Brief
Stouvoth Hot Tubs, an invented online seller of hot tubs, swim spas and their chemicals, takes its orders mainly in spring and early summer. Google Ads and Meta both bid towards brochure requests, finance counts delivered orders after cancellations, and the two figures have drifted apart. An agency proposes a server container for every tag, the board wants revenue reported after cancellations, and Stouvoth has agreed to buy a smaller spa retailer whose ad accounts, analytics property and customer list overlap its own.
Each request is reasonable alone; made together, or in season, they leave nobody able to say which one moved the numbers. The advanced performance marketing level built clean signals, holdouts and profit targets. This plan changes how the programme collects, bids on and reports its results while it keeps trading.
Inputs and constraints
Stouvoth Hot Tubs: figures, tags and conversion actions
| Monthly average | Brochure requests | Orders placed | Cancelled before delivery | Delivered orders |
|---|---|---|---|---|
| March to June | 5,200 | 340 | 45 | 295 |
| July to February | 1,900 | 110 | 12 | 98 |
- Year on year: brochure requests up 38%, delivered orders up 3%. Reported revenue counts orders placed: £15.2m last year, £13.4m after cancellations. Delivery takes two to six weeks, and cancellations fall in that gap.
- Web container: 16 vendor tags, 4 unnamed; the configurator and finance calculator render on the page; a chat widget fires before the consent choice. Mobile product pages are slow, largely from vendor scripts.
- Google Ads: brochure request primary; online order and phone call secondary. In season the online order action records about 190 a month against 340 orders; the tag fires on every test order; buyers take weeks and change device.
- Meta: pixel only, optimising for brochure requests; the team would like its server integration.
- Hosting: a contract developer, three days a week, with nobody on call. Agency plan: every tag in one February release.
Stouvoth Hot Tubs: dashboards, tests and the acquisition
- Dashboard inventory: 10 dashboards read the revenue key; the agency report and the board pack calculate it in their own fields. A price-promise page test, judged on revenue per visitor, ends in August.
- The spa retailer completes in early January. Its Google Ads account bids on Stouvoth's name and 80 shared generic terms; 2,100 of its 9,000 customers are Stouvoth customers too, held in a separate CRM. Its analytics scorecard: consent gate missing (tags fire before choice), key events partial (brochure downloads beside purchases), reconciliation never checked. Its order system shows about 70 orders a month in season.
Constraints:
- The plan runs from July to the following June, and nothing structural (a cutover, a goal switch, an account merge, a list change) happens from March to June.
- The head of marketing owns the plan and approves exceptions; finance signs anything it reports; whoever handles data protection decides what may be collected and who may be contacted. Nothing here is legal advice.
- State the direction each change should move its reading, and the evidence that would show it; leave the size open.
Task
Work through the plan in the following sequence.
- Owners and calendar. Name an owner per change and draw the year: prerequisites, cutovers, switches, reads, completion, the freeze and the peak. Each change enters the measurement change log with its expected effect written first, so the change logA dated record of changes made to tracking, accounts, reports, lists or rules, with who made each and why, so later movements in the figures can be traced. explains every step in the year's charts.
- Server-side decision. Match each platform's gap to its fix with the enhanced conversions or server-side comparison. For server-side taggingRunning tags on a server the business controls instead of in the visitor's browser., apply the case test in the server-side tagging guide and list the prerequisites.
- Cutover. Plan it destination by destination through the server-side tagging go-live checklist, as in the riding equipment retailer exercise.
- Bidding goal. Define a delivered order and the conversion actionA Google Ads definition of one action to count, such as a purchase or lead, with its source, category, counting method, value and conversion window. that brings it back from the order system, then stage the switch with the conversion goal change checklist.
- Revenue. Run the metric definition change checklist for revenue after cancellations, in the order the craft kit subscription exercise used for a north star.
- Acquisition. From completion day, follow the merger and acquisition tutorial: overlap rows, a freeze with one approver and a BaselineAn estimate of what a metric would have done without a given change, campaign or promotion, against which the effect is measured. for each business, the spa retailer's read from its orders while its analytics scorecard fails a gate. Audit its account with the Google Ads account audit checklist and gate its customer list with the CRM sync change checklist, as the cloth nappy maker exercise did.
- Reading rules. For each change, a row of the marketing decision log: owner, date, the reading that judges it, a check date and the signal that would reverse it.
- Summary. One page the head of marketing can act on: the order, why, and what each change waits for.
Evaluation criteria
| Criterion | What a strong answer shows | Common gap |
|---|---|---|
| Server-side decision | Names the gap each platform has, chooses enhanced conversions, a server container or neither yet, and passes the case test with the requests removed, the vendors called from the server and the person on call | A server container bought to recover conversions it cannot recover |
| Cutover | Moves one destination at a time with the consent state passed through, personal data stripped, deduplication confirmed, a parallel run with an agreed end and the previous version ready to restore | Browser tags removed on the day the server tags go live |
| Bidding goal change | Writes the definition of a delivered order, observes the new event against the business's records before it carries bidding, changes one platform at a time and resets targets only after the window closes | Both platforms switched in the same week, inside the spring peak |
| Metric redefinition | Treats revenue after cancellations as a new version, restates history where the rows allow, lists every dashboard that reads it, re-bases targets with a guardrail and leaves running tests on their own rules | The definition edited in place and explained afterwards |
| Acquisition | Works the overlap rows, freezes both programmes, keeps both baselines with the acquired one read from its orders until its failed analytics gates are fixed, and gates the CRM sync change before any customer moves | Accounts merged before either baseline is recorded |
| A calendar that can be read | Sequences the changes so each is logged as a break and read alone, with nothing structural inside the spring peak | Every change made in the same month |
The inputs already point the way. Google's shortfall sits where the base tag fires and buyers take weeks across devices: a matching gap for enhanced conversions in the web container. The server case rests on Meta's integration, page weight and control, and waits for the chat widget's consent fault and an on-call owner. The February release leaves no room for a parallel run before the season.
Brochure requests have outgrown delivered orders, so bidding on them buys brochure readers. Delivered orders arrive weeks after the order, so the new event is observed long enough to see late arrivals. The revenue change is reporting, not bidding, so it keeps its own switch date, and the price-promise test finishes on its old rule. Completion in January starts the freeze and both baselines, sends the consent gate to the data protection owner at once and holds account and list moves until July. The sequence is what the grade turns on.
Reference figures for self-checking
- Delivered orders per brochure request: 295 ÷ 5,200 ≈ 5.7% in season; 1.03 ÷ 1.38 ≈ 0.75, so about a quarter fewer per request than a year ago.
- Cancellations: 45 ÷ 340 ≈ 13% in season. Revenue after cancellations: £13.4m ÷ £15.2m ≈ 88% of today's figure.
- Google Ads online orders against the order system in season: 190 ÷ 340 ≈ 56%.
- Shared customers: 2,100 ÷ 9,000 ≈ 23% of the spa retailer's list.
- One workable order: July, consent fix and unnamed tags removed; August, enhanced conversions; September and October, server destinations one at a time, Google Ads last; revenue switched on 1 October; delivered orders observed from September, Google switched in November once its cutover has settled, Meta in late January once Google's window has closed, a freeze exception the head of marketing approves; completion, freeze and both baselines in January; acquisition moves from July.
