Win-Back and Resurrection: Bring Dormant Users Back and Prove It Worked

Why this matters

Every product has a growing pool of people who used it and stopped. For a company a few years old, that pool is often several times larger than the active base. These people already know your brand, already gave you their contact details and may already have their payment details saved. Bringing a fraction back can cost less than acquiring the same number of new customers.

But win-back is also the most over-credited activity in CRM. Dormant customers return on their own all the time: payday, a seasonal need, a friend’s mention. Send them a discount the week before and every one of those natural returns shows up in your campaign report. Teams then scale the discount, train customers to lapse and wait for an offer, and never learn that most of the “recovered” revenue was coming anyway.

This lesson covers both halves: how to bring people back, and how to know you did.

The framework

User states and definitions

You cannot manage resurrection without fixed definitions. Duolingo’s public growth model is a clean template, and you can adapt the windows to your own cycle.

Resurrected users usually retain worse than current users but better than brand new ones. Track them as their own cohort, because a campaign that doubles resurrections but halves their 30-day retention may add nothing.

Timing: anchor to your cycle

Find the gap between consecutive orders (D2C) or active periods (SaaS) and look at its distribution. If 80% of reorders happen within 45 days, a customer at day 50 is past the point where most people return unaided. That is your first win-back trigger. Earlier and you pay to nudge people who were coming back anyway. Much later and response tends to fall as habits form elsewhere.

Offers: the win-back ladder

Order the reasons to return from cheapest to most expensive and move down the ladder only when the previous rung fails.

Product-led resurrection

The strongest reasons to return live in the product, not in the inbox. Examples that work across categories:

• Progress that persists. Streaks, levels, saved carts, history. Remind people what they built.
• New value since they left. Ship a feature that addresses a known churn reason, then tell exactly the users who left for that reason.
• Collaborative pull. A teammate mentions them, shares a document or assigns a task.
• Re-onboarding. Returning users land on a “what changed” screen, not the main dashboard they left months ago.

Step-by-step playbook

1. Fix the definitions. Write the active, at-risk, dormant and resurrected windows from your interval data. Get product, CRM and finance to sign off so everyone counts the same way.
2. Size the pool and its value. Count dormant users by segment, past value and reason for lapse where known (delivery issue, price, cancellation survey answer).
3. Set up the holdout before launch. Randomly assign 10 to 20% of each eligible segment to a no-contact group. Keep the assignment for the whole campaign window.
4. Build the ladder as a journey. Free rungs first. Escalate to incentives only for users with high past value or strong recent intent signals (site visit, app open, email click).
5. Pick channels by cost. Email and push carry the early rungs. Use WhatsApp or SMS only for high-value segments where the holdout shows it earns its cost.
6. Re-onboard on return. The first session or order after a long absence gets a dedicated experience: what is new, a quick win, and a reason to come back the next week.
7. Measure incrementally, then follow the cohort. Report incremental returns, cost per incremental return and the 30, 60 and 90-day retention of resurrected users.
8. Sunset. Users who ignore the full sequence move to a low-frequency or suppressed list. This protects deliverability and channel quality.

Worked example

Say a D2C personal care brand has 50,000 dormant customers (no order in 120 days). It runs the ladder above with a 10% holdout. Average order value is ₹1,500 and the incentive rung is 20% off.

Raw result. 45,000 are messaged and 1,530 order within 30 days: a 3.4% return rate. The team’s dashboard says 1,530 customers recovered.

Holdout. 125 of the 5,000 held-out customers ordered anyway: 2.5%.

Incremental. 3.4% - 2.5% = 0.9 points. Incremental returns = 45,000 x 0.009 = 405, not 1,530.

Cost. Say every returning treated customer used the 20% code: 1,530 x ₹300 = ₹4,59,000 in discount, plus ₹40,000 in messaging costs = ₹4,99,000.

• Naive cost per recovered customer = ₹4,99,000 / 1,530 = ₹326.
• True cost per incremental returner = ₹4,99,000 / 405 = ₹1,232.

If the brand’s blended CAC is ₹1,600, the campaign still looks reasonable. But the next test is obvious: drop the discount for customers who were already clicking, where the natural return rate is highest, and see whether incremental returns hold up at a fraction of the cost.

Holdout size check. To detect a lift from 2.0% to 2.6% at 95% confidence and 80% power you need about 9,800 users per group:

Small segments therefore need a longer window or a pooled holdout across several win-back campaigns.

B2B SaaS vs D2C: what changes

For cancelled SaaS customers, the best window is often when you ship the thing they asked for. Tag cancellation survey reasons in your CRM so you can message exactly those users when it lands.

How to measure it

The gold standard is a long-running global holdout: a small random share of dormant users who never receive win-back messages for a quarter. It tells you the true baseline return rate across all campaigns combined, which is the only reliable way to avoid crediting your CRM programme for returns that were going to happen anyway.