Deflecting churn: a 3-phase retention strategy

Hotjar is a subscription-based analytics tool. Over two years, cancellations had grown 23% compared to 2022, voluntary cancellations representing 75% of all churn. The product had a generic deflection modal that reduced churn by 5% (contributing $800–900K in ARR), but it treated every cancelling user the same, regardless of why they were leaving.

The hypothesis: if we could personalize the content response to the user's specific churn reason, we could meaningfully reduce voluntary churn. This became a year-long initiative across three phases.

The existing modal was generic. My hypothesis: personalizing the content around what the user would concretely lose (not just a list of features) could shift their decision. We run 2 sets of user interviews where we elaborated 2 versions, and then a A/B test running at the same time.

V1: Usage data + before/after + feature list by plan. Focused on sessions and responses they'd lose, with plan-specific feature comparison and an exact cancellation date.

Research findings:

👎 Users didn't express concerns about losing coverage, since it was expected.
👍 Users understood the structure but focused more on feature loss than usage data.
👍 Before and after with the exact date was positive.

V2: Feature-first hierarchy, clarified cancellation language. Bigger focus on the features they use most (from interview data), and specify what it meant "to continue" and what cancellation means.

Research findings:

👎 "Downgrade" vs "cancel" language created confusion in the same flow.
👎 Some users thought the metrics were generic, not personalized.
👍 Users understood the purpose of the modal, when changes will be applied, and what they'll lose.

As seen before, the churn deflection modal works and brings awareness to what will happen, however, that isn't a incentive by itself. What is? An offer.

So my team and I analyzed churn survey data to identify the biggest cohorts and decided to offer 1 month free to these users. "Budget reasons" and "Too expensive" together represented ~38% of churning users, the largest addressable group. Other than these 2, we also decided to focus on "Rarely used" and "Temporary pause", the main 2 reasons. By targeting these first, we could cover 80% of all churn across three cohorts with tailored content, which could mean $1.4M in ARR if we deflect 5% of the churn.

How did I come up with the content proposition?
Reviewed several churn surveys, grouped by their reason, and went through all their written feedback. With this, I was able to create several personas, their main concerns, and what could potentially make them stay.

Getting users to accept an offer is only half the problem. If they don't find value during the free period, they churn anyway at the end of it. Phase 3 was about building a content strategy to make the free month(s) actually count.

I organized a cross-team brainstorm, mapped churn reasons to personas, and worked with Marketing and Education to identify which resources drove the most signups and account creation. Then I matched the top-performing content to each cohort — not generically, but with headlines rewritten to speak to that cohort's specific motivation.