Fair credit attribution across the entire customer journey
The Treasure Data Multi-Touch Attribution model uses cutting edge machine learning to give you accurate, actionable insights into which marketing channels have the most impact through the entire customer journey.
Other rule-based Multi-Touch Attribution solutions leverage models that only capture a portion of the ad impressions distribution curve like first-touch, last-touch or linear models. These models also tend to overemphasize 1-3 channels and miss the importance that other touchpoints play during a customer’s decision process.
Our solution uses the machine learning driven shapley values model, which extracts meaningful patterns from omni-channel customer journeys so you can see which channels have actual impact during different stages of the customer journey. Thus enabling you to easily see what spend is most effective during which phase of the customer journey, where high quality leads are coming from and just as importantly where they are not.
Image is taken from Causally Driven Incremental Multi Touch Attribution Using a Recurrent Neural Network and modified by Treasure Data.
This may sound complicated, but at the end of the tutorial below you will be able to generate a simple dashboard like the one below.
If you are interested in trying this out on your Treasure Data account you can either dive into the technical implementation below, or contact your Customer Sales Representative for consultation on implementing it in your Treasure Data account.
- Analyze the customer's path to conversion and understand the impact of every touch point.
- Optimize your marketing budget allocation by understanding key touchpoints and prioritizing individual channels.
- Maximize KPI (ROAS/ ROMS) based on the optimal resource allocation.
We have a video that describes how this Box technically works.
To fully leverage the Multi-Touch Attribution model we will assume you have cleaned up your data set. If you are unsure where to start on cleaning your data set please see the further reading section for guidance on how to use server side cookies and the ID unification process.
The workflow assumes we have a following touchpoints table that collects user behaviors and conversion events with their sources (i.e., marketing channels):
If you have a
pageviews table collected by td-js-sdk, the table can be easily transformed to the required format with some lines of queries.
Next you will upload the workflow linked in the github repo below to your Treasure Data account and execute it. There are some optional configuration parameters that you can tweak in the config file as well.
For the latest code and instructions please follow along on the github repo linked below.
The model is fully customizable for your data depending on your own definition of conversion. Contact your Customer Success Representative if you are interested in building and testing the advanced MTA solution.