Personalization opportunity
They're using it less. The reply is a coupon.
Find out what changed. A discount isn't a diagnosis.
Step 1 · Signals
What do we know about this person?
Connected
Trailing usage and key-event completion relative to the account's own baseline, not relative to the heaviest users.
Example eligibility rule
Trigger when weekly key-event volume falls below 50% of the account's prior eight-week median for two consecutive weeks, with no planned pause, seasonality flag, or known outage. Require a minimum baseline before calculating decline.
Step 2 · The need
What’s really going on?
Sessions down, key features untouched, seats idle. The inference is churn risk. The honest response is to find out why, not to spray a win-back.
Avoid this assumption
Inferred churn is a hypothesis. Treat it like one: a test, a holdout, a way to be wrong without being loud.
Step 3 · Worth it?
Do the math before you build.
A proportionate, dismissible check-in plus a useful reason to return addresses the possible problem more directly than an automatic discount.
Reach
How many eligible people can you reach in the chosen channel and time period?
Cost
What will content, delivery, QA and ongoing maintenance cost?
Return
What incremental outcome would cover that cost, and what is it worth?
Planning inputs, not measured results. Estimate reach × incremental outcome rate × value per outcome, then subtract cost. Use a range and a consistent time period.
Step 4 · The response
Pick how you help.
Example format: Content
A "what's changed?" note with two useful paths (refresh a workflow, pause).
Compare lighter and deeper responses
Choose the lightest response that solves the need.
Assist: Offer useful help
Silence on the product. One thoughtful check-in, easy to ignore.
Adapt: Change a meaningful part
A "what's changed?" note with two useful paths (refresh a workflow, pause).
Transform: Tailor the journey
A save offer only if they have already said they might leave. Never as the first move.
What to watch out for
- Comparing them with the heaviest users instead of with their own baseline.
- Leading with a discount before asking what broke.
- No holdout, so you never know if you caused the save.
Step 5 · The moment
Did it help? A/B test it.
Control
No outreach, or the default win-back.
Treatment
Baseline-aware, proportionate check-in.
Success measure
Reactivation of the key event, then retained revenue.
Guardrail
Opt-outs and "this feels like surveillance" replies.
Among accounts that met the same decline rule, compare the check-in with no outreach. Measure recovery of the key event and retained revenue, while tracking opt-outs and replies showing the inference was wrong.
Win or lose, record what worked, for whom and why. What you learn becomes the next signal.
Tools that can deliver the response
- Pendo in-app guides Product experience platform combining usage analytics with in-app guides targeted by user or account segments.
- Braze Cross-channel messaging platform (push, email, SMS, in-app, WhatsApp, and more) that orchestrates sends off real-time customer behavior.
- Iterable Cross-channel journeys (email, SMS, push, in-app) built from a single visual studio, triggered by real-time customer behavior.
- Customer.io Event-triggered lifecycle messaging across email, push, SMS, and in-app, built on a customer profile and behavior-data layer.
Try the framework on another situation
They keep doing it the hard way. Show the easier way while they're still doing it by hand.
They started setup, then disappeared mid-step. Remind them of the unfinished step. Then stop.
Work through it in the hub
Score your personalization program, explore more customer situations, and compare the tools that can deliver them.
.svg)
