The scenario
The retailer sells watches, cufflinks, wallets and a smaller line of women's accessories, with two physical kiosks in Kuwait's busiest malls alongside a Shopify store that carries the online orders. Watches sit at a higher average ticket than most of the category, which makes Tabby installments and the National Day and graduation seasons the two biggest levers in the calendar, and it makes a wasted click more expensive than in almost any other jewellery-adjacent business.
Six years of running paid ads had left an account that had never been rebuilt, only added to: new campaigns launched for every promotion and never retired, an audience split by age, gender and interest that had ballooned into dozens of overlapping ad sets, and a creative library that had not grown at the same pace as the spend it was expected to carry.
- Monthly revenue band
- pending client sign-off
- Average order value
- pending client sign-off
- Fulfilment
- Warehouse stock plus two mall kiosks, gift packaging, next-day courier across Kuwait
- Team
- Owner-operator, two kiosk staff, one part-time media buyer
Every extra dinar of budget made the account worse, not better
The retailer had genuine, proven demand: the kiosks sold well in person, the Shopify store converted at a respectable rate, and National Day and the June graduation run were reliable spikes on the calendar. The problem only appeared at the moment the owner tried to spend more ahead of those spikes. Cost per purchase climbed within days of any budget increase, and it climbed fastest on exactly the campaigns meant to carry the extra weight.
The account structure explained why. Meta alone was running fourteen prospecting ad sets, most of them near-identical splits by gender and watch style, all bidding in the same auction against each other for the same pool of Kuwaiti shoppers. Purchases were tracked from the browser only, which was undercounting iOS conversions and making Snapchat look weaker than it actually was. And the whole structure was being asked to carry a seasonal spend increase on a creative library of five ads, most of them over a year old.
Raising the budget on this account did not buy more customers. It bought more overlap, more frequency on the same shrinking pool of people who had already seen the same five ads, and a cost per purchase that rose in direct proportion to how badly the owner wanted National Day to be the retailer's best month. The account had a structure problem wearing a budget problem's clothes.
What we did — the creative that carried the spend
The losers are here on purpose. A test with only winners was never a test.
Fourteen overlapping prospecting ad sets folded into one per platform
WinnerGender and watch-style splits removed, replaced with one broad prospecting audience on Meta and one on Snapchat
- Format:
- Account restructure, not a new creative
The fourteen ad sets were one idea split fourteen ways, all bidding against each other for the same shoppers and starving each other of the purchase volume the algorithm needed to learn. Folding them into one broad audience per platform let each platform learn from a single, larger pool of data instead of fourteen thin ones, and cost per purchase stabilised within the first budget cycle after the merge.
Server-side tracking connected before any budget moved
WinnerMeta CAPI and Snapchat's server-side API connected to the same purchase event, deduplicated against the pixel
- Format:
- Tracking infrastructure, run before scaling began
Browser-only tracking was undercounting purchases on iOS specifically, which is a large share of this retailer's higher-spending customers, making Snapchat look like the weaker platform when it was actually being under-credited. Fixing measurement first is what made every scaling decision afterward trustworthy instead of a guess wearing a dashboard.
Confirming six proven creatives before the first budget step
WinnerProduction sprint added three new creatives to the existing library of five before any spend increase
- Format:
- Creative supply check, gating condition on the scaling plan
Adding budget to a five-ad library concentrates spend on two or three winners, which is exactly how frequency climbs and cost per purchase drifts up while everyone blames the platform. Requiring six proven creatives before the first budget step meant the increase had somewhere to go besides the same faces the audience had already seen a dozen times.
Twenty percent budget steps, held five to seven days
WinnerBudget increases capped at twenty percent per step, with a rollback rule after two consecutive days above the cost ceiling
- Format:
- Scaling rule applied to the consolidated prospecting campaign
The owner's previous instinct was to double a promising campaign's budget the day before a sale started, which is what taught the account to spike and crash on every occasion. Smaller steps held long enough to read gave the algorithm time to find new buyers at each level instead of just showing more ads to the same ones.
Adding TikTok as a third platform immediately, to chase National Day reach
LostTikTok launched with the same broad prospecting audience the week before National Day, before Meta and Snapchat had finished stabilising
- Format:
- New-platform launch, run concurrently with the consolidation work still in progress
A reasonable idea run at the wrong time, and it is worth publishing because the mistake was sequencing, not the platform itself. Launching a third platform while the first two were still absorbing the consolidation split the media buyer's attention and the account's early purchase data three ways during the exact week the account most needed a stable read, and TikTok's own cost per purchase never had a fair chance to settle before National Day traffic hit.
What we did — the optimizations, in order
Fix measurement before touching a single budget
Connected Meta CAPI and Snapchat's server-side API to the same deduplicated purchase event, and put every platform's reporting against one blended marketing-efficiency number rather than each platform's own claimed return.
Why: Every decision after this one depends on knowing which platform actually produced a sale. Scaling on top of a measurement gap just scales the gap, and the account had been doing exactly that for years.
Consolidate the account into one campaign per job, per platform
One prospecting campaign, one retargeting campaign and one campaign for past kiosk and online customers, per platform, replacing the fourteen near-identical ad sets that had accumulated over six years.
Why: Consolidation is what let the budget the account was already spending start learning as one signal instead of fourteen competing ones, before a single extra dinar was added. Most of the early improvement came from this step alone.
Confirm at least six proven creatives before any budget step
Ran a short production sprint to bring the library from five ads to eight, spanning watches, cufflinks and the women's accessories line, before approving the first budget increase.
Why: Six is the floor below which extra spend has nowhere honest to go except more frequency on the same faces. This gate stopped the owner's instinct to raise budget the same week a sale was announced, before the account had anything new to show for it.
Raise budget in twenty to thirty percent steps, held five to seven days
Every increase was capped at thirty percent, held for at least five days, with a rollback rule triggered by two consecutive days above the cost ceiling rather than a gut call.
Why: Doubling a budget the day before a sale and judging it two days later is not scaling, it is a stress test the account usually fails. Steps held long enough to read gave the algorithm room to find genuinely new buyers instead of just paying more to reach the same ones faster.
Add the second and third platform only after the first is stable
Snapchat was scaled second, once Meta had held its cost ceiling for two full weeks, and TikTok was deliberately delayed past National Day to the following graduation season rather than launched under seasonal pressure.
Why: Launching a new platform inside an occasion week, which the account had already tried once and lost on, splits attention exactly when the existing platforms most need stable management. Waiting for the next calm window let TikTok's own learning phase run without seasonal noise distorting the early read.
Shape National Day and graduation season budgets in advance
National Day and June graduation budgets are now planned four weeks ahead with a pre-approved ceiling for the higher cost per thousand both seasons carry, instead of being decided the week traffic starts moving.
Why: Both occasions raise the cost per thousand across the whole market, watches included, and accepting that in advance with a planned ceiling is a very different experience from discovering it mid-campaign and panicking about return on spend.
Recognise when the ceiling has been reached
Once return started falling at every budget step regardless of which creative was carrying it, the team stopped pushing spend further and moved the conversation to average order value and kiosk cross-promotion instead.
Why: Continuing to feed frequency past this point is buying the same customer twice rather than finding a new one. Recognising the ceiling honestly is what protected the margin the earlier scaling work had just built.
What changed
The table above carries the figures once the client signs them off, and the shape worth flagging is that spend rose while cost per purchase came down, which is the only version of scaling worth reporting. Most of the early movement came from consolidation and measurement, before a single extra dinar was added — the account had been paying a structural tax on every purchase for years.
The TikTok misstep is the finding worth keeping. It confirmed that this account's real constraint was never platform count, it was sequencing: adding a new platform under seasonal pressure re-created the exact instability the consolidation work had just fixed on the other two. Waiting for a calm window is now a written rule rather than a lesson the account has to relearn every occasion.
By National Day, the consolidated structure meant the seasonal budget step went almost entirely to reach and frequency on proven creative rather than to untangling overlapping ad sets, which is the difference between a scaling plan and a scramble.
What we would do next
Bring TikTok back in properly ahead of graduation season, in a calm window with its own creative rather than the recycled prospecting assets it launched with under National Day pressure, and give it the same five-to-seven-day step discipline the other two platforms now run on.
Second, connect the kiosk point-of-sale data to the same measurement stack so a customer who discovers the brand online and buys in person at the Avenues or Salmiya stops looking like a lost online sale, which is currently understating how well the paid account is actually performing.