This newsroom publishes method, not hype. Each article explains one part of testing and measuring HVAC marketing, cites the official Google documentation it relies on, and labels every worked example as illustrative. None of them reports invented experiment results or industry benchmarks.
The articles follow the four steps of the Conv.ca method, which Conv.ca names as research and analysis, strategy development, implementation, and testing and refinement. Ali Sedighi founded Conv.ca, and HVAC.Science applies the same sequence to heating and cooling companies. HVAC.Science is a marketing and measurement service; we are not an HVAC contractor and never install or repair equipment.
-
How big and how long: sample size and duration maths for HVAC A/B tests
Before you launch a test, decide the smallest lift worth detecting, then let the formula tell you how many visitors that takes. For most HVAC sites the answer is weeks, not days.
-
Call tracking and attribution: how HVAC phone leads get counted, and miscounted
Most HVAC leads arrive by phone, so attribution rests on how calls are recorded. Separate calls from ads, calls from the website and taps on a number, then pick one attribution model and stick with it.
-
Google Ads experiments and offline conversions: testing on booked jobs, not clicks
A Google Ads experiment is only as smart as the conversion it optimizes for. Pair custom experiments with offline conversion imports so tests are judged on booked HVAC jobs, not on calls or clicks.
-
Holdout tests: measuring whether HVAC ads create jobs or only claim them
Attribution reports tell you which ad was clicked before a call. A holdout tells you whether the call would have happened anyway. For brand search and remarketing, the gap can be large.
-
Reading GA4 for HVAC leads: five settings that change what you see
GA4 rarely lies, but it often answers a slightly different question from the one you asked. Five settings and behaviours explain most of the confusion HVAC owners run into.
-
Reading a conversion test result: use the interval, not the verdict
A test result is a range of plausible effects, not a single winning number. Read the interval, compare it with what a change is worth to your business, and only then decide.
-
Seasonality controls: stopping the weather from winning your tests
In HVAC, the weather moves demand more than any page change. A test is only readable if something in its design cancels out the season: a concurrent control, a demand index or a matched comparison.
-
Testing when your phone rings a few dozen times a month
A shop with 40 calls a month can still learn from experiments. It has to change how it measures and how it decides, not pretend it has the traffic of a national brand.
Planning a test
Start here if you are about to run your first experiment, or if your last one ended in an argument.
- Sample size and duration maths. The two-proportion formula with a worked example, how to turn visitors into calendar weeks, and what Google's Search Central guidance says about test length, redirects and canonical tags. It teaches you why small sites must test bold changes.
- Seasonality controls. Why the weather beats most page changes, how to use a demand index from Search Console, and when Google Ads seasonality adjustments fit. It teaches you to choose test windows on purpose.
Testing with real-world constraints
Most HVAC companies do not have the traffic of a national retailer. These pieces deal with that honestly.
- Testing with few calls. Switchback schedules, a qualified-call metric, decision thresholds set by the cost of being wrong, and a blind call-scoring rubric. It teaches you how to decide when certainty is out of reach.
- Holdout tests. Geographic, time-based and audience holdouts, with difference-in-differences arithmetic and the Google Ads experiment limits to plan for. It teaches you to separate jobs your ads created from jobs they only claimed.
Measuring correctly
A test cannot be better than the data it is judged on. These two articles are about getting the counting right.
- Call tracking and attribution. The three call signals Google Ads documents, the minimum call length setting, dynamic number pitfalls, and GA4's current attribution models. It teaches you to stop double-counting phone leads.
- Reading GA4 for HVAC leads. Key events, data thresholds, the retention setting that limits explorations, attribution settings and the gap with Google Ads. It ends with a 20-minute health check.
Reading and acting on results
- Reading test results with intervals. A worked confidence interval you can check by hand, the difference between statistical and business significance, and a table of four outcomes with the next step for each.
- Google Ads experiments and offline conversions. How custom experiments split traffic, why calls are the wrong finish line, how GCLID-based imports work, and the 2026 changes to how imports are uploaded.
How we write these articles
Our editorial rules are short and strict:
- Official sources only for platform facts. If we describe how Google Analytics or Google Ads behaves, the article links the help page we loaded and lists it with a date.
- No invented numbers. Worked examples are labelled illustrative. We do not publish lift percentages, conversion-rate benchmarks or "typical results".
- Maths you can repeat. Every formula comes with the arithmetic laid out step by step.
- Dated and revisable. Platform documentation changes, so each article shows when it was last updated.
Suggested reading order
| Your situation | Read first | Then |
|---|---|---|
| Never run a test | Sample size maths | Reading results |
| Small market, few calls | Testing with few calls | Seasonality controls |
| Ads spend feels wasted | Holdout tests | Call tracking |
| Reports never agree | Reading GA4 | Call tracking |
| Leads up, jobs flat | Google Ads experiments | Reading results |
When you want the method applied to your own account, see our services, the plans on pricing, or the HVAC results on results. Call, text or email and you will speak with a senior strategist.
Frequently asked questions
Are the numbers in these articles real client results?
No. Every worked example is labelled as illustrative and exists to show the arithmetic. Real results from our HVAC work, with client names and links, are on the results page.
How often are the articles updated?
When the documentation they rely on changes. Each article lists its sources with the date we loaded them, so you can see how current it is and check the original yourself.
Do I need a statistics background to use them?
No. Each formula is shown with a worked example you can repeat on a calculator, and every article ends with a checklist you can use without the maths.