Multi-Touch Attribution (MTA) collects individual, or user-level data, for trackable addressable media and conversion events in order to determine the impact of each media event to the desired conversion at the customer level. By summing the impact of each addressable media touchpoint on each customers’ likelihood to convert, MTA quantifies the total media channel lift provided by addressable media. MTA does not account for the impact of non-addressable media, and furthermore much addressable media is either non-trackable or lost due to the innumerable challenges of tracking data at the user level.

 

Multi touch attribution is no longer viable due to the demise of third-party tracking cookies, new privacy laws, low identity resolution, pixel related data loss and many other factors.

Follow Us:

How is a multi touch attribution model implemented?
In order to implement a multi-touch attribution (MTA) model you must be able to track all of your media and connect them to an individual. It is a complex and difficult process. There are essentially two ways to go about implementing MTA: DIY or use an MTA provider (more on MTA, or attribution solution providers below). The DIY method, or internally building an MTA model can be done with the right tracking tools and the right data science and engineering resources. It is a time consuming and labor intensive process but can be set up to deliver results far superior than first or last click reporting. Tracking can be performed using Google Analytics, a tool like Segment or one of the many Open Source tracking pixels. Essentially you’ll be building user click paths. Keep in mind that in almost all cases, you will not be able to capture impression level data and pipe it into your models, as many publishers and walled-gardens do not share it. Impression views are a major portion of the overall picture and the lack of this visibility is a big detractor to using MTA. It’s important to note that incrementality measurement takes into account all click and impression data and is not subject to the data reconciliation issues of MTA. Once tracking is set up you’ll need to consider which type of model you’ll use. Some common multi-touch attribution Models are: In most cases you’ll need a data scientist or engineer to implement the model, algorithmic being the most complicated and time consuming. How do you make an attribution model?
Some common multi-touch attribution models -

• Rules Based Weighted Distribution ex) 60% first touch, 30% last touch, 10% other touchpoints - This is a common model and puts the majority of the weight on the first and last touches. The problem with this mode is you still must decide what you want the weights to be for each touch along the path to conversion. It requires much diligence, review and revisions often to keep it close to a version of the truth.
• Rules Based Even Distribution - Credit is divided up equally across all touchpoints in the path to a conversion. It is not a common model and is less accurate than weighed or Algorithmic.
• Algorithmic - This model uses machine learning to objectively determine the impact of marketing stimuli along a consumer’s path to conversion. Building this type of model is extremely time consuming and labor intensive. It is also fraught with data breakage/leakage and a total lack of impression visibility in many major marketing channels.

In most cases you’ll need a data scientist or senior data engineer to implement the model, algorithmic being the most complicated and time consuming. What is an attribution platform, or attribution solution provider?
An attribution platform, or MTA platform, is a marketing technology software solution that captures individual user-level events (interactions with media) across marketing channels and leverages machine learning to apply an algorithmic (or heuristic) model which then assigns appropriate credit to the media touchpoints in the consumer’s journey to a conversion event. There are tools in the market today that provide more robust tracking and segmenting functionality than Google Analytics and can help advanced marketing teams build their own analytics practice and attribution models internally. The ‘Pros’ of doing this internally are that there is more control over the tracking, modeling and reporting process. If a marketing team has an advanced data science and analytics practice, it could be feasible to go this route. There are also “full-service” providers that instrument the tracking of the user-level events across media publishers and platforms, apply their own proprietary attribution models and deliver a bespoke reporting tool. They produce reporting and insights that give a clearer view of marketing’s performance and is more directionally informative than last click reporting provided by publishers. In both scenarios, there are quite a few “cons.” In today’s privacy rich environment, MTA is an extremely difficult exercise to land and produce the desired results due to the demise of the third-party tracking cookie, low identity resolution and many other factors. There are also huge gaps in user match rates, data reconciliation issues and no walled garden visibility. The setup can often take months and is very costly in time and dollars invested. MTA is not for the faint of heart. Alternatively, Incrementality Measurement can answer the MTA, or cross-channel attribution, problem statement; however it goes about it in a much different way. Because incrementality experiments are designed at the group level, DoEs are not subject to all of the user level data challenges encountered by MTA requiring only that campaigns exhibit enough reach to establish statistical significance at the group level. The experiments are deployed within the publisher platforms themselves, so marketers can gain a true understanding of the incremental contribution of each marketing channel down to the most granular level achievable. Additionally, setup is usually completed in several weeks and the ongoing costs are typically considerably less than MTA. You can read more about incrementality testing here: https://measured.com/faq/what-is-incrementality-testing/ What is the difference between a wholesome attribution model and a fractional attribution model?
A wholesome attribution model assigns credit to the first touch, or the last touch. A fractional attribution model spreads credit across all marketing touchpoints in the consumer journey that led to a conversion event. Can MTA be used for forecasting?
Unlike MMM, best in class MTA models estimate propensity to convert rather than demand and are therefore not directly applicable to forecasting. While demand curves can be inferred from MTA models, they typically do not have much validity at the campaign level and only inform tactical decision making at the sub-channel level without MMM’s ability to forecast or support strategic decision making.