Clario Analytics partnered with a leading provider of permission-based email marketing solutions, strategy and services. In order to enhance their client services, the provider wanted the ability to provide an “engagement” score to their clients, in order to gauge the quality of lists, segments of lists, or individual recipients on lists. Clario Analytics, working with data from several clients, built a predictive model to classify email recipients in terms of likelihood to engage in email activities such as opens or clicks. By engaging, a recipient shows interest toward additional communications and interactions with a brand. While conversion is often the end goal of marketing operations, engagement is an important early indicator of conversion.

Model
Clario Analytics defines “email engagement” as the number of positive activities (opens, clicks, share to social, forward to friend) as a ratio of the number of emails sent. To build the model, we first created recipient attributes from a historical period, such as time since opt in, # of opens/# of sends, time since last open, etc.). We then built a model to predict level of engagement in the measurement period, given the attributes in the historical period. Regression techniques were utilized in developing and validating the model.

In addition to a continuous score, ranking recipients from highly engaged to not at all engaged, we defined five distinct segments from the model score. The graph here depicts the actual engagement ratio for five model segments, based on a sampling of ten clients.

Applications
There are several potential applications for the engagement score and segments:

1. Use the engagement score to group your customers into the five segments, and then devise a contact strategy based upon level of engagement. For example, set the email frequency for the customer groups as follows: send all available emails to the Enamored group, and reduce the treatment accordingly to the next four groups.
2. Establish your company's overall benchmark for the engagement score, then measure email campaigns against this benchmark to ascertain winner and loser campaigns.
3. Over time, monitor the engagement score distribution of your email customer list. You will get an indication of whether your list quality is improving or declining.
4. Use the engagement score as a potential predictor in any email campaign models, or simply suppress those with extremely low scores from your email campaigns and use your print channel for marketing.
5. If you are also mailing via the print channel, leverage the engagement score to test reducing print costs for the highly engaged customers. Also add the engagement score as a potential predictor to print campaign models.
6. Predict opt-outs. We know there is value in keeping opt-in email addresses for customers; said another way, there is an opportunity cost for opt-outs. An opt-out model could be an additional model; however, we believe that the engagement model may actually predict opt-outs (either actual opt-outs or “effective opt-outs”, customers who never touch an email).

Enhancements
With additional information beyond email activity (such as conversions, survey data, web browsing data and purchase history), you can customize and improve upon a generic engagement model. For example, pulling in web analytics data and purchase history should provide a model that better rank orders recipients according to engagement, compared to only email send/open/click data.

About Deb Campbell
Deb is a senior marketing analytics leader and is Clario’s Vice President of Modeling Solutions and a co-founder. She has spent the majority of her career developing and implementing analytical solutions, transforming data into intelligent marketing strategies. Before co-founding Clario, Deb worked at Fingerhut where she parlayed her advanced knowledge of database marketing, predictive modeling and optimization techniques and achieve significant revenue milestones from these efforts.