Blended CAC. $290.
From $1,200 baseline. A 76 percent reduction. Sustained across sixteen months of scale, not a one-month dip.
Case study . 01 . HVAC . Ontario . 16 months
Go Lime HVAC.
How we took Go Lime’s blended CAC from $1,200 to $290, scaled ROAS from 1× to 4×, and grew monthly installs from 30 to over 700. Sixteen months. Against Reliance and Enercare.
Live signature
Blended CAC
$290
From $1,200 . Down 76%
Peak ROAS
4.2×
From 1.0× . Up 4×
Monthly installs
700+
From 30 monthly . Up 23×
Spend managed
$200K+/mo
Currently in production
01 . The starting situation
No attribution. Blind decisions.
When the engagement started, Go Lime had no proper attribution infrastructure. Ad spend was running across Google and Meta with no clear view of what was actually driving revenue.
Blended CAC was $1,200. Nobody could explain why. Google Ads had emergency, replacement, and maintenance intent all running in the same campaigns. Meta attribution was client-side only, which meant roughly thirty percent of conversions were invisible to reporting post-iOS 14. Monthly reports showed impressions and CTR. They showed nothing about booked jobs or unit economics.
The decision-making layer was running on data that didn’t connect to the business. That was the first thing that had to change.
02 . The work
Fix the measurement. Then fix the campaigns. In that order.
01
Phase one . Tracking rebuild . Month one
Server-side. Real attribution.
Server-side tracking implementation. GA4 ecommerce configuration with proper data layer. BigQuery pipeline for raw event data. Call tracking connected to campaign, ad group, and keyword level. Blended CAC dashboard live by the end of the month.
Deliv 01Server-side stack
Deliv 02GA4 + BigQuery
Deliv 03CAC dashboard
02
Phase two . Google Ads restructure . Month two
Intent. Segmented.
Separated emergency, replacement, and maintenance intent into isolated campaigns. Built Local Services Ads as a separate system. Restructured Performance Max with service-category asset groups and conversion value rules.
Deliv 04Intent split
Deliv 05LSA system
Deliv 06PMax rebuild
03
Phase three . Meta rebuild . Month three
CAPI. Real audience tiers.
CAPI implementation. Cold, warm, hot audience architecture. Separate creative strategies per tier. Incrementality testing to measure true lift.
Deliv 07CAPI live
Deliv 08Tier creative
Deliv 09Incrementality
04
Phase four . AI automation layer . Months four to six
n8n. Compounding ops.
Reporting automation through n8n. Lead routing into Zoho CRM with full campaign attribution. Competitive intelligence pipeline for Reliance and Enercare ad creative tracking. Creative production automation for ad variants.
Deliv 10Reporting auto
Deliv 11Lead routing
Deliv 12Comp intel
05
Phase five . Optimize against real data . Ongoing
Compounded for sixteen months.
With accurate attribution, every budget decision became sharper. Channels driving booked jobs got more budget. Channels getting credit they hadn’t earned got less. CAC went down. Volume went up.
03 . The results
The biggest lever wasn’t the campaigns. It was the data underneath them.
Peak ROAS. 4.2×.
From 1.0× baseline. Maintained across a period of growing spend, which is the harder test than peak ROAS in a small account.
Monthly installs. 700+.
From 30 monthly at the start of the engagement. Peak month included Black Friday 2025.
Spend managed. $200K+ per month.
Currently in production. Across Google Ads, Meta, programmatic CTV, and OOH.
Organic sessions. 2× year over year.
On a twelve-month base. Driven by the SEO and content velocity layer built during the engagement.
Testimonial . in approval
Quote in approval. Named attribution from Go Lime executive team to be added before launch.
04 . The stack
The tools underneath.
- 01Google Ads. Search, Performance Max, Demand Gen, YouTube, Display, Shopping, Local Services Ads.
- 02Meta. Facebook and Instagram with CAPI.
- 03GA4 and BigQuery for the analytics layer.
- 04Elevar for server-side tracking on Shopify.
- 05Zoho CRM with gclid capture and offline conversion sendback.
- 06n8n for the automation and reporting layer.
- 07Custom dbt models for blended CAC calculation.
05 . The reframe
The situation isn’t unique.
The problem that started this engagement, spending on campaigns without knowing what they were actually costing, is the same problem most operators have. The numbers are different. The structure is the same.
The first call calculates what your version of this number is. Most operators find out it’s different from what they expected.
06 . Next
Find out what your version of this number is.
Book a 30-minute call →Or nabil@novrix.ca.