The Algorithmic Attribution model in Analysis Workspace uses statistical techniques to dynamically determine the optimal allocation of credit for the selected metric.
Hi, this is Trevor Paulsen form Adobe Analytics Product Management. Attribution IQ is a popular feature within Analysis Workspace that allows you to apply any attribution model to a freeform table column to any metric or dimension. For example, in this marketing channels report, I can easily compare first touch revenue to last touch revenue, by changing the column settings.
This allows me to compare any two attribution models that I prefer, to one another, to evaluate how my marketing channels are performing against my conversion metrics. Before now, Attribution IQ has only supported rules-based models and while rules-based models are very useful, they can often leave important marketing decisions up to heuristics or trial and error, rather than a more statistical, data-based approach. So to help address this problem, we’ve added a new algorithmic model to Attribution IQ for our Analytics Ultimate customers. You can find it in the list of models with all the other models. By clicking on my model dropdown, you can see the new model here, Algorithmic. This new model uses techniques from cooperative game theory to statistically determine which channels deserve the credit for conversion. Specifically, the model applies a technique known as the Harsanyi dividend, which is a scalable generalization of Nobel Laureate Lloyd Shapely’s work on fairly distributing the winnings of a team of players in a game. Or in our case, attributing the conversion events in your data, to a set of marketing channels. When I apply the model, you’ll also notice that it supports every type of lookback window, including custom lookback windows. So I can specify a 30, 60, 90, or any other time period of lookback window that I want to compare. One of the most amazing aspects of the new model is that it can be applied to any metric and any dimension, just like every other Attribution IQ model. So you’ll notice if I bring in Online Orders, in addition to revenue, I can also use the Algorithmic model. In fact, you can use algorithmic models in calculated metrics, in breakdowns, on any dimension, or even in the Attribution IQ panel. By dragging in the Attribution panel, selecting Marketing Channel as my dimension or any other dimension, selecting my success metric, I can now include Algorithmic in the list of models that I want to compare. Additionally, I can add any degree of custom lookback to them as well. This gives you the power to compare the algorithmic model with any other rules-based model within Attribution IQ.
Algorithmic models finally give you the ability to take heuristics and trial and error out of the equation, allowing your organization to make better, more data-based decisions. Good luck. -
For more information, see the documentation