Attribution - who should be credited, and how? ...let the data do the talking.
What is attribution?
Attribution is considered by most as the identification of a set of user actions ("events" or "touchpoints") that contribute in some manner to a desired outcome, and then the assignment of a value to each of these events.
Marketing attribution provides a level of understanding of what combination of events, in what particular order, influence individuals to engage in a desired behaviour, typically referred to as a conversion.
The key here is to have a clear outline of what it is you are trying to achieve, and what is your objective? Once we have outlined that, we have a clear milestone to work from and can address an appropriate attribution method.
What are the benefits of attribution?
As mentioned, one of the key reasons for attribution is to define where you get the biggest bang for your buck. However, attribution goes deeper than that - in order to understand this we need to breakdown what is driving the success of the objective and what isn’t, therefore breaking down and understanding the customer journey and experience to inform us of the following areas:
Currently, customers gain information via a number of sources prior to reaching a business object, e.g., a purchase. This does add complexity to attribution as there are more data points being generated in several locations. However, this is also advantageous as it will refine nuances in customer journeys, potentially defining segments in your audience that you can personalise to a greater extent, increasing conversion potential.
Understanding the channel mix to an objective such as a purchase provides us with the information and a starting point as to what tactics to deploy, e.g. from the below, we see that a combination of Email, Social and Paid Search provides us with a 5% increase in purchase. From there, we must calculate how much was spent for this increase in order to justify the return.
So, does this mean that each of these channels should be credited an equal weighting in the success of the purchase? This is where the contention occurs, you might have sent 10 emails but only 1 social post, is that fair to credit to all? This really relates to what you are trying to achieve as to the logic which you apply to attribution which I’ll cover later. What this does determine is that channels generally don’t have one purpose e.g. media driven awareness and email driving sells in isolation, but they have a supporting interplay between their roles at optimising objective success.
The customer journey isn’t a linear path which everyone follows, we need to understand the customers’ attributes and traits to comprehend commonalities in journey touchpoints undertaken. Using a rules-based logic in which to determine attribution brings in bias as to what you are giving credit to. Most clients are still using ‘last touch’ attribution, mainly because this is the easiest to apply as they don’t have the data in place to determine a fuller customer journey and they’re focused on conversions. However, though the endpoint of the objective may be the same, there are lots of routes that customers can undertake in getting there. As outlined below, understanding the entry points and audience types, people’s paths will change and therefore the importance of certain touchpoints might have a higher weighting to differing audiences. If you’re restricting attribution to last touch, there is no consideration into the previous interactions that customers undertook in getting there and therefore may miss a trick. This analysis also has the benefit of understanding where someone is in their journey and therefore can nudge them into the next important interaction or event, pushing them into an optimised path which they might have strayed from.
This proves very handy when you are trying to reach your objective, as you can build out an appropriate measurement framework outlining the drivers that are scaling up into the completion of the objective, rather than making an educated guess - we now have the facts. From there, the ‘levers’ which marketing plays can influence each stage of the measure framework dependent on what tactics are required to be implemented against the benchmarks originally set.
Time and frequency to convert
Other parameters which need to be considered in attribution are ascertaining a time period of maturity for the objective set. Every client’s purchase cycle is different depending on the sector or the expense of the product/service, as this will have influence on the customers decision making e.g. I am not going to spend 6 months researching a pencil worth 20p, but I would if I were buying a house. Therefore, we need to analyse from the point of an objective completion and work backwards to understand the events which were taken prior completion. This will give a host of interactions we can analyse, how many people interacted with those events providing the relevancy of it, plus the standard deviation as to the likelihood of when people conducted that interaction prior completion. Based on these factors, we can determine the appropriate window and events which are driving completion.
Now that we have outlined the component parts for consideration of attribution, there are tools available, such as Google, which use rule-based logic to help us. This is where I feel the contention lays, as depending on your viewpoint you’re going to use the model which fits your directive e.g. media will use first click as they’re prominently higher up the customer funnel vs CRM who would favour last click. Therefore, in summary I have outlined the uses of each:
Let the data do the talking
As you may have guessed, rule-based logic isn’t my preferred choice in how to attribute appropriately as it can be subjective depending on the viewpoint. If the data is there, then it will tell its own story as to what is relevant and what isn’t in an agnostic manner.
With the advancements of machine learning in this field, we are getting closer to determining which data points are relevant at speed. Just like the human eye, information is captured from the field of vision and then processed by the most powerful computer in the world, the human brain. The brain then determines which visual cues (or data points in marketing) are relevant providing weighting and which to ignore. The brain then formulates based on these prominent cues what to do next e.g. if you are driving down a country lane with vast fields of sheep, then at a junction another car pulls out, the brain will block the irrelevant cues such as sheep and focus on relevant cues such as the car pulling out, which then triggers the decision to brake…and then followed up with appropriate hand gestures! The same approach can be applied to the customer journey as depicted below:
In conclusion: The next step and evolution for attribution
As with the saying ‘If you don't know where you've come from, you don't know where you're going’, with the correct attribution using machine learning, we’ll know how objectives are accomplished and how they may adapt over time. We can create models/algorithm in which to accurately scenario plan if I spend XX on YY campaign, then potentially we will generate ZZ conversions through the funnel. This can then be applied to a centralised orchestration tool in which to construct and optimising customer journeys across all channels, increasing customers experience and potential to convert in the most efficient way.
However, there are things to bear in mind with attribution:
- Its only as good as the data you put into it, if events aren’t included in the attribution model then they can’t be weighted, creating bias to another channel or event, rather than interpretating the full customer journey
- Identity too may blur the lines creating duplicate records with pertaining events fragmenting the customer journey and skewing the event which actually generate successful objectives
- The model used, as noted before logic based has rules and weighting assigned, but a little further under the skin there might be other factors to consider e.g. Google Analytics attribution incorporates cross channel event but only on a click basis, whereas Google Campaign Manager only includes social and media channels but does incorporates impressions, however this is more from a campaign level basis
Depending on the objectives you’d like to review, and the sector will determine the approach to take e.g. if reviewing sales which are conducted online and offline then customer data platforms would be recommended to capture and consolidate identities and events. Whereas if its an e-commerce platform then the majority of events and sale would be online, therefore Google might be your best bet!