Vilma Caviar by Majestic · Cosmetics · Meta Ads · Creative testing
Vilma Caviar needed a new creative direction. The test produced 387 attributed purchases and a 4.38x reported ROAS.
Vilma Caviar by Majestic entered the engagement under pressure. The previous advertising direction was unreliable, some creatives barely received delivery, and the brand did not have a clear answer about which products or formats deserved the next budget. More creative volume would not have solved that. The brand needed a verdict.
OJC Labs rebuilt the test around products, formats and purchase outcomes. Static and video variants competed across multiple product families. Weak delivery was identified early. Budget moved toward combinations that could produce purchases at a sustainable cost, while the results established a clearer direction for the brand's next creative cycle.
This is the role of a connected Meta Ads and conversion-tracking system: give the business evidence about where the next unit of budget should go, not merely a dashboard full of activity.
$2,811.42 spend · 387 attributed purchases · $7.26 per purchase · 4.38x reported ROAS
The constraint
The account did not need more ads. It needed to know which ads deserved to exist.
The creative mix covered acne care, intimate-area brightening, beach kits, hair removal, stretch-mark care and a premium Gold Pack. These products had different audiences, buying motivations and levels of urgency. The account also contained variants that spent only a few dollars and generated no purchase, alongside ads that could produce dozens of conversions.
Without a disciplined test structure, the brand risked treating every weak result as an audience problem and every strong result as proof that the whole creative direction worked. OJC Labs needed to isolate the product-format combinations that generated purchases, not merely delivery.
For ecommerce teams, that distinction connects creative production to the wider ecommerce operating system, where product pages, purchase events and campaign decisions need to tell the same story.
What OJC Labs changed
We turned the Meta Ads account into a decision system.
The new structure compared static and video formats within product families, then read performance at three levels:
- Can the creative earn enough delivery to produce a usable signal?
- Can it generate purchases at a sustainable acquisition cost?
- Does the attributed purchase value justify further budget?
Variants with weak delivery or no purchase signal were contained. Stronger product-format combinations were allowed to accumulate enough volume to become useful evidence. The goal was not to make every creative look successful. The goal was to discover what the brand should do next.
This testing loop is part of OJC Labs Growth Engineering: measurement, controlled experimentation and budget decisions working as one commercial system.
Test architecture
- 01Separate campaigns by product family.
- 02Compare static and video formats.
- 03Optimize toward the purchase event.
- 04Monitor cost per attributed purchase and reported ROAS.
- 05Move budget toward repeatable conversion signals.
- 06Contain low-delivery and zero-purchase variants early.
- 07Read product-level evidence instead of hiding behind the account average.
Format evidence
Video created the most purchase volume. Static creative remained highly competitive on efficiency.
Video ads generated 219 attributed purchases from $1,495.22 in spend, producing a blended cost per purchase of $6.83 and a reported ROAS of 4.20x. Static ads generated 134 purchases from $893.79, producing a slightly lower blended cost per purchase of $6.67 and a reported ROAS of 5.48x.
The static ROAS includes one unusually high result from the Anti-Vergetures poster. That single ad reported 40.55x ROAS and materially lifts the static and account averages. We therefore preserve the platform-reported 4.38x blended account ROAS but also run a conservative sensitivity check: removing that outlier leaves the account at 3.74x reported ROAS across 376 purchases. The conclusion remains positive without requiring the suspiciously heroic number to carry the story.
Format comparison
| Format | Spend | Attributed purchases | Cost per purchase | Reported ROAS |
|---|---|---|---|---|
| Video | $1,495.22 | 219 | $6.83 | 4.20x |
| Static | $893.79 | 134 | $6.67 | 5.48x* |
| Unspecified format | $422.41 | 34 | $12.42 | 2.72x |
*Static ROAS includes the Anti-Vergetures poster outlier described above.
Product evidence
Not every product needed the same creative format, and the account finally had the numbers to prove it.
The beach-kit Video 3 creative generated the largest single purchase volume: 104 attributed purchases from $782.11 in spend with 4.41x reported ROAS. It showed that video could carry scale even when its cost per purchase was not the lowest in the account.
The intimate-area brightening product produced one of the cleanest efficiency signals. Its poster and video generated 58 attributed purchases from $276.80, with a blended $4.77 cost per purchase and 5.00x reported ROAS.
Hair-removal campaigns generated 105 attributed purchases from $552.25 at $5.26 per purchase and 4.14x reported ROAS. Within that family, the men's video reached 21 purchases at $3.30 per attributed purchase and 6.68x reported ROAS, making it one of the strongest combinations of efficiency and return.
Anti-acne creative was materially weaker, generating 42 attributed purchases from $564.92 at $13.45 per purchase and 2.37x reported ROAS. That did not mean the product had no market. It meant the tested direction was not earning the same budget priority as the stronger product families.
Product-family comparison
| Product family | Spend | Attributed purchases | Cost per purchase | Reported ROAS |
|---|---|---|---|---|
| Intimate-area brightening | $276.80 | 58 | $4.77 | 5.00x |
| Hair removal | $552.25 | 105 | $5.26 | 4.14x |
| Beach kit | $916.87 | 115 | $7.97 | 4.17x |
| Gold Pack | $323.28 | 43 | $7.52 | 3.69x |
| Stretch-mark care | $177.30 | 24 | $7.39 | 12.92x* |
| Anti-acne | $564.92 | 42 | $13.45 | 2.37x |
*The stretch-mark family contains the 40.55x poster outlier and must not be used as a forecast.
Validated results
The creative-testing system recovered both purchase volume and strategic direction.
From 10 July to 16 August 2026, Vilma Caviar's campaigns spent $2,811.42 and delivered 1,242,963 impressions to 419,157 people at a frequency of 2.97. Meta attributed 387 purchases to the included ads. The blended cost per attributed purchase was $7.26, and the platform-reported purchase value implies approximately $12,316.59 in attributed revenue and 4.38x blended ROAS.
More importantly, the test produced a usable creative hierarchy. Video could carry volume. Static creative could compete on efficiency. Intimate-area brightening and hair-removal products deserved continued attention. Anti-acne required a different direction. Low-delivery variants could be removed without a committee meeting lasting until the next fiscal year.
Verified result summary
- $2,811.42total spend
- 1,242,963impressions
- 419,157people reached
- 2.97average frequency
- $2.26CPM
- 387Meta-attributed purchases
- $7.26blended cost per attributed purchase
- Approximately $12,316.59in platform-attributed purchase value
- 4.38xplatform-reported blended ROAS
- 3.74xreported ROAS when the largest outlier is excluded
Measurement note
Purchase counts, purchase value and ROAS are Meta-attributed results from the supplied Ads Manager export. The rows use more than one attribution setting, including 7-day click, 1-day view and 1-day engaged view combinations. The 387 figure comes from the export's Purchases column; the optimization Results column totals 364 because Meta reports these fields differently across attribution configurations. This page uses the explicit purchase column and states the distinction.
The export does not contain refunds, product margins, fulfilment costs or independently reconciled backend revenue. Reported ROAS is therefore an advertising-platform attribution metric, not audited profit.
This is why OJC Labs connects campaign optimisation to growth systems and reliable conversion tracking before treating a platform dashboard as the company's financial ledger.
The takeaway
Creative testing is a system for eliminating weak directions and funding strong evidence.
Creative testing is not content production with a spreadsheet attached. It is a system for killing weak directions early and giving strong ones enough budget to tell the truth.
Vilma Caviar did not leave the test with one universal winning format. It left with something more useful: a product-format map, a conservative performance baseline, and evidence about where the next creative cycle should go.
Frequently asked questions
What marketing and ecommerce teams ask about Meta Ads creative testing
A useful creative test has to explain what the platform measured, what the business can safely conclude and what should receive the next budget.
Why report both 4.38x and 3.74x ROAS?
The platform reports 4.38x across the complete export. One ad reported an unusually high 40.55x result. Excluding that ad produces a more conservative 3.74x sensitivity case, showing that the overall conclusion remains positive without depending on the outlier.
Were videos or static ads better?
Video produced more attributed purchase volume, while static ads produced a slightly lower blended cost per purchase. The correct conclusion is not that one format always wins. Different product families responded to different formats.
Does 387 purchases mean 387 independently verified orders?
It means Meta attributed 387 purchase events in the export. Backend order reconciliation was not included, so the page labels these as platform-attributed purchases rather than audited fulfilled orders.