Attribution is a tough challenge that is top of mind for every Marketing and Analytics leader. While marketing strategies and technologies may have evolved, the most important question has not changed - Is our marketing working? The right attribution solution should help you answer that question. But how do you find the right solution?
Unfortunately, there is no turnkey attribution solution that perfectly solves all of your measurement challenges. Each business has unique attribution challenges and there are a seemingly infinite number of vendors and methodologies. As a result, I created a framework to navigate the increasingly complex multi-touch attribution market, understand the trade offs between solutions and identify the optimal attribution solution.
What is the point of attribution and why is it hard?
At its heart, attribution (specifically multi-touch attribution) is a very simple problem - how do I divide credit for a conversion (e.g., pdp view, order completed, customer acquired, app downloaded) and give that credit to the factors that influenced the customer’s behavior. This is important, because as an effective Marketer, you need to know the ROI of your marketing strategies. In order to calculate ROI you need to have both the cost and the results (not necessarily $s) at the strategy or tactic level (e.g., channel, campaign, creative).
The complexity arrives when you realize you do not have perfect information. This challenge comes in two flavors:
- Data we can’t collect - When someone converts (until we have chips in everyone’s brains) we do not know of all the ways they were exposed to our brand (Did a friend tell them over drinks? Did they see a billboard? Did they see an advertisement on an unknown device?). We also don’t know which of those events impacted their decision to convert.
- Data that is collected, but you can’t access - The social networks and search engines we use have developed rich cross-device graphs, which they use for the targeting and measurement of advertising. These platforms do not expose their cross-device graphs to external parties, instead the platforms present aggregates. In some cases, these platforms will also restrict the capture of impression data.
Introducing the framework
In this world of incomplete information, how do you find the most optimal attribution solution for your business model? I distilled the ideal attribution solution to five key characteristics. While all five may not be achievable in an attribution solution, this framework will allow you to understand the trade offs between solutions.
- Cross-device: Include all touchpoints across all devices
- Granular: Attribution, reporting and data access at the most granular level
- Neutral: Neutral, third party view of marketing performance
- Actionable: Reporting and analysis you can actually action on
- Comprehensive: Reporting and analysis that covers all channels in detail
If you don’t properly account for the multi-device customer journey you are only seeing half the picture. Today’s customer journey typically includes two (or more) devices. Without a robust cross-device solution a two device customer journey looks like two visitors - one non-converting and one converting. This means the tactics occurring on the ‘non-converting’ device look ineffective (and are given less attribution credit), when they may have had a tremendous impact on driving the conversion on the ‘converting’ device. The impact of this shortcoming is further compounded - if the tactic is given less credit than it ‘deserves’, the credit is given to another (over-credited) tactic.
Analysts love to dive deep into data. As a Management Consultant I would ask clients for the most granular data available - no dataset was too large. While my industry has changed, my attitude to granularity has not. The basis of this approach comes from a ‘you never know what you’ll need, so get it all’ attitude. However, there are genuine insights and unique company specific metrics that you will never be able to calculate using aggregates alone. Generalized solutions can get you pretty far, but will not be tailored for your specific use case. For this reason, an ideal attribution solution enables access to the underlying raw data.
Marketers and Analysts need an objective view of their advertising’s performance. The saying ‘you shouldn’t grade your own homework’ is incredibly true. Insights and operational reporting provided by the advertising platforms themselves tend to cast the platforms in a very positive light. Biased presentation of data can take the form of fluff pieces (e.g., conference presentations, webinars and case studies), subtle tweaks in how a KPI is calculated / presented (e.g., broken Y-axis) or even outright mistakes in measurement.
Attribution, due to its complexity and black-box nature, leaves plenty of room for artistic license. Therefore, it is incredibly important to have an objective attribution solution that has neither intended nor unintended bias.
In the military, when firing a piece of artillery there are two main roles: the person firing the round and the spotter. Both roles are incredibly important. The person firing the round must operate the equipment to hit the target in often less than optimal environmental conditions (sounds like a Marketer). The spotter must observe and measure the trajectory of the round and provide guidance to further improve the effectiveness / accuracy (sounds like an Analyst). Communication is key - without communication, it turns into a game of guess and check.
In our world, a Marketer’s weapon of choice is the advertising platform, and the Analyst prefers an attribution solution over binoculars. Communication is still key. However, communication breaks down when the advertising platform and the attribution solution have structural differences. For example, the attribution solutions determines the tactic works for audience segment A, but not B; however, the advertising platform does not allow for audience segment targeting (or does not have the same segments). Or perhaps the attribution solution provides results at tactic level, but the advertising platform can only target at sub-tactic level. In order for an attribution solution to be actionable, both the attribution solution and marketing platform must speak the same language.
If a bucket has one hole is it any better than a bucket with ten holes? Some would argue the one hole bucket is more reliable. I say, if you are trying to transport water, both buckets are just as bad.
Attribution solutions are pretty similar - an ideal attribution solution should not have any holes (i.e. missing data). Multi-touch attribution solutions use behavioral data to credit the tactics and strategies that influenced a conversion. Missing (or inconsistent) data biases the model and in turn produces biased results. Missing data is typically caused by tactic specific limitations (i.e. marketing platform X does not allow you collect data point Y). Tactic specific inconsistencies mean that the attribution results for this entire tactic is biased. The bias, however, is not limited to the tactic with the missing data - if the tactic is given more / less credit than it ‘deserves’, the credit is being stolen from / given to another tactic.
Data inconsistencies may come from the unique characteristics of a tactic (e.g., marketing platform X is incapable of collecting data point Y due to technology or design limitation). However, the most common (and increasingly so) cause is the limitation due to walled gardens.
Attribution is confusing and unfortunately there is no perfect, turn key solution. At the end of the day, you can view these five factors as an unsolvable dilemma or an empowering framework to find the optimal solution. As you can imagine, I prefer the latter. Good luck!
This is just version one of the framework - if you have thoughts or feedback, please share it here or send it my way!