The scenario
The brand sells gold-plated and sterling-silver necklaces, bracelets and a small line of name necklaces, direct to consumer, and started as an Instagram shop before moving onto its own Shopify store with KNET, Tabby and cash on delivery at checkout. One founder, a part-time photographer and no in-house media buyer, in a category where the customer cannot touch the product before she pays for it and has to decide, from a photo, whether the piece is worth the price.
Demand was never the issue. Eid and the June graduation run both push a spike in gifting, and the founder's own following had grown from genuine word of mouth rather than paid reach. What the brand had never solved was the doubt sitting underneath every sale: whether the necklace in the photo would look like that once it was actually on a neck, in a car mirror, at a wedding, six months from now.
- Monthly revenue band
- pending client sign-off
- Average order value
- pending client sign-off
- Fulfilment
- Own stock, gift box packaging, next-day courier across Kuwait
- Team
- One founder, one part-time photographer
Every ad looked flawless, and every unboxing photo online said otherwise
When we tagged the previous quarter of ads, every single creative was some version of the same shot: the piece macro-lit on velvet or marble, ring light reflected in every facet, colour graded to a warmer gold than the plating actually carried. It photographed beautifully, and it was doing real damage. The comment sections and the DMs told a second story running underneath the first: customers asking whether it would tarnish, whether it was actual gold or plating, and whether it would look this good once it was out of the box.
Nobody on the team was lying in these ads. Nobody was being especially honest either. The studio lighting was simply the industry default, the same treatment every jewellery seller in Kuwait uses, and it had quietly become a promise the delivered product had to live up to. When a customer opened a lightly duller, lightly smaller piece than the one glowing in her feed, the gap between the two was read as being cheated rather than being sold a demi-fine product honestly described.
That gap showed up as returns, as one-star reviews mentioning photos, and as a rising cost per purchase the founder assumed was a targeting problem. It was not targeting. The ads were selling a fantasy version of the product, and the delivered piece was quietly being compared against a photo nobody could actually manufacture.
What we did — the creative angles
The losers are here on purpose. A test with only winners was never a test.
The tarnish question, answered on camera
Winner"Yes, plated pieces can tarnish. Here's exactly how to stop it."
- Format:
- Vertical, 15 seconds, phone camera, polishing cloth in shot
This was the single most repeated worry in the brand's comments and DMs, and no ad had ever said the word out loud. Naming the risk before the customer has to ask it reframes the ad as honest advice instead of a sales pitch, and it disqualifies the buyer who wanted solid gold in the first two seconds rather than after a disappointed unboxing. It was the clear winner on hook rate and it held its cost per purchase as the budget increased, which the glossy studio creatives never did.
Plating thickness stated in plain numbers
Winner"1.5 microns of 18k gold plating — here's what that actually means."
- Format:
- Vertical, 14 seconds, text-on-screen over macro footage
Most competitors in this category never state a plating thickness at all, which trains the shopper to assume the worst. Publishing the actual number, plainly, did more for trust than any lighting choice could, and it specifically won with buyers who had been burned before by a plated piece sold as if it were solid.
On a real neck, in daylight, no filter
Winner"No studio. No filter. This is what it looks like at Friday brunch."
- Format:
- Vertical, 12 seconds, unedited phone footage in daylight
This directly answered the fear that sits behind almost every hesitant DM: does it look cheap in person. Showing the piece exactly as it would arrive, at the scale it is actually worn, in ordinary light, set expectations the delivery could meet, and it converted particularly well on Snapchat, where the audience already discounts anything that looks too produced.
For the girl who just finished her finals
Neutral"She survived June. This is what you get her."
- Format:
- Vertical, 18 seconds, gift-wrap unboxing filmed at home
Built for the two-week window around June graduations, when mothers and aunts buy on behalf of a graduate rather than for themselves. It performed respectably but narrowly, strong only inside its own short calendar window, which told us this belongs on the seasonal production schedule rather than running as an always-on creative.
Type her name and watch it appear
Neutral"Type any name. Watch it turn into a necklace."
- Format:
- Vertical, 10 seconds, screen-recording of the name-necklace preview tool
This carried strong engagement, plenty of comments tagging friends with their own names, and a genuinely weak cost per purchase. The interaction it earns happens in the ad itself rather than on the product page, and it turned out to work far better as a retargeting asset for people who had already visited the personalisation page than as a cold-audience opener.
The heavy-glow filter shot
Lost"✨ Shines like a diamond ✨"
- Format:
- Vertical, 10 seconds, sparkle overlay effect added in editing
This is the loser worth publishing, because it lost for the exact reason the whole engagement started: the sparkle overlay made the piece look like fine gold, buyers arrived expecting that, and the plated reality landed as a letdown. It produced a spike of orders in its first week and a matching spike of return requests two weeks later, once the packages had actually arrived, which is the worst possible shape for a cost-per-purchase number to take.
Priced against the gold souq by the gram
Lost"Why pay souq gold prices for a look that doesn't need to be gold?"
- Format:
- Vertical, 16 seconds, split screen comparing per-gram prices
A reasonable hypothesis that confused more than it converted. Comparing a demi-fine piece against solid gold by weight invited exactly the wrong question, made shoppers wonder what they were actually buying, and pulled in comment threads debating gold purity that had nothing to do with the product. It is a comparison worth avoiding rather than winning.
What we did — the optimizations, in order
Tag the back catalogue before shooting anything new
We exported the previous quarter of ads and tagged every one by angle, lighting treatment, format and language, then worked out what share of spend had sat on creatives below the target cost per purchase.
Why: The tagging is what surfaced that every single ad shared the same over-lit studio treatment, which nobody had noticed because each ad had been approved on its own rather than read as a set. That pattern, not any single ad, was the actual problem to fix.
Build the angle sheet from returns and DMs, not a mood board
We read a year of return reasons, DM questions and comments, and pulled the seven worries that came up most: tarnish, plating thickness, real gold versus plated, size in real life, engraving turnaround, and whether it suits her skin tone.
Why: A jewellery brand's return reasons are a more honest brief than any brainstorm, because they are written by the customer who almost did not buy, or who bought and regretted it. Every angle on the sheet traces back to a real sentence someone wrote.
One testing campaign per platform, everything else frozen
One campaign per platform, one ad set per angle, equal budgets, six days, with the audience, the offer and the landing page held identical across every ad set.
Why: Jewellery buyers are unusually sensitive to how a piece is framed, so letting the offer or audience shift alongside the creative would have made it impossible to say which change actually moved the needle. Freezing everything else is what let the honesty-first hooks be judged fairly against the glossy ones.
Read hook, then hold, then click, then cost
Each creative was judged in that order, and we wrote down the exact step it failed at rather than a single pass-or-fail verdict for the whole ad.
Why: The gold-souq comparison and the sparkle-overlay ad look identical in a cost-per-purchase column, and they needed opposite fixes: one confused the offer, the other created a delivery mismatch. Diagnosing the step is what made the next production round cheaper rather than just different.
Promote winners into a scaling campaign that never gets rebuilt
Winning creatives moved into a dedicated scaling campaign, while the testing campaign kept its own ring-fenced slice of the budget for the next round of angles.
Why: Mixing new tests into the campaign carrying the brand's revenue resets its learning every time an ad is added. Two campaigns with two separate jobs scales far more predictably than one campaign trying to do both.
Write the honesty rules down for future shoots
The photographer now shoots one daylight, on-skin frame for every product alongside the studio shot, and every ad states the plating thickness or metal content in the first line of copy.
Why: This is the rule that outlives any single winning ad, which will fatigue within a season. It is what stops the brand quietly drifting back into the same over-lit habit the next time a new collection needs photography.
Plan the Eid and graduation variants a month ahead
Seasonal hooks for Eid gifting and June graduations are now scripted and scheduled four weeks in advance, cut against the winning honesty-first bodies rather than as a separate campaign built from scratch.
Why: The graduation-gifting angle performed only inside a two-week window, and shipping it late meant missing most of that window. Planning it ahead of the calendar rather than reacting to it is what lets a narrow seasonal angle actually earn its keep.
What changed
The headline number sits in the table above once the client signs it off, and the shape of it is what matters: cost per purchase came down while budget went up, driven by the share of spend sitting on honesty-first creatives moving from a minority to a majority. No single ad went viral. The pattern did the work.
The clearer finding was about which kind of ad this category actually rewards. The three winners all named a real doubt out loud — tarnish, plating thickness, how it looks in daylight — before the customer had to ask. The two losers both did the opposite: one manufactured a sparkle no delivered piece could match, and the other invited a comparison the product was never trying to win.
There was a second effect the founder noticed within weeks. Because the winning ads answered the tarnish and plating questions up front, the volume of DMs asking those exact questions dropped, and the reviews that arrived stopped mentioning photos at all. The ad was doing the reassurance work a human used to type out by hand every evening.
What we would do next
Take the honesty-first pattern onto the product pages themselves, because the tarnish and plating answers now living in the ads still are not repeated where the customer makes her final decision. That is a page fix, not a creative one, and it is where the next block of returns gets prevented rather than refunded.
Second, build the Eid and graduation variants against a shared production calendar rather than around whichever founder has time that week, so the seasonal window is planned rather than sprinted through in its final days.