Customer purchasing journeys aren’t what they used to be. In today’s hyper-connected world, the progression from awareness to consideration to purchase is more loop-de-loop roller coaster than straightforward funnel. The path can involve multiple devices, dozens of website visits, interactions at brick-and-mortar stores, and multiple encounters with marketing campaigns across the web, on social media, and in email. That’s why it should came as no surprise that consumers today own an average of 3.2 internet-connected devices, according to GlobalWebIndex.
In such an environment, the question of attribution becomes a far more complex one for marketing teams. Which interactions are actually impacting conversion, for which consumers, buying which products, at what time? What’s the relative contribution of each interaction to conversion or another marketing goal?
According to Nielsen, 79% of CMOs expect to increase their investment in marketing analytics and attribution in the next 12 months. And while many marketers have relied on straightforward attribution models like first or last-touch, teams are increasingly recognizing the need for more sophisticated multi-channel attribution.
Successful multi-channel attribution quantifies the relative impact of each customer interaction (both online and offline) on revenue, giving marketers a picture of the full customer journey to more effectively allocate spend across channels.
There are two main types of multi-channel attribution models:
- Rules-based (e.g., first touch, last touch, linear, u-shaped, and time decay)
- Algorithmic
Rules-Based Attribution
Some common rules-based attribution models include first- and last-touch (as mentioned above), as well as linear, u-shaped, and time-decay.
- First-touch: All credit goes to the first interaction.
- Last-touch: All credit goes to the last interaction.
- Linear: Credit is divided equally among all interactions.
- U-shaped (or position-based): First and last touches get the most credit, with the rest divided between the interactions in the middle.
- Time-decay: More credit is given to interactions that happened nearest the conversion.
While rules-based models are simple to implement and provide straightforward reporting, they often provide inaccurate results, which cause marketing teams to invest in the wrong channels. A big reason for this inaccuracy is they only include consumer interactions that led to conversions, despite the fact that non-converters are the vast majority of interactions and provide a huge and valuable data set with many lessons for marketers. “Even if a rules-based approach works right now, there’s no certainty it will continue to work just as well in the future, since rules-based approaches don’t take into account changes in your environment,” says Georgios Kourogiorgas, Data Scientist at Syntasa.
Algorithmic (or Data-Driven) Attribution
Unlike rules-based models, algorithmic attribution models are objective, relying on statistical models to determine how much weight each consumer interaction should receive. Algorithmic attribution, also known as data-driven attribution, is one of the most advanced ways to model attribution data. Successful models take advantage of massive amounts of data and use machine learning and predictive analytics to map each individual’s journey through your online and offline assets. Ultimately, teams can use the results to answer business questions, such as:
- Do all channels work equally well for all products?
- Which channels are more effective with loyal customers/high spenders?
- Are there segments in which normally effective channels are underperforming?
Some worry that algorithmic attribution solutions are black boxes, making it hard for teams to understand why machine learning models have come up with a certain result (and hard to persuade decision-makers of the algorithms’ effectiveness). Employing a transparent and highly customizable solution that works within your big data environment and taps all customer data — including behavioral data from websites, apps, and devices, along with offline activities and enterprise data — will provide explainability. The most effective approach combines algorithmic attribution with rules-based methods, in order to compare and contrast results and maximize your team’s understanding of the precise impact of every touchpoint.
User-level journey data is one of the key prerequisites to benefiting from a multi-channel algorithmic attribution model. Because data at an aggregate level is not accurate for attribution, it’s necessary to have a dataset which captures the success or failure of each visitor in reaching the goal event (e.g., buying a product or renewing a subscription) at any given time. (For companies without this user-level data, Syntasa’s algorithmic attribution app can help construct buyer journeys on an individual level to use in algorithmic attribution models.)
Eager to better understand which attribution model is right for your team? Read Syntasa’s detailed guide on algorithmic attribution and then request a briefing to see what an algorithmic attribution app looks like in action.
