ChannelMix can help users measure the impact of different marketing channels on target events with ChannelMix's data-driven attribution model in our Multi-Touch Attribution dashboard.
Data-driven attribution is designed to be more accurate than rule-based attribution. Instead of relying on a one-size-fits-all rule, our data-driven model looks at all the specific interactions that specific users had with each of your marketing interventions to measure those interventions’ influence on target activity.
And unlike rule-based models, ChannelMix's data-driven attribution looks at both converting and non-converting paths to assess and assign value.
Data-Driven Attribution and Pathing Data
ChannelMix's data-driven model runs on your OneView Media dataset, which captures data from all your media channels, and website pathing data from Google Analytics.
A path is the string of interactions that someone has with your marketing on the “path” to filling out a website form, making a purchase or some other target activity. (Google Analytics’ Multi-Channel Funnels are a good example of pathing data.)
ChannelMix gathers pathing data through the use of:
- Google Analytics, which measures website activity and traffic sources
- ChannelMix ID, which can track website visitors across multiple sessions
- ChannelMix Conversion Tracking, to accurately measure conversions on your web property
One person’s path might look like the following:
- Monday: Clicked on a paid search ad and visited a product landing page on a company’s website
- Wednesday: Clicked on an organic social media post and visited a blog post on the company’s site
- Friday: Used a search engine to find the company’s site again, where they completed a lead form
Or put another way:
Paid Search > Organic Social > Organic Search > Target
A path can also end with no target event. Maybe the person decides not to become your customer.
At a high level, ChannelMix's data-driven model looks at pathing data for multiple users over a period of time to determine how often users engaged with a marketing channel before converting.
For example, if most of your users click one of your paid search ads at some point before converting, then paid search would be viewed as more influential and valuable than an interaction that is rarely involved in target events.
More specifically, ChannelMix's data-driven model employs a form of analysis called a Markov chain to determine the likelihood that a specific path will lead to a target event and, thus, calculate its target value.
A Markov chain uses pathing data to determine the likelihood of what a user’s next interaction will be, based strictly on the most recent interaction.
For example, a Markov chain might say a customer has a certain probability to interact with paid search immediately following paid social. But that judgment would be based only on the fact that the customer just encountered paid social and not on a slew of previous interactions.
This allows us to summarize all customer journeys, as we do in the image below. It shows all of the transitions from one channel to the next that are present in the pathing dataset and assigns a probability to each transition.