Preventing Churn with Real-Time Sentiment Analysis: A Hydrant Case Study

Purpose of Analysis

Churn is a HUGE issue for any and all companies. But a common misconception is that identifying a customer at risk of churning is too difficult to do with accuracy.

The truth is, there are actually a few effective ways you can identify unhappy customers before it’s too late. A big one is tracking purchasing patterns, browsing behavior, and ordering frequency to see if there are any alarming changes over time (which Cotera can do). But a commonly overlooked method is sentiment analysis.

There are a few uses for sentiment analysis that have to do with churn.

For example, customers who write a negative review have probably made up their mind about churning. But if you have a tool that tracks and tags reviews in real time, you can identify negative reviews as soon as they’re posted and promptly resolve the issue, increasing your chances of changing that customer’s mind.

Or, you might have customers who ended up having a few issues with your product but decided to give you 4 or 5 stars anyway. If you can identify the negative sentiment within reviews that are overall positive, you can come up with personalized solutions and prevent top customers from churning.

To give you a quick example of this, we ran an analysis on Hydrant’s reviews for 3 of their products: Hydrate, Energy, Immune.

The Analysis

So we ran thousands of product reviews for these 3 products through our sentiment analysis program. Our goal here was to 1) identify common reasons for churn by singling out negative reviews, and 2) show you how to identify signs of future churn in reviews that are largely positive in sentiment.

After filtering for reviews written by customers who likely churned, this is the distribution of reasons for churn that we ended up with.

We found that the majority of customers ended up churning because of the taste, more than anything else.

And when we had our program tag any negative feedback found within positive reviews, we found around 40 comments like this.

Despite customers leaving 4 or 5 stars, each still had a bit of negative feedback to offer. And skipping over these pain points can be costly in the long run.

Meaningfulness of Results

An obvious use for sentiment analysis is simply collecting feedback to improve the product or customer service. But sentiment analysis is taken to a whole other level when you have a tool doing it 24/7 and in real time.

Say a customer leaves a review like this: “The elderberry was delicious! But our Apple Cider had some issues. Some of the packets were solid, as If liquid got in at some point. Others had a sticky substance on it.”

In the blink of an eye, a tool like Cotera could isolate this review, tag the negative sentiment and sort it into categories like “Apple Cider Flavor,” “Packaging,” and maybe even “At-Risk of Churning.” It could even alert the customer service/experience team, who might decide to send over a few apple cider packets free of charge to this specific customer in order to make up for this poor experience.

Or, let’s use the product Hydrate as an example. 53 people churned because they simply had issues with the taste. So a good idea might be to send each of these people free samples of alternative flavors they might like better or samples of unreleased flavors for feedback. And if Hydrant is lucky, a few of the customers they win back may even revise their initial reviews.

If they can leverage tools like these to get existing customers to stick around for longer, acquiring new ones (an even MORE difficult task) becomes much less important.

‍