Electric scooter e-commerce in Sibiu. Average price 3,300 RON per unit. Running Google Ads for eight months with a previous agency. Weekly reports showed ROAS between 3 and 4. One day the founder looked at his bank statement and saw nothing was coming out. That's when the work started.
How we ended up looking over his account.
He wrote us the second week of January. Not because he was desperate. Because he didn't understand what he was seeing. The weekly PDF from his current agency had just landed. First page in bold: ROAS 3.2, 162 orders, campaign X „consolidating performance". Underneath, his bank statement. Seventh-grade math said he was down 9,000 lei for the month.
We didn't promise anything on the first call. We said it straight: „Give us access for a month. We won't touch anything. We look over everything, run a real profit audit, and only then do we tell you if we can actually do something together." He agreed. Then he said something I remember to this day: „If you tell me things I already know, we're paying for nothing again."
First week we only pulled data. Google Ads, Google Analytics, Shopify, SmartBill, courier. Spreadsheets on spreadsheets. The problem started taking shape on day three, when we tried to match Shopify sales with real SmartBill costs. They didn't match. Not even close. Where Google said a product was making profit, SmartBill said you were losing money on every order. By the time we finished matching numbers across all 120 products in the catalog, we understood what was happening: 84 of them were being advertised at a loss and nobody knew.
What we found under the hood.
It's not all of it. It's the rest that matters. The first two weeks of digging surfaced things that three agencies before us had collectively ignored.
- 01
Real POAS across the account was under 1×. Meaning every order from Google was losing about 50 lei once you subtracted product, shipping, courier fee, returns, and Shopify's cut. The 3× ROAS looked pretty only because nobody was deducting those costs.
- 02
There was no link whatsoever between SmartBill and Google. Two parallel universes. Google optimized for what it thought was „good". The accountant counted separately. The founder, in the middle, kept wondering why profit wasn't showing up anywhere.
- 03
Catalog of 120 products. We ran a simple exercise: net margin per product. Took two days. Out of 120, only 36 had margin over 25% after all costs. The other 84 were either break-even or straight negative. Worse, Performance Max was pushing the 84 aggressively because they had higher search volume. The algorithm was doing exactly what you asked it to: conversions. You couldn't blame it. You'd asked for the wrong thing.
- 04
Google Shopping feed was weak. Truncated titles, missing base specs (range, top speed, weight, battery), inconsistent imagery across products. People clicked, saw a page that felt half-baked, left. CTR 1.8%. Conversion rate 0.19%.
- 05
Enhanced Conversions off. Zero offline conversion tracking for returns. Meaning Google thought it had closed a sale and kept it in stats even after the customer returned the scooter two weeks later. The algorithm was learning on lies.
What we did, in the order it happened.
Month 1, week 1. A real profit audit.
Linked Shopify, SmartBill, and Google Ads in one spreadsheet. Product by product. Cost by cost. Net margin after shipping, courier, returns, and platform taxes. The result went into a 40-row document sent to the founder at 11pm. His first reaction, on WhatsApp: „I feel sick." His second, an hour later: „Ok. What do we do?"
Month 1, week 2. 84 products out.
Split the catalog into three tiers. Top: 18 products with margin over 35%, all premium scooters, the ones actually making profit. Middle: 18 more, margin 25-35%, premium accessories, helmets, bags. Bottom: 84 products that just had no reason to be in campaigns. Pulled all 84 in one evening. Founder drew a deep breath when we told him. We warned him he'd see total volume dip for about a week. He said ok.
Month 1, week 3. Performance Max restructure.
Killed the single PMax and built two separate Maximize Conversions campaigns. One dedicated to the 18 premium scooters, aggressive budget, specific asset groups. One conservative for accessories. Each with its own creative, no mixing. Cut the 300% tROAS bidding and let the algorithm do its thing on clean conversion objectives.
Month 1, week 4. Feed rebuilt from scratch.
Titles rewritten based on search intent. Full specs in description: range, top speed, battery, weight, everything. Imagery redone on one template, neutral background, product in context. Custom labels for auto-segmentation (tier, margin, season). Enhanced Conversions on with first-party data. Offline conversion tracking for returns, so Google loses the order from stats when we send the money back. Feed clean by Sunday midnight.
Month 2. Meta Ads in parallel.
Once Google had stabilized on the new setup (conversions were coming in at 33.68 lei average vs 120 before), we opened Meta too. Small budget, 15,998 lei across the whole 60 days, funded from the savings on the trimmed catalog. Full funnel on the eight hero products. Pixel with eventId dedup, CAPI up. Meta conversions: 195. Purchase ROAS between 43 and 65 depending on audience. Google captured existing intent. Meta generated new demand, for buyer segments that didn't know that specific scooter even existed.
Month 2. Daily POAS reporting.
Final step, the one that changed how we talked to the founder. Auto report in Google Sheets with POAS per product, per campaign, per day. No more ROAS talk. No more click talk. Net profit per unit sold. Two weeks in, he was going into the report himself and telling me what to scale, what to cut.
Captures straight from the platforms.
Nothing to hide. The numbers in these images are exactly what you'd see if the founder took you into his account. Verify them anytime, on a call.
Google Ads, January 27 to March 29, 2026
Meta Ads Manager, same window
The numbers, short form.
60 days, January 27 to March 29, 2026. Google Ads plus Meta Ads. Total spent on ads: 37,062 RON.
What we didn't get right.
Not perfect. Three things we'd do differently next time.
- ✕
We pulled 84 products in a single day. Organic traffic dipped 15% for about a week until Google settled on the new structure. Next time we cut in waves of 20-30 products per week, not everything at once. We went for efficient, came out with a bit of noise.
- ✕
Week two, after cleaning up the dedup issue, the dashboard ROAS visually dropped from 3.2 to 2.1. The founder called panicked. Took us two calls and a 12-minute Loom to explain that the drop was the truth surfacing. Next time we prepare that document upfront, not after.
- ✕
We launched Meta too early. Week 5, while Google hadn't fully settled on the new setup. First 10 days on Meta were modest, ROAS between 8 and 12, with creatives that hadn't gathered enough data. Had we waited two more weeks, we'd have walked into Meta with cleaner POAS from day one.
What we took with us into the next projects.
Nothing gets touched in month 1 without an SKU-level margin audit. On this account, that single exercise saved everything. We now repeat it on every client with more than 50 products in catalog. It never gets skipped.
ROAS systematically lies when there's no offline conversion tracking for returns. On the scooter store, return rate was around 8%, enough to make ROAS look a third better than reality. That's why returns tracking is now the very first thing we set up at onboarding, before any campaign.
A small healthy catalog beats a large bloated one. 36 profitable products pulled 2.16 million RON in 60 days. 120 mixed products had been pulling roughly zero net in the months before. Less means more when every euro of budget knows exactly where it's going.
Opening Meta too early eats budget you don't recover. Waiting one full month after Google stabilizes is the difference between clean POAS and two weeks of corrupted data. On the next project we waited exactly 4 weeks before launching Meta. First 30 days of POAS came in twice as clean.
„For eight months the old agency kept telling me „we're optimizing this week." Eight. First week with you lot you told me to pull 84 products from ads and I thought you were joking. Then I looked at the numbers. You were right. Two months, 2.16 million lei from 21 thousand. I'm still wondering why I didn't look at POAS myself sooner."
Client's words: Founder, electric scooter store Sibiu, WhatsApp, March 30, 2026