01 Context
Retention went up. So did friction.
Cancelling meant calling the contact centre, whatever the reason. Nobody asked whether the policy needed to be cancelled at all, so people who only needed to update an address or remove a vehicle left anyway.
In Q4 2024 the Save desk pilot made that rule universal: every customer had to call in order to cancel. Retention improved. So did wait times, call volume and frustration for the people who had a legitimate reason to leave.
↑ The pilot flow sent users to the contact centre for any cancellation reason, which hurt NPS. Recorded session sourced from FullStory.
Challenge
How might we retain customers who can be assisted digitally,
whileletting those who truly need to cancel navigate the flow without friction?
02 Goals
Three perspectives, one flow
To clarify the project's objectives, I documented its goals and key considerations across the three stakeholders the flow had to serve at once.
Retain where possible, and take pressure off the phones.
measured by call volume + churnKnow whether cancelling is even necessary, then do it without a fight.
measured by SEQ + completionFit existing policy systems, honour provincial rules, and ship inside pilot constraints.
measured by compliance sign-offGoal 01 and goal 02 pull in opposite directions. Everything downstream is an attempt to satisfy both at once.
03 Process
Five stages, each tied to a goal
The work was structured into five stages, from understanding user behaviour to validating measurable outcomes, ensuring every design decision mapped directly to a business goal.
- 01
Research spike
Why customers cancel, and what the business needs.
→ save moments - 02
Strategy & concept
Retention logic translated into flows covering every scenario and province.
→ rebuttal model - 03
Design & iteration
Modular, reason-based flow built with stakeholders in the room.
→ hi-fi journeys - 04
Delivery & QA
Design accuracy and logic verified against the build.
→ edge-case matrix - 05
Monitor & measure
Behaviour tracked, results read, next experiments planned.
→ 78% deflected
🤖 How AI accelerated this
AI ran alongside the whole process. Copilot mapped the edge cases in the moving flows and drafted French copy for professional translation to refine. I prototyped rebuttal concepts with AI before committing to mid-fi, and drafted the flow logic with it, cutting documentation time for the dev handoff.
04 Research spike
Finding the "save moments"
I reviewed session recordings and sat with contact-centre advisors to understand why people leave and where retention was actually possible, looking for patterns digital design could handle.
Find "save moments"
Many customers were not leaving out of dissatisfaction. They discovered benefits or offers they had not known about, and those moments convinced them to stay.
Map cancellation reasons
Categorizing reasons like moving or selling a vehicle surfaced chances to guide customers to a helpful alternative instead of cancellation.
Design digital empathy
Rather than replicate the call, we reimagined its intent: the right offer, guidance or reassurance surfaced in context, the way an advisor would raise it.
05 Strategy & concept
Rebuttals, right when they matter
The goal was to bring the contact centre's retention logic into a personalized digital experience that covered every pattern and edge case. I started with a hypothesis.
"If users see relevant benefits, offers, or guidance at the right time, we can reduce unnecessary cancellations and call volume while maintaining user trust."
I introduced rebuttal steps into the flow for the two most saveable reasons, moving and no longer owning a vehicle, covering bundled policies and out-of-province cases.
From there I added a new step to the journey, the rebuttal: what you already have with Sonnet, and what more you get by staying, such as the bundle discount and complimentary TELUS Health benefits. I mapped optimized flows for the two most saveable reasons, moving and no longer owning a vehicle, starting with journey maps and flow charts.
User journey: cancellation due to moving
prioritized: also covers “i sold my vehicle”
Stage
Awareness
Reason selection
Rebuttal ← new
Decision
Confirmation
Steps
Thoughts
Touch‑points
Emotion
Confused
Unsure if cancelling is even required
Neutral
Expecting a quick exit
Informed
Learns about update options
Relieved
Resolves it without cancelling
Satisfied
Confident and informed
↑ The rebuttal stage is the one step that did not exist before, and it is where the curve turns from confused to informed.
One system, three surfaces
Customer flow
Reason-based cancellation journeys and rebuttal cards.
Advisor logic
Contact-centre retention playbooks, digitized into save moments.
Policy systems
Eligibility, bundles, and the geospatial servicing database.
Every screen decision ripples into advisor workload and system constraints, so each flow was checked against all three surfaces.
Rules that shaped the flow
Policyholders can cancel at any point within 15 days of their policy start date, so the flow has to let them through.
Sonnet does not write auto there, so a customer moving in is a genuine cancellation, not a save opportunity.
Home and auto together means cancelling one changes the price of the other, which the flow has to disclose.
Every rule above was agreed with compliance and legal before it reached a screen.
"I am moving or have moved" was the most complex reason. It surfaced bundled home and auto customers leaving the province, customers moving somewhere Sonnet does not write auto, such as Alberta and Manitoba with their provincial coverage, and integration with the geospatial servicing database. After aligning with compliance and the business, I defined a modal and logic that let customers confirm their eligibility and cancel online when Sonnet genuinely cannot insure them.
06 Design & iteration
Digestible, on-brand, and built to keep users
The flow had to be easy to navigate, cohesive with Sonnet's established brand, and persuasive enough to keep customers. To avoid overwhelming people, I introduced collapsible components so each step stays digestible.
Lo-fi to mid-fi: finding the structure
Lo-fi exploration
Lo-fi, the selected direction.
Mid-fi, walked through with the dev team. Clip from the actual working session, tab through the reason-based states as the team pressure-tests the logic. The walk through took place before the design is fully finalized.
Hi-fi: tailored journeys for each cancellation reason
The final structure. Each reason routes to its own journey, with collapsible steps keeping the flow digestible.
Each cancellation reason routes to its own journey, so the rebuttal a customer sees speaks to why they are actually leaving.
Hi-fi in the browser: the moving journey, fully opened
Interactive mockup. It's a progressive closure in the flow, but this one is showing the state where all cards are opened. Feel free to collapse each rebuttal!
Delivery & QA: owning the logic
Every flow shipped with its logic defined. I maintained the acceptance criteria and edge-case matrix, covering bundles, out-of-province moves, non-serviced provinces and the Quebec 15-day rule, alongside annotated flows, then reviewed the test build against them before release.
07 Post-launch
Watching it work, then evolving it
The team ran a review session to watch and analyze real customer behaviour: what is working, what to improve, and where the journey goes next.
↓ 78%
requests deflected digitally
↑ 44%
successful self-serve completions
→ What's next: more cancellation reasons covered by rebuttals, and AI-personalized rebuttals that adapt the save moment to each policyholder.
The screen doing the work: the rebuttal a mover sees before deciding.