WeightWatchers

Reducing churn and improving retention with AI

Redesigning the cancellation funnel

While managing the growth design team at WeightWatchers, we redesigned the cancellation experience with the goal to reduce churn and improve retention. To do this, we wanted to explore the potential for AI to generate specific save offers which would increase the likelihood of users keeping or extending their membership subscription. This project was run as an experiment against the existing cancel flow and is now the current experience. WeightWatchers is a health and wellness product for weight-loss, offering both behavioral and medication-assisted programs.

Role: UX & UI design

Tools: Figma, Amplitude

Team: 2 product designers, 1 product manager, 2 AI engineers

Timeline: 4 months (January 2023 - April 2023)

PROBLEM STATEMENT

The problem

A large percentage of members were cancelling their membership immediately after signing-up and immediately after their plan was set to auto-renew at a higher monthly price. The most frequent cancel funnel visits occurred on day 1 (2.7%) or day 91 (1.4%). Churn rates peaked at up to 40% when a member’s initial 3 or 6-month long-term-commitment (LTC) ended, especially within the first month post-LTC and this trend was increasing.

Success metrics

Cancel Conversion Rate: Decrease the percentage of users who complete cancellation after entering the cancel funnel

01

Save Offer Acceptance: Improve the acceptance rate of personalized offers presented during the cancellation funnel

02

Membership Revenue Impact: Track the revenue per member to evaluate the impact of retention compared to previous experience

03

Membership Duration Extension: Measure the average length of continuous membership extensions from the new cancellation funnel

04

Data analysis

We reviewed months and years worth of data on cancellation trends across different memberships as well as save offers and tactics which had previously reduced churn. We also reviewed countless surveys and user interviews on cancellation reasons, giving insights into the emotional and psychological factors contributing to cancels.

30%

Financial reasons

30% of users choose ‘financial reasons‘ when asked in the cancellation flow. Users reported that WW was “out of their budget“ and they couldn’t justify the cost because they’d fallen off their program.

45%

Churn post-recur bill (RB)

45% churn rate for people who were transitioning from LTC to RB. This number had been increasing as the business has drove more LTC subscriptions.

18%

Save offer acceptance rate
18% of users accepted a 30-day free time credit when offer was presented in the cancellation experience but 80% of those who enter the cancel funnel were “strugglers“ and not using the product.

Competitor Research

Analyzing competitors provided valuable insights into industry standards and opportunities for improvement in the FitLife app's onboarding process.

HelloFresh

Strengths:

  • Asks questions that clearly reflect user research

  • Offers are compelling and attractive

Weaknesses:

  • 8+ steps depending on reasons

  • Long scrolling screens with high cognitive load and too many options

Headspace

Strengths:

  • Appealing visual design

  • Value props restated (FOMO)

Weaknesses:

  • Requires written response after selecting cancel reason

  • Does not have offer monetary incentive

Noom

Strengths:

  • 3 steps to cancel

  • 1 save offer at last step

Weaknesses:

  • Feels impersonal

  • Asks only one cancel reason question

User Flow

This flowchart illustrates the steps a user takes during the cancellation process, from initial log in to cancellation confirmation or membership retention.

Screen Flow

This screen-flow shows the old UI and top-3 cancel reasons and only save offer. This flow and save offer (2 months free) had a 9% acceptance rate.

Variant exploration

EXPLORATION ONE

Cancel survey & feature/benefit emphasis and AI feature offers

Pros

  • Mixing user-centered cancel reasons with AI-generated offers.

  • Reminds users of and product value/features they may not use.

  • Does not rely only on monetary offer.

Cons

  • May redirect users out of cancel flow and cause frustration

  • May add steps lengthen user flow

EXPLORATION TWO

Cancel survey with branching & AI generated offer as last step

Pros

  • Engages users through a more personalized cancel flow.

  • Creates more opportunities for various save offers and tactics

Cons

  • Requires more development effort.

  • More complex and design-intensive, high-risk for an MVP experiment

EXPLORATION THREE

No survey, only AI-generated save offers

Pros

  • Quick and straightforward, reducing speed to testing AI offers.

  • Targets 30% of users cancelling for financial reasons.

Cons

  • Overly simplistic and not user-centered

  • Sacrifices learnings from self-service cancel surveys

  • Mostly designed to test the AI bandit.

Selected variant

The cancellation reasons survey and offer flows were chosen for optimization because these interactions had the highest potential for retention success versus focusing on switch plan, pause or downgrade options as part of the MVP. In addition to the design team’s non-AI cancel flows and offers we also determined how best to leverage AI, train the AI bandit and use in-product behavior analytics to offer timely discounts, recommendations, or extensions that might appeal to members.

Despite some lively debate, we ultimately chose the 3rd option, removing all cancel survey questions and immediately presenting users with a feature benefit highlight (based on their past in-product behavior) and then a final AI-determined discount offer. The intention here was to support the Loss Aversion hypothesis, while recognizing users were attempting to avoid the ramp-up price after the end of their existing subscription.

Subscription details and feature list

Step 1 of the new cancel flow simply shows the user their active subscription and the features/benefits and tools of their program. This reduced the number of screens from the old flow and exposed subscription details which had been hard to find.

Loss aversion and exit levers

Based on a user’s engagement status, past in-product feature use and the growing sophistication of the AI bandit (which would be learning through the test), we chose two feature highlights to encourage people to re-explore the app and exit the cancel funnel.

AI-selected save offers

For users with high-intent to cancel, low engagement status and minimal app feature adoption, discount offers were presented first, optimized for plan extensions and long-term commitments at a lower monthly price with higher potential impact on LTV.

Results

Initial results after 4-week test delivered a dramatic reduction in Cancel Conversion Rate, an increase in Save Offer Acceptance and an increase in Membership Revenue Impact and Membership Duration Extension.

~300%

Retained Customer Revenue

Positive impact across all success metrics indicated an early estimate of 200% increase in retained customer value which, by the end of the test period, jumped to an estimated 300% in retained customer value.

+ $4M

Annual Revenue

Variant 3 ultimately added several million in annual revenue with the AI-driven offers and feature highlight redirects deemed a huge success.

Takeaways

Diagnosing with data: The vast quantity of funnel data and in-product user behavior analytics allowed us to clearly determine the opportunities to improve the performance of the cancellation experience and quickly develop design iterations and variants to test. Qualitative data from user interviews and cancel surveys helped us validate our assumptions and identify user attributes to help train the AI bandit and make it more effective.

Disagreeing and committing: Although the design team were supportive of a more personalized user survey in combination with AI driven offers, the product team’s preferred option to better evaluate the potential for AI was still a success and delivered the intended results and a new experience for further experimentation on.