A Customer Just Unsubscribed! What Are They Telling You? Are You Even Listening?

Most of us marketers consider unsubcribes a fact of life in email. We all-too-often take them for granted. They are considered the unavoidable collateral damage that each passing campaign creates. But what if they were a missed opportunity to learn how our sending behavior may be impacting our effectiveness?

We set out to learn the answer.

I think most of us email marketers understand that unsubscribing is generally a healthy process. It allows us to purge the deadwood from our database. A little like a controlled-burn forest fire, cleaning out the underbrush to prevent large scale wildfires.

This may be true, at least in part, but I can’t help wondering how much of this is normal and how much is self-inflicted, cause by careless targeting, poor timing and ill-conceived campaigns. Not that we’re guilty of that. No, no, no…

So, for one of our clients, we took a deep dive into the master suppression list, the engagement data and the sending behavior of their email platform, database, CDP and Web tracking looking for patterns in the subscriber behavior. And looking at the results, it is an exercise that I strongly recommend to everyone. Here is the process we followed.Here

Our hypotheses:

What if there are distinct patterns in unsubscribe behavior that can be understood and addressed? We posited that a potentially significant portion of the unsubscribes shared similar characteristics. Among our hypotheses, we considered that perhaps:

• Certain types of emails were more likely to triggers an unsubscribe and others
• Unsubscribers might be more likely to share certain attributes or characteristics
• Unsubscribes may be more likely to occur at certain times of the year
• Certain markets may be more likely to unsubscribe than others
• Sending “too many” emails might trigger unsubscribes (sorry Dela)
• Poor customer service experience might lead to higher unsubscribes
• Product availability and pricing issues may lead to an unsubscribe

There were a few others, perhaps more far-fetched, but the data was unavailable to validate them. You may have other hypotheses that are specific to your sending practices or the nature of your business that you would add to the list.

Our Methodology

Once we understood what we wanted to examine through our hypotheses, we went about identifying which datapoints need and which sources contained the required data. Here are the sources we identified. Depending on your architecture and tech stack, these sources may vary. If you have a true 360 view of your customers (lucky you!), there may be a single source.

Using the Email platform’s API, we exported from the email platform the following data:

• The unsubscriber list from the last 2 years. This window was chosen because the engagement data lookback period was 2 years
• The engagement data from the past 2 years
• The sending data from the same period
• Tracked Website visit behavior (captured via the website tracking code)

We then went to complimentary data source to enrich the information using:

• The main contact database of active subscribed customers for comparison purposes
• CDP customer data including RFM and other scoring data
• Product purchase behavior.

In addition, we used specific queries in the database to identify groupings of customers that shared certain characteristics and know behaviors.

Drowning in an Ocean of Data!

After hours of work running SQL queries to merge sources and synthesize the data, we produced a .csv file containing 86 columns and tens of thousands of customer records (de-personalized of course), just begging to be analyzed.

But first we needed to prepare the dataset for analysis. Some of the transformations that were needed included:

• Transforming dates into days from, days since or days between.
• Aggregating email sending behaviors (count of emails sent 3 months, 12 months prior to unsubscribing)
• Averaging things like open rates, clickthrough rates, emails received, etc. prior to unsubscribing
• Compiling web visits, visit frequency, pages viewed prior to unsubscribed, including the last email received, the last web page visited, etc.
• Including NPS data for customer satisfaction prior to unsubscribing
• Cataloging seasonality values showing the month and season (summer, winter, etc.) of the unsubscribes.

Data Visualization Was Critical to Gaining Insights

You don’t necessarily need to do a lot of sophisticated modeling to gain valuable insights from this data. Using Looker Studio, we created an unsubscriber dashboard that allowed us to filter unsubscribers and compare them to subscribers using various graphics and charts.

The contrast between the two audiences immediately revealed certain features that were more present with unsubscribers. And some features that we did not expect to find in the unsubsribers.

Looker Studio has the wonderful ability to filter data based on any attribute available in the dataset just by clicking that attribute. If you know what you’re looking for, often Looker will provide the answers. Or at least allow you to do uncover them by cross-referencing various datapoints.

Are They Really Deadwood?

You would expect unsubscribers to be mostly inactive contacts who haven’t purchased in ages and who are cleaning out their inbox. But it turned out that certain campaigns systematically produced significant levels of unsubscribers who were active and loyal customers. Their database longevity was high, and they had recent purchases. Not good!

Is there seasonality?

Even if you don’t have Black Friday campaigns, you should expect any email activity to unleash a wave of unsubscribes. This is colorfully referred to “Unsubscribe Season”. If you are not a retailer and can avoid mailing frequently between Black Friday and early January, by all means do so. At the very least, limit the number of campaigns as much as possible and focus on a very specific and narrow audience for whom the offer is highly relevant.

Our client had very specific seasonality patterns and these were clearly reflected in the data. The seasonality influenced the relevance of their products for certain customers who were more likely to unsubscribe as the season changed.

Relevance is the key.

What we found came as no surprise. Poorly emails produced the highest unsubscribe rate. But how do you define “poorly targeted”? We always believed that the best offers of the season would be of interest to most customers and that it would be a good tool to reactivate lesser active customers.

The thought was that if inactives unsubscribed, good riddance since they were inactive anyway. But if they became reactivated thanks the “best offer”, they were brought back into the active universe and ended up producing incremental revenue that would otherwise not have been accrued. This was true, but only to a certain degree.

Photo by Tarik Haiga on Unsplash