Revenue up 62.55%on a budgetthat went down 5.12%.
Marpur manufactures mattresses and beds in Poland, runs a single showroom in Kepno and sells mostly online. Revenue from new customers had been stopping at the same ceiling for a long time, website fixes and SEO work had not moved it, and the Meta budget had been cut back shortly before the work started. Below is the year-on-year read across the whole business, and the order the account was rebuilt in.
- +62.55%
- revenue across the entire business, year on year
- -5.12%
- advertising budget over that same year
- +74.09%
- average return across the whole store, year on year
- 1
- Advantage+ campaign in place of the whole Meta structure
A ceiling everyone felt and nobody could locate.
A ceiling nobody could locate
Revenue from new customers kept stopping at the same level, and there was earlier evidence that higher was achievable. Regular website fixes and SEO work, both aimed at traffic volume and conversion rate, never showed up in the results. A constraint that survives the two obvious fixes usually sits somewhere else entirely.
Sales that leaned on people who already knew them
Most revenue came from returning customers and referrals rather than from campaigns. No channel was predictably responsible for a solid share of new-customer revenue, which makes a good month impossible to repeat on purpose and a bad one impossible to explain.
A decision that takes weeks
A mattress is thought about, not picked up. The window between first contact with the brand and the purchase is long, so conversions spread across time and across touchpoints, and attributing a sale to one performance channel stops being reliable. Every budget decision here had to be made on evidence that arrives late.
Split one side, collapse the other, then raise the budget slowly.
Performance Max split in two
One shared Performance Max campaign was covering mattresses, the core of the business, and beds, which accounted for a fraction of sales while consuming budget on a par with the main product. It was split into separate structures, one for each. Campaigns sharing a budget and a goal bid against each other, and the algorithm has no way to learn which product is the one that pays.
Meta collapsed into a single Advantage+ campaign
The previous Meta structure was large and pointed mostly at remarketing, so the budget kept going back to people who were already interested. It was simplified into one Advantage+ campaign with audience segments arranged so that the destination of every message was known, aimed at broader reach and new customers instead. The budget stayed low while the campaign learned, which was a decision rather than a delay.
Budget raised in steps, judged on the store
Increases were small and controlled, and each one waited for the return to hold before the next. The number that decided them was the return across the entire store, not the attribution each platform reports about itself. When the client signalled that peak season was coming, budgets went up in Google Ads and Meta Ads together, on evidence collected during the quiet months.
Creative refreshed before the numbers forced it
More video, UGC and dynamic product ads went in alongside the structural work. When performance started slipping at an unchanged budget, which is what creative fatigue looks like before anyone names it, a new video went out, and that one creative became what carried the next rise in Meta budget. TikTok was launched off content the client already had, as a brand channel rather than a sales one.
One year against the one before. Two figures rose, one was cut on purpose.
The study reports a year-on-year comparison measured across the whole business rather than across the ad platforms. Three figures come out of it. Two of them are on the chart below, and the third is the reason those two count.
Revenue rose while the money behind it went down, and the second bar is those two facts in one figure: more sales on less spend. The 5.12% cut sits in the strip at the top of this page and not on this chart, for two reasons. A reduction drawn as a bar beside two increases reads as a third increase. And the three figures do not quite close on each other: the revenue rise set against the budget cut lands on a smaller improvement than the 74.09% published here. The source labels that figure an average return, written as ROAS across the whole business rather than as the ROAS a platform reports about itself, and an average of period returns is not the same arithmetic as one year of revenue divided by one year of spend. Close enough to trust the direction, not close enough to draw as one calculation.
“Working with MTA runs very smoothly and at a high standard. The team is committed, communication is quick and to the point, and everything is delivered on time. The results of our joint work are clearly visible – sales are growing steadily, and the marketing and sales activities we undertake bring measurable results. We are pleased with the cooperation and look forward to growing together and to further successes.”Mariusz Kaczmarek, Owner, Marpur
The change worth copying looks like doing less.
Run one question against your own account: are two of your campaigns bidding for the same budget with different jobs? A core product and a marginal one inside a single Performance Max campaign hand the system money and no instruction, and you will read the result as a demand problem when it is a structure problem. What happened after that is specific to the category. A mattress is a considered purchase in a market with a fixed number of buyers each year, and yours may pay back patience at a very different rate.
The second one is harder, because for months it looks like inaction. The budget here stayed low while the rebuilt campaigns gathered enough data to be optimised on, and it only moved in small steps once the return held. The return that decided each step was the whole store, not the figure each platform reported about itself. If you cannot sit through the quiet months, the scaling you do in season is a bet rather than a decision.
The third is a question about your own contract. The source records that the fee here grew with the client revenue rather than with the budget spent. Ask what your agency is paid on. If the fee grows with spend, waiting costs them money, and you will rarely be advised to wait.
This work was delivered by MTA Group, which Zero Fluff Digital is part of, by the same specialists who would work on your account. Every figure here comes from the case study MTA published, republished here with their consent. The reviews behind it are public and verified on Clutch, where clients rate the work rather than the agency describing itself.
Are two of your campaigns bidding for the same buyer?
30 minutes, no deck. You leave knowing where your budget competes with itself, and what your store-level return says that the platform dashboards do not.
Book a callNext case studyPet tech: cost per customer cut from $332 to $163 →