Accurate attribution is complicated by the sheer volume of data associated with the user journey or "paths to conversion." That’s why marketers have resorted to last-ad attribution for so long, despite its inherent flaws. It’s simply the most convenient and readily available model.
But last-event attribution modelling has created serious problems - from misalignment of incentives - where marketing teams compete and are incentivized based on channel-specific goals - to gaming last-event attribution in RTB.
So what are the alternatives to last-touch?
Heuristic multi-touch models have risen to prominence over the last few years, and initially these models seemed preferable to crediting everything to a single event. Multi-touch, rules-based models found favor with analysts who were able to weigh different events in the user journey to evaluate how attribution would look if spun from an alternative perspective.
However, these insights are inferential and can result in assignment of credit that is just as arbitrary as last-event analysis.
Some analysts are beginning to apply more sophisticated techniques to processing and analysing attribution data. Instead of imposing a fixed model on path-to-conversion data, analytics teams can use algorithms or statistical models with log-level data to understand relationships and attribute credit. The problem here is the amount of resources, skill and expertise required to understand the results and how they were derived. Agencies may struggle to justify significant time and financial investment in these more robust attribution methods without first knowing what the benefits will be. No investment is approved without evidence, yet no evidence is possible without first investing a significant amount of time and resources.
This dilemma is further compounded by the "black box" solutions offered by some vendors, choosing between a heuristic rules based or algorithmic approach, and the fact that the market may not be comfortable with the application of an advanced statistical methodology to attribution.
Key considerations when tackling attribution
Here are some of the key requirements and considerations marketers should take into account when examining attribution models:
Time to insights: data processing for advanced attribution analysis can be time and resource intensive. A software-based approach to attribution means marketers can access their data more easily and efficiently.
Presentation of data: marketers need to be able to access and consume insights through an interface that facilitates understanding from both holistic and granular levels.
Actionable Output: many attribution solutions are descriptive in terms of output. Essentially they describe the data but provide little direction in terms of the actions that should be taken. Marketers need to know how media spend could be optimized as a direct consequence of the analysis.
Media Independence: A media neutral source prevents any potential conflict of interest in reporting performance results.
Accurate methodology: marketers need an accurate alternative to rules based or heuristic modelling. Ideally the model must be mathematically proven to work but transparent enough to explain to a non-technical audience.