Case studies / Care Window
Care Window · Beauty & personal care · Saudi Arabia · Client since Apr 2025
How Care Window grew monthly sales 3.3× and held a 3.89× blended ROAS across four platforms
A Saudi Salla store growing on a catalogue where the best sellers sell out without warning and new creative arrives slowly. We built an automated winners engine with an availability guard, so the budget only ever follows a product the store can actually ship, and judged every platform on one number the store could confirm.
Last updated 12 Sep 2026 · figures through Aug 2026 · source: Salla and the four ad platforms
3.89×Blended ROAS17 months, Salla sales ÷ all-platform media
+53%Sales year on yearJan–Aug 2026 against the same eight months of 2025
9,700OrdersSalla, Apr 2025 – Aug 2026
The case in 30 seconds
The challenge
A catalogue whose best sellers go out of stock without warning, a slow supply of new creative, and four platforms all buying from the same feed with four different scoreboards.
What we changed
A weekly winners engine that ranks products on the store's own revenue and writes the labels itself, an availability guard on every product set, a defined job for each platform, and one blended number.
The result
3.3× the average monthly sales, 9,700 orders and a blended ROAS of 3.89× held across seventeen months and four platforms.
Our diagnosis
The budget kept landing on products the store could not ship.
One map of the catalogue and the four accounts before changing a campaign.
- A catalogue that movesBest sellers change week to week and sell out without warning
- Four platforms, one feedGoogle, Meta, Snapchat and TikTok all buying from the same catalogue
- Creative supply is the limitNew assets arrive slowly; the account cannot refresh its way out of fatigue
- Winners picked by handA buyer's memory of what sold last week
- Four scoreboardsFour platform ROAS numbers and one Salla total that disagreed with all of them
The opportunity
Automated winners engineAvailability guardA job per platformBlended measurement
→Budget that only follows what is in stockBefore → After
What actually changed
BeforeAfter
BeforeWinners picked by hand
AfterA weekly engine ranks products on 60-day store revenue and writes the labels itself
BeforeSold-out products still promoted
AfterAn availability clause in every product set: a product that goes out of stock leaves the campaign on its own
BeforeOne budget spread evenly across four platforms
AfterEach platform funded for the job it proved it could do, and one of the four closed
BeforeFour platform ROAS numbers
AfterOne blended number: Salla sales divided by total media spend, every month
BeforeRefresh waiting on new creative
AfterThe catalogue does the refreshing; the product mix rotates even in a month with no new assets
Growth strategy
One growth system. Four connected levers.
01 — Media
Google for volume, Snapchat for cost, Meta for the middle, TikTok tested and closed.
Prospecting→Catalogue→Retargeting→Repeat
Across the reported months Google carried 58% of purchases; Snapchat produced 27% of them on 9% of the media.
02 — Product
A weekly winners engine: rank on the store's own 60-day revenue, enter above a threshold, leave below a lower one.
The same ranked list feeds Google Merchant Center and the Meta product set, so all four platforms promote the same winners.
03 — Creative
A limited supply of new assets, rotated against a catalogue that updates itself every week.
When new creative is scarce, the product mix has to carry the novelty.
04 — Measurement
Salla sales divided by total media spend, every month, across every platform.
Blended ROAS stayed between 2.51× and 5.80× for seventeen consecutive months.
Performance over time
What happened next
Salla and platform data by month. The engagement began in April 2025; January to March 2025 is the store's own run-rate before it. Monthly sales are indexed so the trend is visible without disclosing revenue. Two things are worth reading together. Individual months swing hard, because a sold-out best seller or a month without new creative moves them; August 2026 is the clearest case, a month in which media was deliberately cut by a third while blended ROAS held at 3.88×. The annual run-rate underneath those swings is the third chart, and it has gone one way.
The business impact
Store numbers, not platform numbers. April 2025 to August 2026 unless stated. Year-on-year comparisons use the same calendar months so seasonality is not doing the work.
3.89×Blended ROAS17 months, Salla sales ÷ all media
9,700OrdersSalla, 17 months
+53%Sales, Jan–Aug 2026 vs 2025orders over the same eight months +22%
+35%Average order valueJul 2026 vs Jan 2025
+34%Annual sales run-ratelast 12 months vs calendar 2025
6.5% → 12.7%Returning-customer rateJan 2025 → May 2026
Absolute revenue and ad spend are confidential to the client. Every figure is a ratio, an index or a count from the client's own Salla and platform reporting for the months stated. ROAS here is blended: total Salla sales divided by total media spend across every platform, which is the number reported to the client each month; each platform's own attribution is lower and is tracked separately.
Creative strategy
The creative system
The pieces that did the work, not the whole gallery.
- Format
- Catalogue ad on the winners set · Conversion
Why it workedThe set rebuilds itself every week from the store's own revenue and drops anything out of stock, so the ad is current without a new asset being made.
- Format
- Snapchat product video · Prospecting
Why it workedThe cheapest purchases in the account: 27% of all purchases on 9% of the media, at a cost per purchase around a quarter of the account average.
- Format
- Bundle and routine sets · Basket
Why it workedBeauty sells in routines, not single items. The average basket ran between three and four products, and the average order value rose 35%.
What we learned
01
When the catalogue moves faster than the buyer, automate the catalogue.
A weekly engine ranks products on the store's own 60-day revenue and writes the labels that Merchant Center and the Meta product set read. Every set carries an availability clause, so a sold-out best seller leaves the campaign by itself instead of collecting budget.
02
The last five months grew on basket size, not on more customers.
Against the same five months a year earlier, sales are up 7% while orders are down 14% and the average order is up 25%. That is what a thin creative supply and an unstable stock position look like in the numbers. It is on this page because it is the problem the next phase has to solve.
03
The cheapest channel was not the obvious one.
Snapchat took 9% of the media and returned 27% of the purchases. TikTok, opened in the same period, produced 132 purchases before it was closed. Both answers came out of the same test, and only one of them is the one an agency usually quotes.
04
Judge this account on the run-rate, not on any single month.
Months move with what is in stock and what creative arrived, so any one of them can be read either way. Over the eight months of 2026 so far, sales are up 53% and orders up 22% on the same months of 2025, and the rolling twelve-month run-rate is 34% above calendar 2025. August was a deliberate one-third cut in media that still returned 3.88×. The system holds its efficiency when it is throttled, which is the property that makes the next stock and creative fix worth making.
Scaling an eCommerce brand but losing efficiency as spend grows?
We review media, creative, products and conversion together to identify what is limiting profitable growth.
