Case study · Sushi · Barcelona
Sushi in Barcelona: a record month, in the month delivery usually dips
A single-location sushi restaurant in Barcelona, selling on three platforms, came to us at the end of July. We replaced the cover image, rebuilt every photo in the listing into one studio style, rewrote the menu for in-app search and reconfigured the advertising. August closed at €23,705.23 against a frozen baseline of €18,846.59 a month — up 25.8%, and the best month in the account’s recorded history. It was also our first month on the account.
Client profile
The restaurant is not named. A delivery account carries pricing, promotion depth and volumes, and a competitor two streets away reads pages like this one. The client agreed to publish the figures; they did not agree to publish the name, and we do not publish it.
- Client
- Name withheld at the client’s request
- Category
- Sushi
- Market
- Barcelona, Spain
- Locations
- 1
- Platforms
- Uber Eats, Glovo, Just Eat
- Work started
- End of July 2026
- Scope of work
- Cover image, full photo rebuild, menu and search copy, advertising
The starting point
This was not a rescue. The account was working: revenue had climbed from around €9,000 a month a year earlier to a little under €19,000 by the summer. Nothing was on fire, and that is the harder kind of account to improve, because everybody involved can point at the trend line and say it is fine.
What was wrong was that the listing had grown by accretion. Photos had been added over two years by different people with different phones in different light, and the result read as a page nobody was looking after. The cover image — the single thing a customer sees before deciding whether to tap at all — was one of those photos. The menu was written the way a printed card is written, not the way a person types into a search box. And the advertising was running, spending money, without anyone asking what it was returning.
There is a seasonal claim I want to be careful with. In our practice August is one of the weakest months of the year in delivery, and doing a record in August is therefore worth more than doing one in November. But this account’s history in our hands starts in the summer, so I cannot prove last August was weak for them — I do not have it. Everything below is measured against the two months immediately before we started, which is the comparison I can stand behind.
What we did
- 1 · The cover image
- Replaced first, before anything else, because it is the cheapest change with the largest reach. In a listing of forty restaurants the cover is the entire first impression, and it is what the advertising buys attention for — a paid impression on a weak cover is money spent to show somebody a photograph they scroll past. After the change, the click-through rate on the ads rose by 3.4%.
- 2 · Every photo, into one style
- We took the restaurant’s existing photographs and rebuilt them into a single studio look — the same background treatment, the same light, the same crop logic across the whole menu. Our designer does this with Photoshop and AI tooling rather than a shoot, which is why it can be done in days for a full menu instead of weeks for a handful of dishes. The point is not that each photo is prettier. The point is that a menu where every photo matches reads as a restaurant that is being run, and a menu of mismatched photos reads as one that is not.
- 3 · The menu, rewritten for search
- Dish names and descriptions rewritten around what people actually type, categories reordered so the things worth selling are not three taps deep, and the whole listing structured as a page to be searched rather than a card to be read. This is the slowest of the four workstreams and the one with the longest tail: search positions move over weeks, not days.
- 4 · The advertising, rebuilt
- Campaigns restructured and re-targeted rather than topped up. The account now runs at a ROAS of 13 on the platform’s own measure — revenue from customers who ordered within seven days of seeing an ad, divided by what the ads cost.
An honest note about that ROAS figure, because agencies quote it without one. A seven-day attribution window credits the advertising with every order placed in the week after somebody saw an ad, including orders from people who would have ordered anyway. It is a comparison tool — it tells you which campaign is working better than which — and it is not profit. The number that is profit is the one on the invoice at the end of the month, which is why we reconcile that separately and why the charges that are not the commission get their own article.
Results
+25.8%
Revenue vs baseline
€23,705.23 in August against a frozen baseline of €18,846.59 a month.
+3.4%
Ad click-through rate
From replacing the cover image, before any other change had taken effect.
The same figures as a table
| Month | Gross revenue | Month |
|---|---|---|
| Sep '25 | €9,092.44 | before |
| Oct | €9,861.87 | before |
| Nov | €10,625.32 | before |
| Dec | €10,448.23 | before |
| Jan '26 | €12,689.23 | before |
| Feb | €12,549.16 | before |
| Mar | €12,394.22 | before |
| Apr | €16,177.30 | before |
| May | €19,536.18 | before |
| Jun | €18,804.68 | before |
| Jul | €18,888.50 | before |
| Aug | €23,705.23 | with us |
Uber Eats
€13,112.70
Glovo
€6,314.28
Just Eat
€4,278.25
August by platform. Three platforms is deliberate: the same menu, the same kitchen and the same photographs work on all of them, so the third listing costs almost nothing to maintain and removes the risk of one platform’s algorithm deciding your month for you.
How this was measured
The baseline is frozen before the work starts and never recalculated afterwards: the average of June and July 2026, €18,846.59 a month, taken from the accounts themselves. August is compared against that figure, not against a target and not against last year.
Everything on this page is gross revenue as the platforms report it, which is not what reaches the restaurant — commission, advertising, promotions and per-order charges come out of it first. That is a deliberate choice: gross is the only figure both sides can verify in the same place. What the restaurant actually keeps is a separate calculation, and if you want to run it on your own account, the price calculator does it line by line.
Our own fee is charged on the growth over that baseline, after the platform commission and the extra advertising are taken out of it — never on gross revenue. In a month with no growth there is no invoice.
Method
- Where the figures come from
- The restaurant’s own platform accounts, to which we have access, and its own advertising reports. The client reviewed this page before publication and agreed to the figures appearing here.
- What is not published
- The restaurant’s name, its commission rate, its contract terms and its dish-level costs. Those are the client’s commercial information, not ours to publish.
- One month is one month
- This is a single month of work, and a single month can be lucky. We are publishing it because the change is large and the mechanism behind it is visible, not because one August proves a trend. The account continues and the next report will say what September did, whichever way it goes.
- The ROAS figure
- Taken from the platform’s own advertising report on a seven-day attribution window. It compares campaigns; it does not measure profit. See the note above.
All cases What platforms actually charge Price calculator
Start where this case started.
Send us read-only access to your platform accounts. We calculate your baseline and your real cost per euro from your own invoices, for free, and show you what the cover, the photos, the menu and the campaigns are costing you today. Nothing changes in your accounts until you approve it.
Get my free audit