Being capable o understanding every interaction with a customer and the value of that impact on the customer journey is what every marketing analytics and data science teams should aim to achieve
Multitouch attribution is one of those topics in Marketing that is largely discussed, often cited as “a gamechanger”, subject of research papers on the benefits of applying Markov Chains to your machine learning process.
However, only a third of marketeers report using any sort of Multitouch attribution initiative as part of their Analytics structure.
On the other hand, Google launched its Attribution capabilities as part of Google analytics in 2013 and the free version in 2017. So most companies could, in theory, be using some sort of attribution in their analytics process.
In Portugal and Spain (Math Marketing main markets in Europe) most companies I´ve talked to go from not doing anything to asking us to design a project so they can start on the path towards Attribution
So why is that the case?
A research from the MultiTouch Attribution Think Tank sheds some light into the difficulties in adopting MTA within companies.
Organizing the data
The main roadblock companies cite in working with MTA has to do with the data.
Both for companies working on or planning to work on MTA projects, issues around accessing, organizing and connecting data are the main pain points.
No surprise here, data is the cornerstone of a MTA project. Without it there is little one can do. Implementing a MTA operations is, first and foremost, about organizing and linking data from several different sources in order to identify single users across multiple platforms.
But how can one go about in order to create a Multitouch attribution project? What are the basic milestones and steps a company needs to take in order to achieve it?
1 - Understand what you are trying to achieve
Multitouch Attribution operates on the realm of advanced analytics. In our analytics continuum framework, it falls within the insights category. That means that in order to achieve it, machine learning and other data science skills have to be available. MTA is not a tool one can acquire, but a capability one has to develop. Having the proper skills available is the key success factor of the project.
2 - Make sure the Basic works
You might have heard of the maximum "Garbage in, Garbage out". That means that a statistical model is as good as the date one inputs in it.
That is also important for the basics of any Digital marketing project. Do you have a proper standard for creating and managing Campaign UTMs? Are they systematically created, with a consistent nomenclature, with organized rules that are available to consultation? IF the answer in no, than this is where you must start
The same is true for you media investments. Is the information available, organized properly structured?
Also, make sure your web analytics solution is properly implemented. Users ID, events and Custom categories should be current knowledge for you and your team before starting your MTA journey.
3 - Use your Google Model
If you advertise with Google Display and Search, and if your search console is properly configured, it is possible to start using Google Analytics build in attribution models. It works very well within the Google ecosystem, with important limitation when trying to connect with other sources, such as Facebook or Twitter. However, it is a good first step.
4 - Finding a Common denominator with your data
This is by far the most complicated part of the process. A study from Salesforce show that clients use 10 channels, on average, to communicate with brands. In order to build a MTA model, one has to find a common denominator on all those different sources, in order to find a single user, the same user, on all those platforms.
For customers that already visited your website once, that should be less of an issue. Either by developing a proprietary pixel or using a combination of google User id and customer ID from CRM one should be able to monitor all user interactions and visits on the website, even for a unknown customer.
That becomes more complicated when we talk about impressions. Given that some (or most) customers will not come to your website all the time, being able to find a single denominator between multiple audiences is key. Here, on will do a mix of information mash up with statistical inferences to connect all the different audiences in a single user. It will not be perfect, and there will be some blind spots. But its a process, and in every interaction your model will become better.
5 - Build your Attribution models.
Once your data is connected and assigned to a user, it is time to work on your attribution model itself. In order to accelerate the return on your capability investments, starting with a basic attribution model (U, Time decay or Linear) is a good place to start. However, we always advocate towards a customized attribution model, in which the value of each impact is evaluated based on your business reality. That requires training the model, adjusting and refining it constantly. Analytics in general, and Multitouch in particular, cannot be treated as a project with a beginning and end. It is mostly an operation, that requires constant improvement and that evolves.
6 - Removal effect and Plateau
Once a customized MTA model is build, two new KPIs will become norm in your optimization daily meeting - Removal effect and Plateau. The removal effect shows how many conversion will be lost by removing a certain media from the marketing mix. The plateau, on the other hand, will predict the maximum expected conversions from a certain media given the current conditions. Those two KPIs will allow for a better media planning and optimization, improving the ROAS of your marketing teams.
Show me the money.
Building a Multitouch Attribution Model is an investment of money, time and team. So one expects that investments to turn into more sales, smaller costs of acquisition or, preferably, both.
Bottom line - Multitouch attribution is the solution to that old saying attributed to John Wanamaker: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” Being capable o understanding every interaction with a customer and, most important, the value of that impact on the customer journey is what every marketing analytics and data science teams should aim to achieve.