important.
ROI reporting continues to evolve within Google Analytics. For a while now, by using the data import function within Google Analytics, you can import the costs associated with each digital channel to help ascertain their ROI. Building on this, the new attribution tools that Google Analytics will soon be releasing will help the true ROI of each channel become much clearer
Thank You Jon for sharing!
If you’d like to catch Jon and learn more about the future of analytics and conversion optimisation, our friends at SearchStar are hosting a half-day conference ‘Analytics and Conversion, 2019 & Beyond’ in Bath on Thursday 28th June 2018. For more information, you can register your interest here >