Our latest solution: Rockerbox Experiments Learn More
All the questions you have on measurement, attribution, incrementality and marketing answered.
Hold-out testing is the process of intentionally omitting marketing from a section (i.e. control group) of your audience for the purpose of comparing the results across the test and control group.
Marketing attribution is the process for determining which marketing touchpoints led to a conversion (e.g. a signup, download or purchase).
An attribution model is the rule, or set of rules, that determines how conversion credit is assigned to different marketing touchpoints. Attribution models can be single-touch or multi-touch.
Single-touch attribution models assign 100% of conversion credit to one marketing touchpoint. Even if a customer saw twenty ads before converting, single-touch attribution will determine that only one of the twenty ads deserves conversion credit. Single-touch attribution models are easy to implement because of their low level of complexity. They’re also the most popular attribution models because of their historical tie to Google Ads (formerly AdWords).
Last-touch attribution assigns 100% of the credit to the last marketing touchpoint. This model would give all the credit to the striker (Player E in the image below). You could argue that the last touch is all that matters because it resulted in the actual conversion, but it doesn’t tell the whole story. Your customers are likely engaging with your brand across multiple touchpoints on various channels before they convert.
While single-touch only gives credit to one marketing touchpoint, multi-touch assumes that all touchpoints play some role in driving a conversion.
Often referred to as linear, even-weight distributes all credit evenly. In a soccer game, all players would receive the same amount of credit for scoring a goal. Even-weight is more sophisticated than a single-touch model—it doesn’t ignore all the touchpoints in the middle of the conversion path. The downside is that not all touchpoints are created equal. This model essentially says that every soccer player gets awarded for participating, which is unfair to the rockstars.
Time decay assumes that the closer a touchpoint is to a conversion, the more influential it is. With a time decay model, the soccer players closest to the goalpost receive the most credit. While time decay acknowledges that not all touchpoints are created equal, it’s still flawed. Let’s say someone reads a powerful article on your website that makes them want your product. The next day, this person is retargeted with an ad. Is the ad really doing more heavy lifting in earning the conversion?
Also known as the U-shape model, the position-based model gives 40% of conversion credit to the first and last marketing touchpoints. The remaining 20% is distributed evenly among all touchpoints in between. In the soccer game, Players A and E receive the most credit while the remaining players are acknowledged to a lesser degree, equally. If the touchpoint that grabs customers’ initial interest and the touchpoint that ultimately converts them are deemed most important, this model works well. The drawback is that if your business really needs a second or third touchpoint to convert a customer, this model won’t properly acknowledge it.
Machine learning MTA models (commonly referred to as data-driven MTA models) use your historical data to create rules for assigning credit to each of your marketing touchpoints. In the soccer game, each player’s value would be determined based on their performance throughout multiple plays. Implementing a machine learning MTA model used to be challenging. For instance, it requires having access to the historical data of your marketing activities, as well as a tool to analyze performance patterns. Challenges aside, this model brings you a lot closer to determining the true value of each marketing touchpoint.