Brief
Zaemirey Party Supplies is an invented online shop selling balloons, tableware and decorations, with a peak before the end-of-year celebrations that carries much of its year. Last peak its best-selling balloon kit sold out while the ads kept running, a code stayed live for days after the deadline, and the review judged the peak on headline sales.
The rest of the year runs on habit. Shopping and Performance Max run on one target from a product feedA structured file of a store's products, prices and availability used to build shopping and catalogue ads. nobody owns. Coded offers go to the whole list most weeks. Lapsed customers get a monthly discount with no end. The owner wants to grow without another peak like the last, and asks for one plan that says who runs what, when, and how anyone will know it worked.
The advanced ecommerce level set profit targets, segmented a catalogue by margin and valued cohorts. This plan is about running those decisions as operations: owned, gated, tested and read.
Inputs and constraints
Zaemirey Party Supplies: the team and last peak
Team: the owner; an ecommerce manager; a buyer who also keeps the feed when asked; a freelance paid media manager; an email and social marketer; a customer service lead with seasonal temps; a warehouse lead; an analyst two days a week, enough to design and read two tests at once.
Last peak: 20% off everything for ten days, ending the week before the celebrations. Contribution margin at full price after goods, packing, delivery and fees: 45%.
| Promoted lines | Orders | Baseline: matching days in a year without a sale |
|---|---|---|
| The ten sale days | 4,200 | 2,600 |
| The three weeks after | 2,300 | 2,900 |
- The review's headline: "orders up 62% on the baseline".
- The best-selling balloon kit sold out on day four; Shopping and Performance Max spent £1,900 on it over the next three days.
- The code was redeemed on 340 orders in the five days after the deadline, through the abandoned basket email and a deal site's page.
Zaemirey Party Supplies: the Shopping account and feed
| Campaign | Set-up | Spend last year | Reported conversion value |
|---|---|---|---|
| Performance Max, all products | One target ROAS of 400%, no brand exclusion, asset groups unchanged for two years | £96,000 | £520,000 |
| Standard Shopping, all products | Low priority, rarely serves | £6,000 | £21,000 |
- Feed diagnostics: 210 items disapproved, 140 of them for price mismatch, recurring weekly because sale prices are typed over normal prices in the store; 45 missing identifiers; best-selling kits among both.
- Conversion actions: purchase and begin checkout both primary. Of 5,200 conversions reported last quarter, 2,100 were begin checkout.
- No custom labels; no monthly comparison with orders.
- Gross margin: balloons 50%, tableware 35%, decorations 55%.
- The paid media manager proposes another £3,000 a month for Performance Max.
Zaemirey Party Supplies: offers, journeys and customers
- Coded offers: an email code to the whole list in 41 of the last 52 weeks. Orders using a code: 58%, against 39% two years ago. Customers reachable on owned channels: 24,000. Personal-offer events the owner would keep if codes were cut: a spring party-season code and an early-summer garden-party code.
- Journeys: welcome with a code; abandoned basket quoting the current code; review request; a "party again?" reminder a year after a birthday order; the lapsed discount. No named owners, no holdouts; the birthday reminder's trigger broke in a platform change and has sent nothing since spring.
- Cohorts were built once, last spring, and never updated. Second order within twelve months by joining season: spring 34%, summer 30%, autumn 25%, peak 14%.
- Service tags exist, but 61% of contacts are tagged "other". Top named tags: arrived after the party date, items missing, kit out of stock after ordering.
| Gap between first and second order | Share of all second orders |
|---|---|
| Within 90 | 30% |
| 91–180 | 15% |
| 181–300 | 15% |
| 301–400 | 30% |
| Over 400 | 10% |
Contactable customers passing a candidate lapse point each month: about 1,400 at six months, about 900 at thirteen.
Constraints:
- The peak build starts at the beginning of September and the change freezeA set period around a peak when changes that are hard to reverse are stopped while routine, reversible work continues through an agreed route. covers the weeks between a finished build and the last day of the window. No test runs or is read across the peak.
- A public summer clearance runs at the end of July.
- Paid budget stays flat until a test justifies more.
- The analyst's time caps how many tests run at once, as the team list sets out.
- State what you expect each change to do as a direction, with the evidence that would show it, and leave the size open.
- Whoever handles compliance reviews offers, price claims and contact with lapsed customers; nothing here is legal advice.
Task
Set the plan out in this order.
- Owners and the year. Name a trading lead and deputy and an owner for the offer rhythm, the Shopping programme and the retention programme. Draw the year: audits, fixes, tests, reads, the summer clearance, the build, the freeze and the window. Log every change with its date in the measurement change log.
- The peak as one programme. Apply the peak season trading playbook: gates from the promotion and its break-even, set with the promotion brief and the break-even calculator, through stock and delivery, build, service brief, freeze and the sale promotion readiness checklist, to the daily stand-up and its decision rights. Write the close-down for every channel from the SOP for ending a sale or promotion, naming the two leaks that kept the code alive.
- Reading the peak. Plan the read with the guide to measuring promotions, counting pull-forwardPurchases a promotion moves earlier in time rather than creates, as customers who would have bought later at full price buy during the offer. and margin given away, and the write-up with the campaign post-mortem template, so its causes reach next year's first gate.
- Offer rhythm. Decide whether personal coded offers are cut and design the test with the full-price protection test: cells, planned events, metric, guardrails, window, cool-down and decision rule.
- The Shopping programme. Apply the Shopping and Performance Max playbook: split product decisions between buyer, feed owner and campaign manager, set the rhythm, score the account with the feed and Performance Max audit gates first, and order its fixes.
- Proving Performance Max. Answer the budget request with the Performance Max uplift test, run past the holdout test readiness checklist.
- The retention programme. Build the journey inventory with owners and, where a journey can afford it, a holdoutA randomly chosen part of an audience kept from exposure to measure what exposure caused., from the lifecycle email programme playbook, score the customer retention audit, set a lapse definition that respects annual occasions, and design the win-back offer test.
- Reading rules. For each test and change: the owner, the read date, the baseline agreed beforehand, the decision rule and the experiment results report it ends in.
- Summary. A single page the owner can act on: the order of changes, the test or reading behind each, and the signal that would halt it.
Evaluation criteria
| Criterion | What a strong answer shows | Common gap |
|---|---|---|
| Peak run as one programme | Names a trading lead and deputy, gates from the promotion and break-even to the review, a freeze with an exceptions route, a daily stand-up with decision rights and a close-down list for every channel | A promotion calendar with no owner, gates or close-down |
| Offer rhythm and full price | Tests fewer coded offers on customers assigned at random over several cycles, read on gross profit per assigned customer with active customers, unsubscribes and public-sale orders as guardrails | Offers cut for everyone and the year compared with the last |
| Shopping programme | Splits product decisions between merchandiser, feed owner and campaign manager, fixes the feed at the source, refreshes labels and assets on a rhythm and orders the audit's fixes by its gates | Performance Max scaled before the feed and conversion value gates pass |
| Proving Performance Max | Designs a regional uplift test read on total account value, with brand, new-customer and blended-cost guardrails and a decision rule written before launch | The campaign's own report read as proof |
| Retention programme | Keeps a journey inventory with owners and holdouts, reads churn by cohort and reasons for leaving before any offer, and tests win-back against a holdout on incremental gross profit | A loyalty scheme launched to fix churn nobody has read by cohort |
| A calendar that can be read | Orders the year so each change and test is read alone, keeps tests clear of the peak and judges every result against a baseline agreed beforehand | Tests, structure changes and the peak overlapping so nothing can be read |
Order decides most of the grade. Both audits come first. The feed audit's gates score missing, since nobody reads the diagnostics and begin checkout counts as a purchase, so bidding has chased the wrong figure and the budget request waits. The retention gates are partial at best: a cohort view abandoned since spring and most reasons tagged "other" mean nobody can yet say which fix keeps customers. The lapsed discount with no end stops at once. Feed and value fixes then need a settled month before the uplift test starts, clear of the summer clearance.
The two customer-level tests both turn on coded offers. A strong plan runs them one after the other, or randomises them independently, stores both flags on the customer record and writes the overlap into both plans; no flow issues a code without checking the cell. The lapse point is the trap in the retention data: many second orders arrive around a year after the first, so a definition borrowed from a shop with a monthly cycle would mail party customers before their next birthday came round. Every test closes and is read before the peak goes live, and the peak's own post-mortem opens next year's plan, as the winter peak exercise and the retention audit and win-back exercise each practised for one part.
Settle your own numbers before opening the reference set below.
Reference figures for self-checking
- Last peak's break-even uplift: 0.20 ÷ (0.45 − 0.20) = 80%. Sale-day uplift: (4,200 − 2,600) ÷ 2,600 ≈ 62%, short of break-even before pull-forward. Weeks after: 2,300 − 2,900 = −600, so the net gain was 1,600 − 600 = 1,000 orders.
- Break-even ROAS by range: balloons 1 ÷ 0.50 = 200%; tableware 1 ÷ 0.35 ≈ 286%; decorations 1 ÷ 0.55 ≈ 182%. One 400% target treats three different economics as one.
- Conversions that were not purchases: 2,100 ÷ 5,200 ≈ 40% of last quarter's reported total.
- Second orders still to come after six months: 15% + 30% + 10% = 55%; after thirteen months, about 10%.
- Win-back enrolment at thirteen months over three months: about 2,700 customers, 900 per arm across three arms.
