Eid Campaign Meta Ads ROAS in Pakistan

The Client

A Lahore-based D2C kidswear brand selling on its own Shopify storefront and shipping nationwide on cash-on-delivery terms through two courier partners. The label ran two distinct lines: everyday playwear priced PKR 1,200-2,800, and festive Eid assortments — three-piece suits, kurta sets, and coordinated family bundles — priced PKR 3,500-6,500. Roughly two-thirds of annual revenue landed in three windows: Eid-ul-Fitr, Eid-ul-Adha, and the winter wedding season that followed.

The company was small and commercially disciplined — around twenty people, an in-house designer, and one performance marketer who also managed the marketplace listings. Meta owned about 80% of the paid mix, with in-season spend running PKR 1.6M-2.4M per month during peaks. That is a mid-size Pakistani advertiser: large enough that structural waste compounds quickly, small enough that nobody had time to audit it.

The trigger for the engagement was the 2025 season. The two Eid launches blended at 1.8x return on ad spend — below the brand’s 2.2x breakeven once cost of goods, delivery, failed deliveries, and COD returns were counted — and the winter wedding capsule that followed came in at 1.6x. The brand was paying to acquire orders it then lost money on. They engaged WeProms Digital in January, ahead of Ramadan, for a 22-week engagement spanning both 2026 Eid launches. This is an illustrative composite built from the patterns we see across clothing brands in Pakistan; the metrics are realistic outcome ranges a buyer can use to sanity-check fit, not audited third-party figures.

The Problem

The first two weeks were spent inside the ad account, the pixel, and the courier reconciliation sheets. Six blockers explained most of the loss:

• Eleven overlapping prospecting ad sets. Interest, lookalike, and broad sets ran side by side with 60-70% estimated audience overlap. The brand’s own ads were its closest competitors in the auction, and the algorithm fragmented learning across sets that each saw too little conversion data to stabilize.
• Six creatives, none seasonal. The oldest had been live for fourteen weeks. The Eid collection message — new cuts, family bundles, delivery-before-Eid — was being carried by generic product studio shots from the everyday line.
• Optimization on placed orders, not kept orders. The purchase event fired at order placement. With 22% of festive COD parcels returned uncollected, the account was optimizing toward audiences that ordered enthusiastically and refused delivery just as readily. Reported ROAS overstated the true, return-adjusted number by roughly a quarter.
• Spend spiked exactly when CPMs peaked. Off-season budgets ran flat, then quadrupled for launch week — the same week CPMs ran roughly 1.8x off-season rates as every apparel advertiser in the country fought for the same feed positions. Campaigns entered learning phase during the most expensive window of the year.
• Retargeting was dark. Dynamic catalog ads had been dormant for eight months after a pixel reroute during a theme update; nobody noticed because reported conversions kept flowing from prospecting.
• No forecast. Budget decisions were made the night before, on instinct, with no model linking expected CPM curves to spend and margin.

None of these were exotic. Every one of them is present in a large share of Pakistani D2C ad accounts we audit. Compounded across a PKR 2M month at 1.8x ROAS, they were the difference between a loss-making season and a profitable one.

Two quieter issues sat underneath all six. The brand had no documented definition of a launch — warmup, peak, and harvest were not distinguished anywhere, so every window was managed as one undifferentiated sprint that started late and ended abruptly. And nobody owned the number: reported ROAS came from the ads manager, return adjustments reached finance two weeks later, and the two were never reconciled in the same room. Fixing accountability for the metric preceded fixing the metric itself.

Phase 1 — Diagnosis and Account Cleanup (Weeks 1-2)

The cleanup phase deliberately touched no budgets. It established ground truth first.

The conversion signal audit found the purchase event double-firing — the browser pixel and an old Conversions API implementation were both sending purchases with no event ID deduplication, so Meta credit was inflated against courier reality. Deduplication was rebuilt so browser and server events shared event IDs, and a Confirmed Purchase custom conversion was defined: an order only counted once the courier marked it delivered and the seven-day return window closed. The performance marketer reconciled this weekly against courier manifests — a two-hour job that became the single most valuable dataset in the account, because for the first time the brand could see return-adjusted ROAS by creative, by campaign, and by city.

Structurally, seven of the eleven prospecting ad sets were suspended. The remaining four were consolidated into a single broad campaign pending the Phase 2 rebuild, which immediately reduced internal auction pressure — CPMs on the surviving delivery dropped roughly 9% within ten days at unchanged spend.

Diagnostic finding	Action taken	Effect within phase
11 overlapping prospecting ad sets	Suspended 7, consolidated 4 into one campaign	Internal auction pressure down; CPM -9% at flat spend
Purchase event double-firing (pixel + CAPI, no dedup)	Rebuilt event ID deduplication	Meta credit aligned with courier reality
ROAS reported on placed orders	Defined COD-confirmed purchase conversion, reconciled weekly	True festive ROAS visible: ~1.4x, not 1.8x
Dynamic catalog retargeting dormant 8 months	Re-enabled against corrected feed	Warm audience of ~180K recent site visitors available again
No baseline reporting	Return-adjusted dashboard by campaign, creative, city	Daily decision-making on kept revenue

The most important output of Phase 1 was not a change — it was a number. The brand learned its real 2025 blended ROAS had been closer to 1.4x than the 1.8x the dashboard claimed. Breakeven was further away than leadership thought, which reset the ambition for the season.

Budget guardrails were also set in this phase, before the season started: a maximum daily spend per campaign, a floor on return-adjusted ROAS below which spend stepped down automatically (the 2.2x breakeven plus a 0.3x buffer), and a standing rule that no budget moved without a logged reason. None of this is sophisticated — it is simply the difference between an account that is operated and an account that is watched.

Phase 2 — Signal, Structure, and Creative Rebuild (Weeks 3-6)

With clean signal flowing, the account was rebuilt into three campaigns with a 70/20/10 launch budget split:

1. One broad prospecting campaign (70%) using Advantage+ audience settings with no interest stacking, carrying the full creative set under a cost-cap bidding test that later reverted to highest-volume where caps choked delivery.
2. One warm retargeting campaign (20%) across 14-day site visitors, video viewers, and Instagram engagers.
3. One dynamic catalog campaign (10%) merchandising festive SKUs to recent viewers, which later carried the post-season harvest.

The creative system was the larger rebuild. Fourteen creatives were produced across four archetypes, each mapped to a parental buying tension we see consistently in Pakistani kidswear: wanting the family to look coordinated for Eid, dressing more than one child at once, fearing poor fabric after years of marketplace disappointments, and the hard delivery cut-off before Eid. This is where most of the ROAS came from — creative, not targeting, carries Pakistani Meta accounts, and the account had been starved of it.

Creative archetype	Format	Share of prospecting spend	Relative CPA (1.0 = best)
Matching-family moments	UGC-style video	35%	1.0
Sibling bundles, two-child value framing	Carousel	25%	1.15
Fabric durability close-ups and wash tests	Static	20%	1.3
Order-by-date delivery cut-off reminders	Static	20%	1.25

Two rules governed the system. First, a 72-hour kill rule: any creative whose confirmed-purchase CPA exceeded 1.4x target for three consecutive days was paused without discussion — creative decisions were delegated to the data, not to internal preference. Second, a pre-season test window: roughly 15% of the planned launch budget was spent in the two weeks before Ramadan advertising began in earnest, so winners and losers were already separated before CPMs inflated. Five creatives entered launch week as proven winners; under the old model, all fourteen would have been learning live during the peak.

Each archetype also carried a defined release order: the matching-family video always shipped first because it set the emotional frame, value-framing carousels followed once audiences warmed, and the cut-off statics were held back for the final fortnight when deadline pressure peaked. Sequencing creative to the emotional arc of the season — anticipation, comparison, urgency — cost nothing and consistently beat launching everything at once.

Throughout, the engagement ran on our standard Meta ads management operating cadence — weekly creative reviews, daily budget checks during season, and a documented change log so every performance move could be traced to a decision.

Phase 3 — Two Eid Launch Windows (Weeks 7-18)

Warmup discipline. Thirty percent of each launch budget was committed to the two weeks before peak — not to drive volume, but to get the consolidated campaign out of learning while CPMs still sat near off-season rates. This inverted the brand’s historical pattern of spiking spend into the single most expensive week. By the time Eid CPM inflation arrived, delivery was stable and the algorithm was optimizing to confirmed purchases rather than placed orders. The effect was mechanical: launch-week CPMs fell 21-23% versus the prior year while in-season spend rose 31%.

Eid-ul-Fitr (launch weeks 7-10). The first launch blended at 2.9x return-adjusted ROAS — above breakeven for the first time in three seasons. The matching-family UGC creative carried 38% of conversions at the account’s best CPA; the delivery cut-off statics overdelivered in the final ten days as deadline pressure built. Budget scaled in +20% steps every 48 hours whenever the three-day rolling ROAS held above 2.6x, a rule written down in advance so scaling was never an in-the-moment decision.

Between windows (weeks 11-15). Spend dropped to maintenance levels. The team refreshed the creative wave — same archetypes, new footage and stills from the Eid-ul-Adha line — and ran a bundle test that lifted average order value from PKR 3,940 to PKR 4,380 (+11%) by merchandising two-child sets at a single-shipment price. Returns were attacked directly: sizing charts moved above the fold on festive product pages, fabric-composition videos were added to ads, and the confirmation-call script was tightened to restate delivery timelines. Festive COD returns fell from 22% at Fitr to 15% by the second launch.

Geography was rebalanced at the same time. Courier data showed confirmed-purchase rates in the major metros ran meaningfully ahead of smaller towns, where refusal rates ran hottest, so delivery promises in ad copy were tied to postcode-level serviceability rather than promised nationally. Spend share to the top three metros rose from 48% to 61%, while the excluded areas’ share of returned orders fell by roughly half — a quiet margin gain visible only in the reconciled data.

Eid-ul-Adha (launch weeks 16-18). The second launch blended at 4.0x with the same structure, same rules, and a proven creative bench — the clearest evidence in the engagement that the system, not a lucky ad, was doing the work.

Metric	Eid-ul-Fitr 2025	Eid-ul-Fitr 2026	Eid-ul-Adha 2025	Eid-ul-Adha 2026
Blended ROAS (return-adjusted)	1.9x	2.9x	1.7x	4.0x
Launch-week CPM	PKR 328	PKR 258	PKR 296	PKR 228
Cost per confirmed purchase	PKR 1,790	PKR 1,340	PKR 1,890	PKR 1,210
In-season spend	PKR 1.9M	PKR 2.4M	PKR 1.5M	PKR 2.1M

Blended across both launches and weighted by spend, the engagement delivered 3.6x against the prior year’s 1.8x — the doubling that the whole 22-week system was built to produce.

Phase 4 — Post-Season Harvest and Compounding (Weeks 18-22)

The weeks after Eid-ul-Adha were treated as a fourth revenue window rather than a shutdown. The dynamic catalog campaign, now fed by a corrected product feed, merchandised residual festive stock to recent viewers at 8% of total spend and returned 4.8x — clearing inventory the brand had previously written down. A win-back sequence offered last year’s Eid buyers early access to the winter capsule, seeding that launch’s warm audiences at negligible cost — a pool of roughly 22,000 past-season buyers that converted at 3.4x when the capsule opened, warm traffic the brand had previously left untouched every year.

The compounding asset, though, was documentation. The engagement ended with a written launch playbook: a season calendar with warmup start dates, creative briefs for each archetype, the kill and ramp rules, and a simple forecast linking expected CPM curves to budget and margin. The winter wedding capsule that followed ran on the same system without our involvement and held at 3.1x blended on roughly 40% higher spend than the prior winter — which is the actual point of a framework, and the standard we hold this kind of work to.

Final Results

Metric	2025 Eids (blended baseline)	2026 engagement (blended)	Change
Blended ROAS (return-adjusted)	1.8x	3.6x	+100%
Launch-week CPM	PKR 310	PKR 242	-22%
Cost per confirmed purchase	PKR 1,840	PKR 1,270	-31%
COD return rate (festive)	22%	15%	-7 pts
Average order value	PKR 3,940	PKR 4,380	+11%
In-season monthly spend	PKR 1.7M	PKR 2.2M	+31%

Every row traces to a phase: the CPM and spend rows to consolidation and warmup discipline in Phases 1-3, the confirmed-purchase cost to the signal rebuild in Phase 1 and the creative system in Phase 2, the return rate to the expectation-setting work between launches, and the order value to the sibling-bundle merchandising test. Spend rose while returns improved, which is the only version of scaling worth attempting.

What Made This Work

1. The account optimized to kept revenue, not ordered revenue. Defining the conversion as a courier-confirmed, return-window-closed purchase changed every downstream decision — which audiences looked profitable, which creatives earned budget, and what the true breakeven actually was.
2. Warmup preceded the peak. Moving 30% of launch budget into the pre-peak fortnight bought stable delivery at off-season CPMs, so the brand stopped paying peak prices to train the algorithm.
3. Creative volume with enforced kill discipline. Fourteen creatives across four archetypes gave the auction genuine choices; the 72-hour kill rule prevented sunk-cost attachment to beautiful ads that did not confirm purchases.
4. Consolidation cut auction overlap. One broad campaign buying reach is structurally cheaper than eleven narrow campaigns bidding against each other for the same Pakistani shopper.
5. Returns were treated as a pre-purchase problem. Sizing charts, fabric transparency, and honest delivery windows in the ad itself reduced refusals at the door — protecting the return-adjusted ROAS that reported ROAS had been hiding.

What Teams Can Apply

1. Compute your return-adjusted breakeven ROAS before the season starts. If 20%+ of parcels come back, your breakeven is meaningfully higher than your dashboard suggests. Optimize to confirmed orders or you will scale a leak.
2. Consolidate before you scale. Audit audience overlap and suspend anything that competes with itself. In most Pakistani accounts we review, this alone moves CPMs before a single new creative ships.
3. Pre-test creative at small spend so winners carry launch budget. Fifteen percent of the launch budget spent two weeks early is the cheapest performance insurance available; learning during peak CPMs is the most expensive.
4. Ramp with written rules, not adrenaline. A +20% step every 48 hours when rolling ROAS holds above threshold keeps scaling boring and reversible.
5. Make COD confirmation part of measurement, not an afterthought. A weekly courier reconciliation turns returned orders from a finance surprise into a targeting and creative signal.