5 Considerations While Building & Executing a Multitouch Attribution Framework
The modern mobile marketer deals with an extremely complex landscape, surrounded by a plethora of data and limited time to deliver impact. In the ever-challenging world of scaling user acquisition and growing marketing budgets with ROAS-led vision, the validity of media channels outside of the duopoly and their persistence in future marketing mix strategy is often called to question in favor of good unit economics.
Smart marketers must balance their portfolio with impactful channels, quickly killing off experimental channels that don’t deliver value to their overall digital efforts. Given individual channel platform limitations for audience exclusion based on factors beyond defined final conversions, and the lack of sophisticated audience exclusion technologies afforded to small-medium startups, there is a large overlap where the same prospective customers are reached multiple times through various channels.
Multitouch attribution (MTA) has many forms, with a clear top 3 use cases across industries. Firstly, applying itself to reconciling online efforts with their offline impact or, inversely, offline advertising translated into an online impact. Secondly, to clear attribution across multiple screens such as laptops, desktops, mobile phones, tablets etc. The third, and the one form of MTA that I will elaborate on, would be attribution across a portfolio of different digital channels.
The major driver to set up a MTA framework and evaluate channel contribution in a portfolio is to validate their assistance towards an end conversion. Judging a channel’s value purely based on an end conversion can show it in poor light, especially if it has a healthy interaction with the prospective customer at earlier stages of the funnel. As in a ball game, a channel’s indirect assistance in driving conversions, and thereby its true impact, should be given importance in budget allocation while scaling UA efforts.
It’s always sound practice to evaluate how important MTA is to your overall strategy and if the time, effort and trade-off to develop your framework is worth forsaking another project instead. While building and executing on the actual framework, there are a few considerations that will help ensure your best efforts. Depending on the business and the customer funnel in question, a variety of approaches exist when it comes to deciding how to assign weightages to each channel interaction in the customer journey. It is up to you as the decision maker to decide which method produces the most cash flow while aligning with the business. Let me quickly list the most popular approaches below:
- First click attribution – All credit for the conversion goes to the first channel interaction in your integrated campaigns.
- Last click attribution – Inversely, all credit goes to the last channel interaction amongst all touch points of the campaign.
- Linear attribution – Each touch point or interaction gets the same weighted contribution in the whole customer journey.
- Time Decay attribution – The channel interaction closest to conversion gets most of the credit, and the touch point prior to that will get less credit. Thereby, over time the weighted relevance of closer interactions to the actual conversion get bigger pieces of the pie. If the earlier interactions were as powerful, they would’ve resulted in the final conversion independently.
- Position based (U shaped) attribution – Here, 40% of the weight is assigned to each the first and last touch points, with the remaining 20% distributed evenly to all the interactions between them.
- Customized attribution – Introduce factors and learnings unique to your business by building on the earlier models. This could extend to look back windows for each funnel stage, the exclusion of certain channels such as newsletters if deemed sensible at certain stages or playing with weightages etc.
In my opinion, starting with time decay and moving on to a customized attribution model has been powerful to all MTA efforts. Whichever is deemed the most relevant approach upon which the framework is constructed, here are five considerations to bear in mind while executing on the MTA framework:
- Clear Objective: Identify and define the objective around which the MTA framework is built. For example, if you want to better understand the incrementality each channel brings to your bottom line, a simple experiment with a test and control group wherein a combination of touch points (retargeting) can be evaluated against the control group that is not exposed to any retargeting. This could be in the form of push notifications or other channel initiatives. The incremental revenue would be what is generated by the test group, on top of the revenue base line (from the control group) because of the channel mix.
- Quick customization: As mentioned before, each business can have its own considerations to account for while arriving at the best MTA model. In my own experience, we found that having a trial period for a paid subscription and the general user behavior of converting into a subscriber at a much later point after the trial has ended, needed to be recognized while developing a customized attribution model. One must also respect the channel platform’s offerings apart from what makes most sense for your funnel. While 7,14 and 30 day look back is common, you can also set a customized window or choose the closest fit depending on the media channel and the options it provides. Make sure you have a uniform format across all your media channels to compare apples to apples.
- Externalities: A major learning we had is executing on the MTA model when the industry seasonality was unbeknownst to us. If you have implemented a new model at a time when you know the market is slow and you cannot achieve statistical significance for solid conclusions, the value of the model is going to be underrated and misjudged. Always look into the timing for the introduction of your model, as well as planned fluctuations in your budgets as they will have an impact on the success of your experiment.
- Channel inclusions: Make sure MTA is necessary for your business. The spend, effort required and expected outcome worth your budget and scaling efforts and your business should have multiple channel campaigns with sufficient budget and micro conversion events for MTA to make sense. Start assigning value to push notifications, email newsletters, SMS and mobile web by tracking these efforts as part of your marketing channel mix. If owned media is not tracked alongside paid media, revenue can be incorrectly double counted if an event is attributed to the both. Also look at the combination of channels that drives the most value and make it a repeat procedure while creating audiences for further funnel treatment.
- Statistical Significance: Any assumption needs to go through a test of fire with statistical significance. Use the multi-armed bandit approach for your conclusions and make sure to respect statistical significance (large enough sample set, enough conversions in each variation) before deciding on the final solution. Don’t arrive at a definitive framework too early without playing around with models and weightages to see what drives most value through micro conversion events such as a sign up.
Ultimately, deciding on MTA depends upon your business, your team and your resources. Figuring out most of your marketing spend is more gratifying than solving every piece of the puzzle. John Wanamaker once said, “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” Let MTA be the guiding light in making sure your business stays ROI positive in these challenging times.
