Choosing a Win-Back Offer Under a Hard Discount Budget
Kettle & Co is a coffee subscription with 48,000 active subscribers and 19,000 lapsed subscribers who have not ordered in 90+ days. Average order value is £34 and gross margin is 41%. Last year a 25% discount to all lapsed subscribers recovered 4.1% of recipients, and 38% of recoveries re-lapsed within 90 days. A prior test of free shipping (worth £4.95, about 14.5% of AOV) recovered 3.6%. A 'we miss you' email with no offer recovered 2.2%. The lapsed file splits three ways: 41% lapsed after one order, 34% after two or three orders, and 25% after four or more. Subscribers who lapse after four or more orders recover at 9% on a no-offer email. The team has a six-week window and a hard budget of £4,000 in offer cost. Leadership wants 'the highest ROI' and has asked for one campaign.
1. Run the unit economics and recommend a campaign structure rather than the single campaign leadership asked for.
Read the worked answer
Do the unit economics first. Gross profit per order at £34 AOV and 41% margin is £13.94. After a 25% discount (£8.50) it is £5.44; after free shipping (£4.95) it is £8.99.
Gross profit per recipient:
- No offer: 2.2% x £13.94 = £0.307
- Free shipping: 3.6% x £8.99 = £0.324
- 25% off: 4.1% x £5.44 = £0.223
The 25% discount recovers the most orders and makes the least money, about 27% below doing nothing. It also breaks the budget: 19,000 x 4.1% x £8.50 = £6,622 against a £4,000 ceiling. Free shipping costs 19,000 x 3.6% x £4.95 = £3,386 and fits.
Do not treat the file as homogeneous. The 4+ order group is 25%, about 4,750 people, and recovers at 9% on a no-offer email: roughly 428 orders x £13.94 = £5,966 of gross profit at zero offer cost. It is also the group most likely to be alienated by a discount.
Recommended shape.
- No-offer 'we miss you' to the 4+ segment first. Free, high yield, protects margin and status.
- Free shipping to the remaining 14,250 in a second wave: about £2,540 offer cost and roughly £4,610 gross profit.
- No 25% discount this cycle. Hold it in reserve for a segment that ignores two cheaper attempts.
Caveat. These recovery rates come from different periods and audiences. Apply a 20–30% decay to any repeat, and measure against a holdout rather than trusting last year.
2. State which metrics you would report instead of recovery rate, and why recovery rate misled the team last year.
Read the worked answer
Recovery rate rewards the campaign that buys the most orders, which is exactly why the 25% discount looked like the winner. Four better measures:
- Gross profit per recipient. It is the only metric that ranks the three options correctly, because it accounts for discount depth as well as response.
- Incremental recovery, not recovery. Some 2.2% of lapsed subscribers would return anyway. The 4.1% from the discount contains that baseline, so the true lift is about 1.9pp. Without a holdout you are paying £8.50 to accelerate a purchase that was already coming.
- 90-day re-lapse rate. At 38%, a recovered subscriber is worth roughly 0.62 of a retained one even before repeat purchases. If the discount cohort re-lapses faster than the free-shipping cohort, which is plausible for discount-trained customers, the lifetime gap widens against the discount.
- Discount dependency. Track what share of each recovered cohort pays full price on their next order. A segment that only buys on discount has materially lower LTV than first-order gross profit implies.
Report them together on one page: segment, offer, recipients, recovery %, incremental recovery against holdout, offer cost, first-order gross profit, 90-day re-lapse, projected 12-month gross profit. That answers 'highest ROI' in the terms leadership actually means.
Set the holdout now. Withhold 10% of each segment from all offers for the six weeks. It costs almost nothing and it is the only way to know next year whether the programme is working or simply harvesting subscribers who would have returned.
3. The team wants to send to all 19,000 lapsed subscribers in week one. Explain how you would sequence the six weeks and what you would do differently for the 41% who lapsed after a single order.
Read the worked answer
Do not batch all 19,000 into week one. Three reasons. It spikes complaint rates and risks deliverability on a file that has not been mailed for 90+ days. It consumes the entire budget in one irreversible decision with no read on response. And it treats three very different churn reasons identically.
Sequencing.
- Week 1: no-offer email to the 4+ order segment (about 4,750). Highest yield, zero offer cost, and it protects the relationship with your best lapsed customers.
- Week 3: free-shipping wave to the 2–3 order segment (about 6,460), after the 4+ responses are visible.
- Week 5: one-order segment (about 7,790), with whatever budget remains after actual redemptions, not projected. Staggering also lets you pace budget against realised offer cost rather than forecasts.
The one-order segment needs a different treatment. These people are trial churn: the first order did not become a habit. The cause is product fit, delivery experience or onboarding - not price sensitivity. A 25% discount here trains them to wait for discounts and does not fix the underlying reason they left. Better: a no-offer email asking what went wrong, with a one-click preference or grind-size update, plus free shipping as the fallback for non-responders. Expect a lower recovery rate and a higher proportion of recoveries that stick.
Measure by cohort, not in aggregate. A single blended recovery rate across the three segments will hide the fact that the 4+ no-offer programme is carrying the result.