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

Analytics › Expert

Capstone: A Measurement Governance Plan for a Used Camera Dealer

The closing project of the expert analytics level. You write a year's measurement governance plan for an invented used camera dealer whose tags were added by three agencies, whose trade-in and finance journeys pass through other companies' software, whose consent and deletion stop halfway through the stack and whose last outage surfaced as a wrong board report. The plan covers one event contract, embedded tools and identity, consent and retention, a scored audit, the outage procedure and the rule for correcting figures.
Capstone projectExpert levelModule 3: Measurement as a governed service8 minUpdated

Brief

Ternovask Cameras is an invented dealer in used cameras and lenses. It sells online, buys stock from customers through a trade-in quote tool from a third-party provider, embedded in a frame on its "sell your kit" page, offers finance through the finance provider's hosted checkout, and lets regular buyers sign in to save searches.

Its tags were added by three agencies over the years. Two tools ignore the consent banner. A customer's deletion request reached the email platform but not the warehouse. And a tracking outage last quarter went unnoticed until a board report was wrong and budget had moved on the strength of it. The head of ecommerce, who owns measurement, wants one plan for the year instead of a fix each time something breaks. Where the advanced analytics level set metric definitions and strategy, this plan governs the collection beneath them.

Inputs and constraints

Illustrative example — not a real result

Ternovask Cameras: the tag inventory

Tag or scriptAdded byFires onConsent handling
Analytics purchaseFirst agencyA push on the order confirmation pageConsent mode
Google Ads purchaseSecond agencyClicks on "Place order"Consent mode
Meta purchaseThird agencyConfirmation page views, value read from the page textAdvertising category
Meta server eventsShop platform integrationEvery order, with a random identifierAdvertising category, read from the stored consent state
Session recording scriptHard-coded in the themeEvery page, signed-in pages includedNone
Trade-in provider's own analyticsThe provider, inside its frameEvery quoteNone
Illustrative example — not a real result

Ternovask Cameras: the tools and last quarter's notes

ToolPersonal data it holdsFed by
Shop platformOrders, accounts, saved searchesCheckout and sign-in
Trade-in quote toolSellers' details and their kitIts own form
Finance providerApplications and decisionsIts hosted checkout
CRMBuyers, sellers, a do-not-contact flagNightly shop sync; trade-in exports uploaded by hand each week
Email platformNewsletter and saved-search alertsCRM sync
WarehouseOrders, the analytics export, a copy of the CRMDaily loads
  • Orders last quarter: 1,650 in the shop platform. Google Ads claimed 1,210 and Meta 980, and the board pack added the two.
  • Finance: 310 orders paid on finance; analytics credited 262 of them to the finance provider as a referral.
  • Trade-ins: 2,100 quotes, 640 accepted and 590 kits received, none of them in any report.
  • Signed-in sessions led to 31% of orders. Accounts are keyed by a customer number printed on every invoice.
  • Incident: a checkout app update stopped the analytics and Google Ads purchase tags for 15 days, with 270 orders in the gap. It was found three weeks later, after the board report showed paid search return halving and £6,000 a month had moved to Meta.
  • Deletion: a buyer's request was carried out in the email platform; her record remains in the warehouse's CRM copy and in a saved customer list file.
  • Also supplied: last quarter's board reports as sent, and a blank copy of the marketing measurement audit scorecard.

Constraints:

  • The team is the head of ecommerce, an analyst for part of each week, a developer on a fixed fortnightly release day and one remaining agency for paid search. The operations director owns data protection.
  • The operations director decides each lawful basis and retention period; where none is decided yet, the plan records an open decision rather than choosing one.
  • No new tool arrives without the onboarding review. Describe each expected effect as a direction and the check that will confirm it; put no invented figure on it.

Task

Write one plan with these sections, in this order.

  1. One collection layer. Write the Data Layer event contract and its mapping to analytics, Google Ads and Meta, with event identifiers taken from the order number, consent checks on every tag, the tests and the tags removed in the same release, following the tutorial on ad platform tags from one data layer. Gate the release with the container publish checklist and the tracking verification checklist.
  2. Trade-ins, finance and identity. Decide where accepted trade-ins and finance sales are recorded in code you control, how the source survives each hop and which system confirms outcomes the browser never sees, from the guide to embedded widgets and hosted pages. Then make the User-ID decision with the guide to User-ID and signed-in users: whether it is worth sending, which value given that the customer number is printed on invoices, when it is set and cleared, its consent rule and the CRM join it enables.
  3. Consent and retention. Build the consent map by purpose with the tutorial on mapping consent across the stack, including the two tools that bypass the CMP, and trace a withdrawal with a marked test record. Write the retention schedule by purpose and a deletion design that reaches the warehouse and saved files, following the guide to data retention across marketing tools.
  4. The audit and its fix list. Score last quarter with the marketing measurement audit, with evidence for each row and the gates first, then rank the findings into a fix list with owners, dates and a rescore date. Say which findings belong to the analytics implementation audit instead.
  5. Outages. Adopt the tracking outage response SOP: the routine check against orders (weekly in the measurement governance playbook's calendar) plus a check after every release, which together would have caught last quarter's gap within days; roles and deputies; who may raise and who declares; annotation and the stakeholder updates.
  6. Corrections. Write the rule for a wrong figure with the tutorial on correcting a number after a report has gone out, and apply it to last quarter's board report as its first case. Every correction gets a Change Log entry in the measurement change log.
  7. Calendar and owners. Lay out the year's checks per change, weekly, monthly, quarterly and yearly, drawing on the measurement governance playbook and the privacy and consent operations playbook, with roles written as a RACI matrix. Each audit, deletion run and correction leaves an Audit Trail.
  8. Summary. One page for the board: what changes, in what order, who owns it and how each change is checked.

Evaluation criteria

CriterionWhat a strong answer showsCommon gap
One collection layerWrites one event contract mapped to analytics and each ad platform with shared event IDs, consent checks and testsEach platform's tag left on its own trigger
Embedded tools and identityRecords trade-ins and finance outcomes where the code is controlled, keeps the source across the hop and decides the User-ID value and its consent rulesFinance sales credited to the finance provider as a referral
Consent and retentionMaps every purpose from capture to each reading tool, traces withdrawals, sets retention by purpose and designs deletion that reaches every copyA banner check with no map of the tools behind it
AuditScores the system on the audit's criteria with evidence, ranks findings into a fix list with owners and dates and repeats on a cadenceA score with no fix list
OutagesAdopts the outage procedure with detection checks, roles, annotation and stakeholder updatesOutages found when a report looks odd
CorrectionsSets how wrong figures are sized, fixed at source, restated and communicated, with a change logQuiet fixes to the dashboard

Most of the marks in a strong plan come from sequencing. One count of orders from the shop platform comes first, because the audit's gate, the outage check and every correction are measured against it. The consent rows for the session recording script and the trade-in frame are decided before the new contract ships, so the release does not carry old consent gaps into new tags. Last quarter's board report is corrected openly in the first month, so the board sees the rule work. The User-ID decision is judged on its reasoning: a strong answer weighs the CRM join against a widely known customer number and may choose a separate internal key.

Check your arithmetic against these figures once you have done it yourself.

Illustrative example — not a real result

Reference figures for self-checking

  • Platform claims: 1,210 + 980 = 2,190 purchases against 1,650 orders, about 1.33 claimed per real order.
  • Finance sales credited to the provider: 262 ÷ 310 ≈ 85%.
  • Trade-ins: 640 ÷ 2,100 ≈ 30% of quotes accepted; 590 ÷ 640 ≈ 92% of accepted kits received.
  • The outage gap: 270 orders over 15 days, about 18 a day missing from analytics and Google Ads.

Submission checklist

0/9

Once the plan is written, set your collection layer beside the cookery school data layer exercise, your consent and retention sections beside the wine club exercise and your outage and correction rules beside the stationery brand exercise.

Pages this lesson applies

Templates and checklists used

Terms used on this page

Also related

Related