Machine Learning Scaled HVAC Leads 

Broad match had to earn its place in the media plan before it earned a national rollout. 

  • 2x

    Increase in click-to-lead conversion rate

At a Glance

Client
Mitsubishi Electric Trane HVAC US (METUS)
Engagement
Paid Search (SEM)
Focus
Improve non-brand lead generation efficiency

Overview

We treated it like a question worth proving, not a claim worth believing. 

Mitsubishi Electric Trane HVAC US needed a smarter way to grow residential lead generation without wasting spend.  

Google had just updated broad match, but a promise of more reach wasn’t reason enough to trust it.  

We treated it like a question worth proving, not a claim worth believing. 

The Proof

We validated the approach across four regions before national rollout, making the result explainable and repeatable rather than accidental. The tests also clarified channel behavior: broad match scaled for non-brand search, while exact and phrase match stayed more efficient for brand search. 

  • 29%

    More efficient lead generation

  • 4

    Regions tested before national rollout


The Goal

Mitsubishi Electric Trane HVAC US wanted a more efficient path to lead generation that could support sales goals and its contractor network. 

  • Discovery

    Find new customers earlier in their research stage

  • Efficiency

    Make non-brand lead generation more efficient


The Challenge

The team needed to support sales goals and contractor demand with better lead efficiency. Google’s broad match promised more reach and relevance, but it had to prove it could outperform legacy campaigns in the real world, not just in theory. 


The Solution

We built a regional testing plan for non-brand search keywords using Google’s improved broad match. After strong results in two regions, we ran a second test in two more regions to confirm the pattern and test scalability.

A person typing on a laptop

Once the tests reached statistical significance, broad match rolled out nationally across non-brand campaigns. We also separated the brand search finding clearly: exact and phrase match remained the more efficient choices there.