The challenge
01
No salesGoogle Ads had run for about two years without delivering sales.
02
Leaking analyticsA free PrestaShop plugin dropped conversion and cart-value data.
03
Tight budgetNo room for spend on unprofitable product groups.
Case studies / 01
How we fixed broken analytics in furniture e-commerce and reached 947% ROAS on a limited budget.
Quarter analysed: 1 March – 31 May 2026 (vs 29 Nov 2025 – 28 Feb 2026)
947%
Average ROAS across the quarter
1 075%
Best month (April 2026)
461.9
Conversions (from 5 before the fix)
64.3kPLN
Revenue from Google Ads
What we did
01
Sealed measurementAn advanced tracking plugin recovered purchase, value and path data.
02
Selected assortmentMargin and potential analysis, then only the key product groups.
03
Rebuilt accountPerformance Max on those groups, a brand campaign and remarketing.
Key takeaways
01
Measurement firstNo account earns if the algorithms learn from broken sales data.
02
Focused budgetOn limited spend, advertise only what actually earns.
03
Closed funnelProduct campaigns, brand and remarketing work together.
Raw platform evidence / 02
Unedited Google Ads screenshots, one row per reporting window. Numbers on the left are read straight from the account.
01 / 05
1 Dec 2025 – 15 Aug 2026 · whole account
- Clicks
- 29.6k
- Impressions
- 943k
- Actual ROAS
- 884.34%
- Conversions
- 1k
Full account view after the measurement layer was sealed. The step-change in the curve is the month tracking was fixed.
02 / 05
1 Mar – 31 May 2026 · analysed quarter
- Conversions
- 461.90
- Actual ROAS
- 947.25%
- Clicks
- 14.3k
- Impressions
- 476k
The quarter we report on: average ROAS 947.25% from only 14.3k clicks.
03 / 05
1–30 Apr 2026 vs 2–31 Mar 2026 · peak month
- Purchases
- 27.95 (+17.59)
- Actual ROAS
- 1,074.91% (+386.20 pp)
- Sales value
- 21.5k PLN (+9.77k)
- Cost / conv.
- 13.46 PLN (−4.27)
Peak month. Return climbed while cost per conversion fell — the signature of accurate conversion data.
04 / 05
1–30 Apr 2026 vs 2–31 Mar 2026 · volume
- Conversions
- 148.81 (+52.53)
- Conv. rate
- 3.24% (+0.83 pp)
- Clicks
- 4.59k (+1.3k)
- Impressions
- 151k (+36.9k)
Same month, volume view: more conversions at a higher conversion rate, not just more traffic.
05 / 05
1 Mar – 31 May 2026 vs 29 Nov 2025 – 28 Feb 2026
- Purchases
- 78.55 (+65.55)
- Conv. value
- 64.3k PLN (+60.5k)
- Cost / conv.
- 14.70 PLN (−176)
- Conv. rate
- 2.95% (+2.80 pp)
Quarter over quarter: 13 transactions became 78.55, and cost per conversion dropped by 176 PLN.
Results
First month after the fix
March 2026
689% ROAS
Google finally received precise signals about valuable conversions.
- Conversions
- 96.28
- Sales value
- 11.7k PLN
- Cost per conversion
- 17.73 PLN
- Conversion rate
- 2.41%
Peak month
April 2026
1,075% ROAS
Return climbed while the cost of each conversion fell.
- Conversions
- 148.81
- Sales value
- 21.5k PLN
- Cost per conversion
- 13.46 PLN
- Conversion rate
- 3.24%
Quarter total
1 March – 31 May 2026
947% ROAS
78.55 transactions, against 13 in the quarter before.
- Conversions
- 461.90
- Sales value
- 64.3k PLN
- Cost per conversion
- 14.70 PLN
- Conversion rate
- 2.95%
March figures are April's reported values minus the month-over-month changes shown in Google Ads for 2–31 March.
More client results / 03
Three more accounts, read from the same system.
Live Google Ads accounts we run in home and furniture e-commerce in Poland. Client names are withheld. The softer months stay in.
Closed months: 1 September – 31 October 2026
02 E-commerce · Tables & occasional furniture · Poland
15% more revenue in a month, on the same budget.
1,769%
ROAS, Sep–Oct
+14.8%
Conversion value, Oct vs Sep
+1.1%
Ad spend, Oct vs Sep
55.9PLN
Cost per purchase
How the account is built
01
Split by priceTables moved into two Performance Max campaigns, under and over 1,000 PLN.
02
Bidding on valueSales campaigns switched from maximising conversions to conversion value.
03
Seven campaignsThe product groups, with brand search and remarketing alongside.
What the numbers show
01
More purchases90.4 in September became 103.3 in October.
02
Higher-ticket tablesThe over-1,000-PLN campaign returned 37.5k PLN on 1.5k PLN of spend.
03
Not brand-ledROAS is 1,669% without brand search, which is 7% of value.
What it teaches
When the budget is tight, tell the algorithm what a sale is worth, not only that it happened.
03 E-commerce · Living room & wall units · Poland
A 1,577% return that doesn't lean on brand search.
1,577%
ROAS, Sep–Oct
1,452%
ROAS without brand search
129
Purchases
109PLN
Cost per purchase
How the account is built
01
By product groupPerformance Max split across eight campaigns in total.
02
Brand protectedA separate search campaign for people already looking for the store.
03
Budget follows resultsThe two wall-unit campaigns get the largest share.
What the numbers show
01
One strong categoryWall units bring 65% of value, the main lever and the main risk.
02
Brand kept small10% of conversion value on 2.3% of spend.
03
Steady month to monthSpend +2.2%, conversion value +2.9%, ROAS flat.
What it teaches
Report ROAS with and without brand. The second number is the one that shows new demand.
04 E-commerce · Sofas & upholstered furniture · Poland
265k PLN in two months, and where the next gain sits.
265.4kPLN
Conversion value, Sep–Oct
957%
ROAS, Sep–Oct
159
Purchases
174PLN
Cost per purchase
How the account is built
01
Largest budget27.7k PLN across 15 campaigns.
02
By category and pricePerformance Max by product, with wall units split by price band.
03
Best performerCorner sofas, 75.2k PLN of value on 3.4k PLN of spend.
What the numbers show
01
Concentrated returnThe top three campaigns bring 62% of value from 45% of spend.
02
First to reviewTwo campaigns spent 1.1k PLN with no recorded value.
03
The softer monthOctober value fell 9% on flat spend, ROAS from 1,000% to 914%.
What it teaches
At this size, the gain is in moving budget toward what already works, not in adding more.
How to read these numbers: conversion value and purchases are as reported by Google Ads and imported daily into our system, on 6 November 2026. ROAS is conversion value divided by ad spend. Figures don't include margin. August starts on the first imported day, so it isn't a full month and isn't used in the totals.