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

Meta Ads › Intermediate

Capstone: Auditing and Improving a Meta Ads Account

The closing project of the intermediate level. You audit an invented bedding brand's Meta Ads account with the scorecard and checklist, establish what its reported results are worth, and write a prioritised plan covering measurement, structure, audiences, creative and the owner's wish to scale.
Capstone projectIntermediate levelModule 3: Measurement, scaling and audits7 minUpdated

Brief

Thistledown Bedding is an invented online brand selling duvets, pillows and bed linen. Its busiest months are the colder ones, and the owner wants to raise Meta spend substantially before they start. Meta is the brand's main paid channel. The account was built by a freelancer who has since moved on, and nobody has reviewed it since.

The owner's questions: are the results as good as the reports say, what should change first, and is it safe to spend more this season?

Constraints:

  • One marketer runs the account part-time; a developer can give a few days this month.
  • The designer can produce a small number of new videos a month, and the brand holds written licences for a set of customer review videos.
  • Spend may move between campaigns now; any increase in the total waits for your recommendation.

The account and its data

Everything below comes from Ads Manager, Events Manager, the shop platform and the owner, over a closed period.

Illustrative example — not a real result

Thistledown Bedding: economics and measurement

Contribution margin after product, delivery, payment fees and returns: 45%. The owner wants 10 percentage points of revenue kept as profit after ad cost. Average order value: £90. Current spend: £400 a day. Proposed: £600 a day.

Last 28 days:

  • Shop platform: 720 online orders from all channels, £64,800 revenue.
  • Events Manager: 1,410 purchase events received, from both the browser pixel and a partner Conversions API integration. The pixel purchase tag sends no event ID; the deduplication share shown for purchases is close to zero.
  • Attribution: Prospecting – Advantage+ uses 7-day click and 1-day view; the other campaigns use 1-day click. The owner's monthly report adds them together.
Illustrative example — not a real result

Thistledown Bedding: campaigns over the last 28 days (Meta-reported)

CampaignSet-upBudgetSpendPurchasesRevenueROAS7-day frequency
Prospecting – Interests6 interest ad sets, purchase-optimised, 5 learning limited£25 a day per ad set£4,200280£25,2006.01.8
Prospecting – Advantage+1 ad set, Advantage+ audience, no exclusionsCampaign budget £120 a day£3,360390£35,10010.42.9
Retargeting – VisitorsSite visitors, 180 days, no purchaser exclusion£100 a day£2,800470£42,30015.19.4
Awareness – ReachReach objective, frequency control set£30 a day£84050£4,5005.41.9
Total£400 a day£11,2001,190£107,1009.6

Further evidence:

  • The overlap tool shows heavy overlap between the six interest ad sets, and between them and the Advantage+ ad set.
  • The customer list used for lookalikes was last uploaded nine months ago; no campaign excludes customers.
  • The account runs four ads. Both prospecting campaigns use the same two videos; one has carried about 60% of prospecting spend for ten weeks, and its click-through rate has fallen from 1.4% to 0.8% against its own first fortnight.
  • Change history: interest ad set budgets edited most weeks; no test records anywhere.
  • The ad account sits in the former freelancer's business portfolio; the brand has partner access only. There is no naming convention.

Task

Produce one document with the sections below, in this order. Use the site's resources as named, state expected effects as direction with a way to check them, and do not invent uplifts.

  1. Gather the evidence. Work through the Meta Ads account audit checklist group by group, recording each failed line with its evidence and the spend it touches. Record anything you cannot check as missing evidence, not as a pass.
  2. Score the account. Rate every row of the Meta Ads account audit scorecard as missing, partial or solid, gates first, and give the band. Say which rows are provisional because of the gate result.
  3. Write the measurement verdict. Explain what the missing event ID does to purchase counts and to delivery, using the deduplication section of the Conversions API guide. Give the fix, how it will be verified in Events Manager and against shop orders, and what happens to the attribution settings and the owner's report.
  4. Set the economics. Calculate the break-even and target ROAS with the break-even calculator, and explain why no campaign can be judged against them until the measurement fix has run for a full window.
  5. Fill the plan lines. Turn each partial or missing row into a line: finding, evidence, fix, owner, rescore date. Rank them by spend touched and speed of effect, as the scorecard's prioritisation section describes.
  6. Propose the structure and budgets. Say what to merge, keep or stop, which budget type each campaign uses and why, applying the learning phase guide and the budget types comparison. The ad set consolidation exercise practised this step.
  7. Fix audiences and exclusions. List the exclusions and refreshes the account needs, using the audience exclusions and overlap guide, and resize retargeting to its pool.
  8. Plan creative. Diagnose the tired video as wear-out or saturation with the frequency and creative fatigue guide, then write one test in the experiment hypothesis template, as in the creative test exercise.
  9. Answer the scaling question. Say whether and when the budget rise should happen, and how, using the stopping rule and checklist in the scaling campaigns guide.
  10. Sequence the first six weeks. Lay out week by week what changes and what is checked before the next change, with a weekly review rhythm from the paid social account management playbook.

Evaluation criteria

CriterionWhat a strong answer showsCommon gap
Gates firstThe server events row scored missing, the band red, and structure and budget rows marked provisionalScoring every row equally and calling the account amber
Measurement fixShared event ID on both sides, verified by deduplication share and server purchases against shop orders, one attribution settingSwitching the server integration off and losing its coverage
EconomicsBreak-even and target from margin, with no campaign judged on reported return before the fixPraising retargeting for its reported return
ExclusionsCustomers out of prospecting, recent purchasers out of retargeting, list refreshed and kept currentRefreshing the list but not applying it
StructureInterest ad sets merged into a structure that can exit learning; retargeting on its own budget, sized to the poolKeeping six ad sets to learn which interest wins
CreativeWear-out diagnosed from steady reach and a falling click-through rate; new concepts added alongside, one logged testReplacing every ad at once
ScalingRise held until clean data, then taken in moderate steps judged on marginal returnApproving the rise because reported return is high
HygieneAccount ownership moved to the brand's portfolio, scheduled earlyLeaving ownership as a footnote
SequenceOne area changed at a time, with checks between changesEvery fix in the first week
CommunicationA short summary the owner can read quickly, with the band, the top lines and direction rather than promised figuresInvented percentage uplifts

To check your arithmetic, the key figures are below. Work them out yourself first.

Illustrative example — not a real result

Reference figures for self-checking

  • Break-even ROAS: 1 ÷ 0.45 ≈ 2.22. Target ROAS: 1 ÷ (0.45 − 0.10) ≈ 2.86.
  • Purchase events per shop order: 1,410 ÷ 720 ≈ 1.96, consistent with almost every order counted twice.
  • Meta-reported revenue (£107,100) exceeds the shop's revenue from all channels (£64,800).
  • Blended ceiling: even if every shop order came from Meta, revenue ÷ Meta spend = £64,800 ÷ £11,200 ≈ 5.8, far below the reported 9.6.
  • Retargeting: 180-day window, no purchaser exclusion, frequency 9.4: saturation, not a star performer.
  • Proposed rise: £600 ÷ £400 = 1.5 times current spend, in one step.

A strong plan treats the reported results as unknown rather than merely inflated, because a broken purchase count also misleads delivery about who buys. It fixes deduplication before anything else, moves ownership of the account early, and uses the Conversions API it already has instead of removing it. It reads the Frequency on retargeting as an audience problem and on prospecting as a creative one.

Submission checklist

0/11

When you have finished, compare your measurement and scaling sections with the scaling decision exercise for a sailing clothing brand, which practises the same judgement on a single campaign, and return to the intermediate Meta Ads level to mark the level complete.

Pages this lesson applies

Terms used on this page

Also related

Related