Data-driven MTA: The only attribution model that counts
A comprehensive analysis of why data-driven MTA is the superior attribution model vs traditional models.
A comprehensive analysis of why data-driven MTA is the superior attribution model vs traditional models.
Attribution models help marketers understand the value of marketing channel touchpoints as their customers move from awareness to conversion.
In the past, single touch attribution, such as ‘last click’ and ‘first click’ models, were often the default. They required relatively little data and were easy to observe in analytics software.
For instance, if I made 1,000 sales on my website during October and I can see 75% of the traffic coming from organic search and 25% is coming direct – it makes sense to turn more of my November investments to my organic search activities.
That would be adopting the last click model. I am only assigning value to the touchpoint my customers have used immediately before hitting my site. So, what of the channels they’ve interacted with beforehand?
As customer journeys have become more complex in recent years, so too have the attribution models available to ensure marketers are giving accurate weight to the channels along the way and are investing in them accordingly.
Multi-touch attribution (MTA) and data-driven attribution (DDA) are increasingly favored. After all, my website visitors may be seeing marketing messaging in a range of places (from myriad social media sites, to any number of offline ads from TV to billboards) before turning to Google to make their search.
Finnian Bradfield, an analyst at AI data and attribution company Fospha, highlights that it is not only the growing complexity of customer journeys which is causing MTA to be more effective today than single touch models.
“The main value with MTA is to enable marketers to better distribute their spend and understand the true performance of every touchpoint,” he says.
“By using solely, a last click model you essentially only value or see the last touchpoint the customer engaged with prior to conversion and pay little attention or regard to what may have influenced the consumer before this. The result is more spending on the last touchpoint, however in reality that conversion may have never occurred if it weren’t for a touchpoint the consumer engaged with at the start of their journey. So, by looking across all touchpoints and attributing a value across each depending on their role means you can get much more effective results from your marketing spend.”
In today’s omnichannel world, MTA models certainly seem to be the superior choice for ensuring marketers have clarity on what’s working and what isn’t. But not all attribution models are made equal.
Let’s dig down into how they differ, their pros and cons, challenges to implementing them, and why data-driven MTA models are the attribution option all marketers should be aiming for.
The ‘linear’ or ‘even credit’ attribution model is arguably the first MTA model we think of when we start thinking beyond first and last click.
It is certainly an improvement on single-touch models. The linear model acknowledges that there is more to the customer journey than simply the channel the customer sees at the awareness stage, or that which they use in the step before making their conversion.
It credits all touches equally, so is only really useful when you have good evidence that all parts of your marketing strategy are performing to roughly the same level of effectiveness.
But its simplicity quickly becomes very apparent, as it fails to account for the fact that different touchpoints will likely have had more or less of an impact on the customer’s decision to convert than others.
‘Time decay’ is another relatively simple MTA model. It credits all touchpoints on a user journey, with an increased weighting the closer the touch is to conversion.
Time decay attribution can be useful. And it would likely be preferable over one of the single touch options. It is often used with time-sensitive marketing campaigns or when businesses are looking to focus on channels which are playing a converting role. But its flaws are quite easy to see.
Of course, there are customer journeys where touchpoints closer to conversion are more important than those near the beginning. But as user journeys get longer and more complex, this is increasingly not the case. Time decay, ultimately, still relies on a certain amount of guesswork and lacks real accuracy.
‘U-shaped’ – or ‘position-based’ attribution is another MTA model which, again, improves on single-touch methods by ensuring all channels receive some credit, but it is still a relatively simplistic approach.
U-shaped credits all touchpoints but gives more weighting – 40% each – to the first and last channels. The remaining 20% of credit is spread among any touchpoints in the middle of the journey.
Again, the u-shaped model has its uses. It can be effective, for instance, in the case of companies who are investing in lead generation and is a strong candidate for longer customer journeys. But it can fail to accurately credit any touchpoints in the middle of the journey which may have had a bigger part in the conversion than the marketer might expect.
‘Data-driven’ attribution (DDA) is another MTA model. It differs from those we have discussed so far in that it uses data across touchpoints to eliminate any guesswork and to attribute credit to channels by how they perform, rather than by what position they are in.
DDA calls on having comprehensive data and a full view of the customer journey. Marketers also need to have the means to be able to use that data effectively – often by utilising a customer data platform.
They can be the costliest MTA model to implement, but one of the key goals in using a DDA model is seeing a better ROI which can be used to offset their extra cost.
Bradfield is clear about when it makes most sense for marketers to implement a data-driven MTA: “A data-driven approach to attribution is the most effective when you are spending across multiple paid channels that include mixtures of paid search, social and affiliate for example,” he says. “The reason for this is simply down to the fact that if you’re spending in marketing, you need to know where is best to spend and which of these channels are more effective for your business KPIs. We tend to see businesses that have larger marketing budgets gain a lot of value from these types of tools.”
In 2019, Fospha worked with a leading holiday provider to stitch together their online and offline data with MTA.
This gave a cross-channel view of the journeys their customers are making and highlighted that their paid search activity was being overvalued with last click attribution.
The automated process did not eat into analyst capacity or time. But more importantly, Fospha were then able to use freed up budget from the paid search channel for growth in new/undervalued touchpoints.
Savings for the year amounted to around £600,000 and their TV marketing strategy was highlighted as an area that had been previously difficult to measure using traditional attribution models. Using MMM (Marketing Mix Modelling), Fospha was able to identify that TV was a strong performing channel with room to grow and bring in further revenue.
As soon as this was surfaced , the business was quick to re-invest around £250,000 back into this vital offline channel.
“Fospha’s independent measurement tools have empowered our team with transparent data, and the ability to link our offline sales to both our offline and online marketing,” the client’s Group Marketing Director said. “We now have the confidence to invest more in our marketing, and we are excited to see the results.”
Increasingly it is not only organizations with big budgets who are turning to DDA. It is the role of companies like Fospha to make such MTAs accessible to all businesses working with multichannel and omnichannel marketing.
“I believe any company who is spending cross-channel should be able to identify the true value of their efforts,” Bradfield continues, “which is why Fospha have broken down the barriers for all types of businesses to implement DDA.”
Cost is not the only barrier to implementing a comprehensive data-driven MTA model. Writing for ClickZ, CEO and co-founder of Measured, Trevor Testwuide cites the politics of adtech, the evolution of “walled gardens” such as Google and Facebook, as well as messy user-level data as significant challenges that need to be overcome in attribution.
Bradfield reflects on this: “Walled gardens from the adtech platforms certainly makes complete accuracy tricky,” he says.
“It’s important to know that there is no perfect MTA model or completely accurate tracking tools on the market. What we find as the main issue is implementing data driven attribution through an adtech platform such as Google due to the inaccuracy and bias data.”
Even having all the data for every customer across every channel touchpoint cannot yet guarantee complete MTA accuracy, but Bradfield does point to one way in which to overcome these challenges.
“It is important to source an independent tool,” he says. “Then there is no agenda for them to be selling ads and you can eliminate some of that bias.”
The linear, time decay and u-shaped MTA models do help ensure that all touchpoints get at least some credit for a conversion. But as we have seen, they can lack in accuracy.
In the case of the linear model, all channels receive even credit. But this option doesn’t take into account the likelihood that any number of touchpoints along that journey could have had more impact on the consumer than others.
Time decay can be desirable for time-sensitive campaigns. But, again, it assumes that certain channels – simply by virtue of seeing an interaction earlier – might be making less of an impact than they are.
And the u-shaped model, while adhering to the logic that the first and last touch of a user journey are often significant, it can seriously under-credit the middle channels – especially in the context of longer purchase funnels.
The MTA option that eliminates this guesswork is the DDA model (sometimes known as algorithmic MTA). While customer data can never be 100% accurate and – as Bradfield points out – even the best MTA model isn’t perfect, the inclusion of data into the attribution mix credits channels by how they are seen to be performing, rather than simply at which stage in the customer journey they appear.
Independent tools eliminate some of the bias and skewed data that marketers have reported within the walled gardens of Google and Facebook. While a solid Customer Data Platform works to consolidate, organize, and tweak marketing strategy in real time.
It is no surprise, then that data-driven MTA models are proving invaluable to big budget businesses. These are often the organisations with many paid channels in their marketing arsenal, but such models are becoming increasingly accessible to medium and small size organizations too – thanks to the ability to offset the cost with better ROI.
In today’s omnichannel world, cost shouldn’t be a barrier to a business of any size considering a data-driven MTA model. But for any organisation that still feels it is not ready to commit to this level of attribution, linear, time decay, and u-shaped MTAs are always a better option than the first and last click single-touch attribution models many are seen to use as a default.
Content produced in collaboration with Fospha.