Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

E-commerce Marketing › Advanced

Valuing Customer Cohorts for a House Plant Shop

An invented house plant shop wants to launch a points scheme and spend more on remarketing to past buyers. You read customer value by acquisition cohort and channel on margin, set acquisition cost limits from payback, cost the scheme with its rewards, breakage and running costs, and design a holdout that shows whether remarketing to past buyers adds repeat orders or only collects credit.
ExerciseAdvanced levelModule 3: Customer value, loyalty and retention spend5 minUpdated

Why this matters

Retention spend is easy to approve and hard to judge, because the customers it rewards were partly coming back anyway. The cohort analysis tutorial builds the table that shows what customers of the same age actually do; the loyalty programmes guide sets out breakage and incremental margin. Here you price two retention ideas before money is committed, and set what each channel may spend per customer.

The shop and its retention ideas

Greenhythe Plants is an invented online house plant shop. The founder wants a points scheme for every customer and more remarketing to past buyers, whose reported return is the best in the account. Finance asks that every channel's acquisition cost (CAC) pays back within six months of contribution.

Illustrative example — not a real result

Greenhythe Plants: cohorts, scheme and remarketing

Cumulative contribution per original customer, after product, packing, delivery, fees and returns:

CohortChannelCustomersCACFirst orderMonth 3Month 6Month 12
Spring, a year agoPaid social1,400£24£8£10£12£15
SpringShopping1,000£17£11£15£19£24
SpringOrganic and email sign-up700£5£10£16£22£30
Autumn, six months agoPaid social1,800£21£8£11£14Not yet
AutumnShopping1,100£18£11£15£20Not yet
AutumnOrganic and email sign-up800£5£10£16£21Not yet

Points scheme as proposed: one point per £1, 100 points worth £5 off, open to all customers. Expected member revenue £700,000 a year, of which repeat orders today are £300,000. Platform £4,800 a year. The proposal assumes 40% breakage. Contribution on repeat orders: 40%.

Remarketing to past buyers on Meta: £3,000 a month; the platform reports 600 purchases a month. The past-buyer audience holds about 9,000 people. Last year, 8% of past buyers ordered again within any 60-day window. Average repeat order £35, so about £14 of contribution.

Reading value by cohort and channel

Read along a row to see where a cohort's curve crosses its acquisition cost, and down a column to compare cohorts at the same age. A blended figure lets cheap organic customers subsidise an expensive channel, so every limit is set per channel. Lifetime value (LTV) here is the twelve-month cumulative contribution, labelled as such; nothing is projected past what the oldest cohort shows.

Task

  1. Find each channel's payback point and twelve-month value, and compare them with the blended view.
  2. Set a maximum acquisition cost per channel from the six-month payback rule.
  3. Cost the points scheme: issuance, redemptions, running costs and the incremental revenue it needs.
  4. Decide whether the scheme should launch as proposed, and what to do instead.
  5. Design a holdout for remarketing to past buyers, with the lift it must show to break even.
  6. Explain how the two tests avoid contaminating each other.

Worked answer

Illustrative example — not a real result

Greenhythe Plants: payback, limits and the scheme's cost

Blended spring view: value (1,400 × £15 + 1,000 × £24 + 700 × £30) ÷ 3,100 ≈ £21.30; CAC (1,400 × £24 + 1,000 × £17 + 700 × £5) ÷ 3,100 ≈ £17.50. It looks healthy.

By channel: paid social reaches £15 against £24 and never pays back within the year; Shopping passes £17 between months three and six; organic pays back on the first order. Autumn paid social is better at month six, £14 against £12, after a creative change, but still short of its £21 CAC.

Six-month limits: paid social about £13, Shopping about £19, organic about £21.

Scheme: points issued 5% × £700,000 = £35,000; redeemed at 40% breakage £21,000; plus platform £4,800 = £25,800 a year. Incremental revenue needed at 40% contribution: £25,800 ÷ 0.40 = £64,500, about 21% more repeat revenue, all of it caused by the scheme. At 20% breakage: £28,000 + £4,800 = £32,800, needing £82,000.

Acquisition. Paid social is the only channel failing the payback rule, and the blend hides it. Narrow it to the audiences behind the improved autumn curve until its cost nears the limit, and move the rest to Shopping. Refresh the limits monthly as cohorts mature.

The scheme. It needs a fifth more repeat revenue that would not otherwise happen, and it only gets cheaper when members fail to redeem. Most rewards would go to habitual buyers. Do not launch it as proposed; pilot a lower-cost design aimed at the second order and at care, such as repotting reminders and early access to rare plants, as the repeat purchase and loyalty guide suggests, with a random share of new customers held out. Judge it on cumulative contribution per original customer at month six against the holdout.

Illustrative example — not a real result

Greenhythe Plants: remarketing holdout design

Hold out 20% of past buyers at random: 1,800 held out, 7,200 exposed, over a 60-day window from audience entry.

Break-even: two months of spend, £6,000 ÷ £14 ≈ 429 incremental orders among 7,200 exposed, a lift of about 6 points on the 8% baseline. The platform would report about 1,200 purchases over the same period.

Rough detectable lift: standard error √(0.08 × 0.92 × (1 ÷ 7,200 + 1 ÷ 1,800)) ≈ 0.7 points, so about 2 points at conventional confidence and power. The test can see the break-even lift with room to spare.

The holdout. Following the remarketing holdout test, randomise once per person and exclude the held-out group from every Remarketing campaign, including catalogue cross-sell. Email continues for both, so the test measures paid remarketing on top of email. Count repeat orders from the order system, watch frequency and cost per incremental order, and record the ratio of incremental to reported purchases. Keep spend above break-even, cut frequency between zero and break-even, and move the budget to acquisition near zero.

Keeping the tests apart. Run the pilot on new customers and the holdout on existing past buyers, so no one sits in both; if they must overlap later, randomise independently and report all four combinations.

Common mistakes

  • One acquisition limit from the blend. Organic customers subsidise the failing channel.
  • Projecting value past the oldest cohort. Label anything beyond it as a guess.
  • Counting on breakage. It pays only when members are disappointed.
  • Trusting reported remarketing returns. Past buyers were likely to return anyway.

Bring it together in the e-commerce profit and scale capstone or revisit the intermediate e-commerce level.

Pages this lesson applies

Terms used on this page

Also related

Related