The scenario
The brand runs one online store selling three different products: fashion sunglasses, prescription frames bought against a prescription the customer brings from her own optician, and coloured contact lenses bought mostly for how they look on camera. The Shopify store takes KNET, Tabby and cash on delivery, and one Instagram and Snapchat feed had always covered all three lines with whatever was on offer.
Demand exists for all three: sunglasses spike before summer and Eid, frames sell year-round on renewal cycles, and coloured lenses move fastest around Ramadan and Eid nights when Snapchat try-on interest peaks. The problem was never traffic. It was that one campaign, one offer and one ad format were asked to sell a fashion item, a semi-medical item and a cosmetic item at the same time.
- Monthly revenue band
- pending client sign-off
- Average order value
- pending client sign-off
- Prescription handling
- Customer's own optician prescription, lenses fitted by a partner optical lab
- Team
- One founder, one part-time content creator
One campaign trying to sell a fashion item, a medical item and a cosmetic item at once
The account ran one prospecting campaign per platform, rotating ads that mixed sunglasses, frames and lenses in the same carousel or the same fifteen-second video. The logic was efficiency: one shoot, one edit, one campaign to manage. The result was an ad that was never quite for anyone — a shopper looking for coloured lenses saw a slide of frames she had no reason to click.
Every metric read as mediocre rather than broken, which made the problem hard to see without pulling the ads apart. Cost per purchase sat close to target on paper, but the split by product line — which nobody had run — showed frames losing money and lenses barely breaking even, while sunglasses quietly carried the account.
The three buyers were not confused about what the store sold. They were being shown an answer to a question they had not asked: the sunglasses buyer wants to know if it suits her face, the frames buyer wants her prescription handled by someone qualified, and the lens buyer wants to see the colour on a face like her own. One creative track was answering none of those three questions well, because it was built to answer all of them at once.
What we did — the creative angles
The losers are here on purpose. A test with only winners was never a test.
Same frame, four different face shapes
Winner"Round face, square face, long face — same sunglasses, four ways."
- Format:
- Vertical, 14 seconds, four quick cuts, one frame per face shape, frame width in millimetres on screen
Built after the split, aimed only at the sunglasses buyer's real objection: will this suit my face. Showing the same product on four face shapes, width in millimetres on screen, answers it directly instead of relying on one model. Clear winner on hook rate for the sunglasses line, and it held cost per purchase as spend increased.
You bring the prescription, the optician does the rest
Winner"Send us your prescription. Our optician partner fits the lenses, not us."
- Format:
- Vertical, 16 seconds, phone camera, the founder walking through the WhatsApp prescription-upload step
The frames track needed to solve a trust problem, not a style one. This ad never claims to correct vision — it states plainly that the store sells the frame and a partner optician fits the lens. That distinction moved frames from the account's weakest line to its second best, and cut a category of WhatsApp questions the founder answered by hand.
The colour on a face like yours, before you order
Winner"Grey, honey, green — see the colour on three different eye tones first."
- Format:
- Vertical, 12 seconds, close-up eye shots, three eye tones, same lens colour, shot for Snapchat
The lens track's own angle sheet, pulled from Snapchat DMs, showed the top worry was the colour looking different on a dark eye than in the product photo. This creative describes tint and appearance only, never eye health or safety, and it won hook rate on Snapchat, where the audience already expects try-on content.
New collection, one ad for everything
Lost"New collection is here — sunglasses, frames and lenses."
- Format:
- Vertical, 15 seconds, carousel-style cuts across all three product types
The generic ad the account had run for a year, kept as the control. It lost on every metric to every successor ad. Worth publishing because it is the clearest evidence here: one ad selling three buyer intents reads as slightly wrong to all three, and the cost stays invisible until the lines are measured apart.
The dupe warning, stated as a favour
Lost"There is a fake version of this frame online. Here is how to tell."
- Format:
- Vertical, 18 seconds, side-by-side of the authentic frame and a known dupe, hinge and logo detail
A reasonable hypothesis worth keeping for what it revealed. It attracted attention but little intent to buy: viewers treated it as content about counterfeits generally rather than a reason to buy here, and it drew price comparisons with other sellers instead of converting. Authenticity works better woven into a product page than led with cold.
The frame width, stated in numbers before anyone asks
Winner"This frame is 138mm wide. Here is how to measure your own face."
- Format:
- Vertical, 15 seconds, text-on-screen measuring guide over product shots
Built to test whether the sunglasses winner was really about face shape or simply a concrete number. It performed nearly as well as the four-faces ad, telling us the buyer responds to a checkable fact over a style claim, in frames as much as sunglasses. It also cut measurement questions before checkout.
The lens colour as a dramatic transformation
Neutral"Watch her whole look change in three seconds."
- Format:
- Vertical, 10 seconds, fast cut before-and-after with heavy filter and music
A neutral result worth recording precisely because it looked like a natural winner. The heavy filter made the change look unrealistic and drew comments doubting the true colour; it neither hurt nor helped cost per purchase. The version that dropped the filter and showed the colour under normal light, described as a cosmetic tint, is what carried the lens line.
What we did — the optimizations, in order
Tag the back catalogue by product line, not just by angle
We exported the previous quarter of ads, tagged each by product line first — sunglasses, frames or lenses — then by angle and format, and split cost per purchase by line for the first time.
Why: A store selling three products under one campaign averages its results, hiding which line loses money. Splitting by product first turned a vague sense of underperformance into three specific, solvable problems.
Write one angle sheet per product, from questions specific to it
Three separate angle sheets: sunglasses from Instagram comments about face shape, frames from WhatsApp about the prescription process, lenses from Snapchat DMs about colour on different eye tones.
Why: The three buyers ask different questions on different platforms, and one shared angle sheet averages those questions away, the way the combined campaign averaged the results away.
Produce three short tracks instead of one long shoot
Twelve pieces across the three lines in two weeks, four per product, rather than one bigger shoot every two months covering everything at once.
Why: Smaller shoots per line let each track react to its own early results, and matched production cost to each line's own margin instead of spreading one shoot's cost evenly.
Test each product line as its own campaign
Three separate testing campaigns, one per line, equal budgets within the line, audience and landing page held identical inside each line while the creative varied.
Why: Mixing product lines inside one test campaign is the same mistake as mixing them inside one ad. A sunglasses ad and a frames ad never competed for the same buyer.
Read each line against its own objection, not a shared scorecard
For sunglasses we read hook rate against face-shape angles; for frames, the drop in prescription-related questions alongside cost; for lenses, Snapchat read separately from Instagram.
Why: A shared scorecard rewards whichever line is naturally cheaper to advertise and buries the rest. Reading each line separately surfaced the frames line's real win — fewer support questions.
Promote winners into three separate scaling campaigns
Each line's winning creative moved into its own scaling campaign, so a frames slowdown never pulls budget from a sunglasses peak and the reverse.
Why: Three products with three seasonal patterns should not share one budget pool. Separate campaigns let the Eid sunglasses surge be funded without starving the lens line during Ramadan.
Build a hook library per product, with seasonal peaks marked
Three hook libraries, each marked with when new creative should be ready — before Ramadan for lenses, before Eid for sunglasses, on a rolling basis for frames.
Why: The lines peak at different points in the year, and one shared cadence rushes one product while wasting effort on another with no season coming.
What changed
The headline figure is in the table above, and the telling number sits underneath it: once the split held, all three lines moved in the same direction instead of one carrying the other two. Sunglasses, already the account's quiet strength, improved further once its creative stopped sharing screen time with frames and lens content.
The frames line moved most in relative terms, which was the surprise. The winning creative was not more polished production — it was one plain sentence stating who handles the prescription and who does not. Removing that ambiguity did more for the line's cost per purchase than any styling change could have.
The lens line's answer was platform-specific in a way never tested before: the same colour content performed differently on Snapchat than on Instagram, and once read apart, the Snapchat-native version became worth scaling while a calmer, less filtered cut worked better on Instagram.
What we would do next
Extend the frames track's trust message from the ad into the product page itself, since the ad now explains the optician handover clearly but the page does not yet repeat it in writing when the customer uploads a prescription.
Pre-build the Eid sunglasses and Ramadan lens creative four to six weeks ahead of each season, on separate production timelines rather than the single shared calendar used before the split, since each line now has its own proven angle to build from.