Conversion research finds where your website loses HVAC leads and why, before anyone changes a page. We combine analytics, call listening and a page-by-page review on a phone, then rank the problems by evidence. The deliverable is a short list of hypotheses worth testing, not a long list of opinions.
This is the research-and-analysis step of the Conv.ca method that HVAC.Science follows. It feeds directly into CRO experiments, and it is usually the first piece of work for a company whose measurement is already in reasonable shape. If it is not, start with measurement setup.
Who it is for
Research suits HVAC and heating companies that already get steady traffic but sense that too few visitors call. Typical signs:
- ad spend is stable but calls per week have drifted down
- a new website launched and the phone got quieter
- one service page (often replacements) gets visits but almost no quote requests
- your team keeps debating design changes with no data to settle them
It is less useful for a brand-new site with almost no traffic, where the first job is getting found.
What is included
- Analytics review. Landing pages, devices and channels, read through the lens of GA4's key events and the settings explained in our GA4 article.
- Path analysis. Where visitors go before calling, and where they go instead of calling.
- Call listening. A scored sample of recorded calls, if you have them, using a blind rubric.
- Mobile page review. Every key page checked on a phone: how fast the number is reachable, what the first screen promises, what objections go unanswered.
- Message match. Do your ads, Google Business Profile and landing pages say the same thing?
- Competitor read. What homeowners see from two or three competitors for the same searches, described factually.
- Hypothesis list. Each problem stated as "If we change X for audience Y, metric Z should move, because evidence W."
Timeline
A typical engagement runs in three phases. Your proposal confirms the actual dates.
- Access and baseline. You grant read access to analytics, ads and call tracking; we pull at least four full weeks of baseline data.
- Analysis. Listening, page review and path analysis run in parallel.
- Readout. A working session to go through the ranked list and agree which hypotheses move to testing.
Research during a shoulder season is fine, but we note the season in every finding, because demand mix changes as the weather turns.
How results are measured
Research does not lift conversions by itself, so we do not claim it does. We judge it on three things:
- Evidence quality. Every ranked item cites the data behind it.
- Testability. Each top hypothesis has a metric and a realistic sample size, using the maths in our sample size article.
- Hit rate over time. Across later tests, how many research-led hypotheses produced a readable result. That figure comes from your own tests, not from a benchmark.
Mistakes to avoid
- Starting with a redesign. A full redesign changes dozens of things at once, so you can never tell which helped or hurt.
- Trusting heatmaps alone. They show where people tap, not why they did not call.
- Ignoring the phone. On an HVAC site the main conversion happens off the page. If you never listen to calls, you are researching half the funnel.
- Asking only the team. Owners and technicians know the work, but homeowners decide what the page must say. Their words, from calls and reviews, are better evidence.
- Ranking by effort instead of evidence. Easy fixes feel productive. Rank by likely impact first, then by effort.
What happens next
Most companies move from research straight into a small program of CRO experiments. Some discover that ad spend is the bigger question and go to holdout testing. Plans and prices are on pricing. Call, text or email to start with a short conversation about what you have already tried.
Frequently asked questions
Isn't this just a website audit?
A generic audit lists everything that could be better. Research ranks problems by evidence and by how many leads they plausibly cost, and turns the top ones into hypotheses you can test. Many audit items never make the list.
Do you need access to our call recordings?
It helps a great deal, but it is optional. If you record calls with customer consent, a scored sample tells us what callers ask about and what they mention from the site. Without recordings we rely more on analytics and page review.
What if the research says our site is fine?
Then we say so, and the problem is usually upstream: not enough of the right traffic, or leads lost after the call. That is still a useful answer, because it stops you paying for a redesign.