SaaS / B2B
-41% CPL, +18% MRR contribution
Trades & Home Services
3.8x ROAS in 12 weeks
Atlas Trade Services — 3.8x ROAS via Google Ads
The challenge
Atlas Trade Services had been running Google Ads for two years. The account had grown organically — campaigns bolted on as new service lines launched, match types never reviewed, bids adjusted reactively when the phone went quiet. By the time they came to us, monthly spend sat at around $18,000 with a blended ROAS of 1.4×. At that return, they were barely covering the cost of the channel, with no room to scale.
The previous agency's reporting told a flattering story: high impression share, strong click volumes, plenty of activity. What it didn't show was where those clicks were actually coming from. A quick search term audit in week one turned up thousands of dollars in spend on queries for suburbs 40 kilometres outside Atlas's service area, competitor brand terms they had no business showing for, and informational searches from people clearly not in-market to book a tradie.
Compounding this was a conversion tracking problem. The account was counting visits to the contact page as conversions — not form submissions, not phone calls, just page loads. Google's smart bidding algorithms had spent two years optimising toward a goal that had no relationship to actual booked jobs. The PMax campaign, consuming 40% of budget, had no audience signals, no asset exclusions, and was serving display inventory across publisher categories completely unrelated to home services.
---
What we did
The first two weeks were purely diagnostic. Before touching any live settings, we pulled the full search term report — 90 days, all campaigns — and ran it against Atlas's actual service list and postcodes. The waste was significant but quantifiable: roughly a third of spend was going to queries with zero plausible path to a booked job.
Conversion tracking came first, because everything downstream depends on it. We rebuilt the tag architecture from scratch: GA4 with enhanced conversions feeding into a linked Google Ads account, primary conversion action set to verified form submissions (confirmed via Thank You page URL), secondary conversion action set to phone call clicks with a 60-second minimum duration threshold. We ran both in parallel for two weeks to establish a clean baseline before touching bid strategies.
With clean data flowing, we restructured the search campaigns. The existing setup used a mix of broad match and legacy ad groups that hadn't been touched since 2023. We consolidated into tightly themed ad groups organised by service line and intent stage — commercial intent terms (book, hire, cost, quote) separated from informational terms (how to, what is, DIY). Informational terms were paused or negated; they generate clicks from people who aren't buying. We added 400+ negative keywords in week one alone, covering competitor names, wrong locations, and non-commercial query patterns.
The Performance Max campaign was the bigger structural problem. Without audience signals, PMax defaults to broad prospecting — expensive and slow to converge for a local trades business with a defined service area. We rebuilt the campaign from scratch: one asset group per major service line, audience signals built from a CRM upload of past booked customers, URL exclusions to prevent serving on blog content, and a geographic radius tight enough to exclude postcodes Atlas doesn't service.
Bid strategy was the final lever. The account had been on Target CPA — but the CPA target had been set using the old, inaccurate conversion data. We switched to Maximise Conversions for the first 30 days to let the algorithm recalibrate against clean data, then introduced a tCPA floor based on the actual cost-per-booked-job from the clean period.
---
The results
The account took about four weeks to fully stabilise after the conversion tracking rebuild — that's normal when you've made significant structural changes and Google's systems are relearning. Atlas held their nerve. By week eight the numbers were moving in the right direction. By week twelve we had the full picture.
The headline result was ROAS moving from 1.4× to 3.8× on equivalent spend. Cost per booked job dropped 54%. Critically, Atlas was confident enough in the efficiency gains to increase monthly spend by 31%, and the ROAS held. That's the real test: can you scale without the return degrading?
The 400+ negative keywords added in week one had immediate impact on impression quality. Wasted spend dropped sharply in the first two weeks just from that change — before the restructure and bidding changes had fully kicked in.
---
> "I'd been told my Google Ads were performing well — the agency kept showing me impression share numbers. When Empire Amplify ran the audit, we were spending about a third of our budget on suburbs we don't even service. Twelve weeks later our ROAS had gone from 1.4× to 3.8×. The phone rings with better jobs now. That's the difference." > > — Marcus T., Director, Atlas Trade Services ---
Ready to find out what your account is actually doing?
We audit accounts every week. If yours has similar symptoms, we’ll find it.
