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

Meta Ads › Intermediate

Deciding When and How to Scale a Meta Campaign

You read an invented sailing clothing brand's results for its strongest Meta campaign, including a recent budget step and notes on how well its purchases are measured. The exercise practises judging readiness on marginal rather than average return, choosing between vertical and horizontal scaling, and fixing measurement before trusting the numbers that drive the decision.
ExerciseIntermediate levelModule 3: Measurement, scaling and audits5 minUpdated

Why this matters

The request to scale usually arrives with an average attached. The scaling campaigns guide explains why the average is the wrong number, and the Conversions API guide explains why a pixel-only account sees fewer purchases than happened. Both matter here: the decision rests on a marginal figure computed from purchases the account may be under-counting.

The results and signal notes

Halyard Sailwear is an invented brand selling waterproof sailing jackets, salopettes and knitwear online. Its owner wants to raise spend on the campaign that has done best this season.

Illustrative example — not a real result

Halyard Sailwear: economics and campaign results

Contribution margin after product, delivery, payment fees and returns: 40%. The owner wants 10 percentage points of revenue kept as profit after ad cost. Average order value: £150.

Prospecting – Waterproofs: Advantage+ campaign budget, two ad sets (Advantage+ audience, purchaser lookalike), both out of learning. The daily budget moved from £300 to £400 at the start of week 3.

PeriodDaily budgetDaily revenue (Meta-reported)ROASPurchases a dayCost per purchase
Weeks 1–2£300£1,2004.08.0£37.50
Weeks 3–4£400£1,5203.810.1£39.60

Signal and delivery notes:

  • Measurement is pixel only. Events Manager received 430 browser purchase events in the 28 days; the shop recorded 560 online orders in the same period.
  • The pixel purchase tag sends no event ID.
  • The shop platform offers a maintained partner integration for the Conversions API.
  • Prospecting 7-day frequency rose from 1.5 to 2.3 over the four weeks.
  • One video, a rain-test film, carries about 70% of the campaign's spend; its click-through rate has slid from 1.6% to 1.2% against its own first fortnight.
  • No new creative has been added since the season started.

Reading readiness to scale

Three questions decide readiness, in order. Is the measurement complete enough to trust? Did the last increment pay at the margin, not just on average? Is there room in the audience and creative for more spend? Use the break-even calculator for the floor, and compare the marginal figure with both the floor and the profit-bearing target. A ROAS that still looks healthy on average can hide a last step that barely paid. Then read frequency and the share of spend on one ad: if those are climbing, the next step is horizontal, through new creative, rather than vertical. The Meta Ads account audit scorecard treats server events and deduplication as a gate for the same reason this exercise does: every later judgement rests on them.

Task

Answer before reading the worked answer.

  1. Calculate the break-even and target ROAS from the margin.
  2. Calculate the marginal ROAS and the marginal cost per purchase of the last budget step, and compare them with the targets.
  3. Estimate how much of the shop's purchasing the pixel is missing, and say what that does and does not tell you about the marginal figure.
  4. Decide whether the campaign is ready for a vertical budget increase now, with your reasons.
  5. Write the order of work for the next month: what to fix, what to launch, what to hold and when to decide again.
  6. Name the measurement step that must be recorded so the later decision is not misread.

Worked answer

The arithmetic first.

Illustrative example — not a real result

Halyard's targets and marginal return

  • Break-even ROAS: 1 ÷ 0.40 = 2.5. Target ROAS: 1 ÷ (0.40 − 0.10) ≈ 3.33.
  • Marginal ROAS of the last step: extra revenue £1,520 − £1,200 = £320 a day for extra spend £100 a day, so 3.2.
  • Marginal cost per purchase: £100 ÷ (10.1 − 8.0) ≈ £47.60, against an average of £39.60.
  • Pixel coverage: 430 ÷ 560 ≈ 77%, so roughly a quarter of orders are invisible to Meta.
  • If the missing orders were spread evenly and all recovered, the marginal ROAS would be about 3.2 ÷ 0.77 ≈ 4.2. Neither assumption is safe.

Readiness. Not ready for a vertical step yet. The average return still beats the target, but the last step returned below target and above break-even: profitable, though under the owner's standard. A further step would very likely do worse. The measurement gap cuts the other way, since the true marginal return is probably higher than reported, but by how much is unknown. Adjusting the figure by hand would be a guess; the fix is to measure.

Order of work.

  • Hold the budget. The last step still cleared break-even, and cutting back would add an edit while measurement changes.
  • Connect the Conversions API through the shop's partner integration and add an event ID to the pixel purchase tag so both sides share it. Verify in Events Manager: test events first, then the deduplication share, then server purchases against shop orders for a full week.
  • In parallel, launch two new concepts alongside the rain-test film. Frequency is climbing and one ad is wearing, so creative is the horizontal lever with the fewest side effects. Add them as new ads, not as a budget change.
  • After a full attribution window of clean, deduplicated data, recompute the average and the marginal figures from the new baseline. If the marginal return then sits comfortably above target, take a moderate step and hold it long enough to judge. If not, look for further horizontal room before more budget.

What to record. The date the server integration went live, in the change history and the owner's report. More matched purchases will raise reported return with no change in the campaign; without the note, the rise will be credited to the new creative. Judge marginal cost per purchase (CPA) on the new baseline only.

Common mistakes

  • Scaling on the average. It still beats target while the last increment did not.
  • Adjusting the numbers for missing orders by hand. The coverage gap is real, but its distribution is unknown.
  • Launching the server integration without event IDs. Purchases double and the account looks far better than it is.
  • Comparing returns across the integration launch. The measurement changed, not the campaign.
  • Raising budget onto one tiring ad. The extra spend lands on the ad that is already wearing out.

Next, bring the same reading to a whole account in the Meta Ads account audit capstone, or return to the intermediate Meta Ads level. The ROAS calculator checks any of the returns above.

Pages this lesson applies

Terms used on this page

Also related

Related