Most heating company owners know something is wrong with their marketing before they can say what. Calls feel lower than last year, the ad report looks fine, reviews are thin, and nobody can explain the gap. This guide is a structured way to find out. It takes about 90 minutes, needs only access to your own accounts and last month's job list, and produces a scorecard you can act on whether or not you ever hire anyone.
Before you start
Gather four things:
- A list of last month's sold replacement jobs, with customer names and sale values.
- Logins (or a staff member with logins) for your website, call tracking if you have it, CRM, Google Ads, Meta and Google Business Profile.
- A phone that is not your office line, for test calls.
- A blank sheet with three columns: test, result, notes.
Work through the tests in order. They follow the direction data should travel through the stack, from capture to proof, so an early failure often explains a later one.
Test 1: the after-hours call (capture)
Call your own advertised number from your website at 6:30 p.m. on a weekday. Then text it.
- Pass: a person answers, or you receive a useful text within a few minutes and a callback the next morning.
- Fail: voicemail with no follow-up, or a text that goes nowhere.
In heating season, after-hours and lunch-hour calls are often urgent no-heat calls. A caller who hears voicemail usually tries the next company in the results.
Test 2: the source check (capture to record)
Open the CRM records for your ten most recent new enquiries. For each one, note whether the lead source is filled in, and whether anyone had to type it.
- Pass: eight or more have a source, filled in automatically.
- Fail: sources missing, all set to "Google", or typed by hand inconsistently.
If this test fails, every later report about which marketing works is guesswork.
Test 3: the job trace (record)
Take three sold replacements from last month. For each, try to answer from your systems alone: how did this customer first find us, how quickly did we respond, how many days from quote to sale?
- Pass: you can answer all three for each job in a few minutes.
- Fail: you have to ask a staff member to remember.
Memory is not a system. In a busy season it's the first thing to fail.
Test 4: the stale-quote count (record)
In the CRM, count replacement quotes sent more than 14 days ago that are neither won nor marked lost.
- Pass: only a handful, each with a follow-up task assigned.
- Fail: dozens, with no next step.
Each open quote is a homeowner who was interested enough to let you into their house. Multiply the count by your close rate and average sale to see what is sitting there.
Test 5: the conversion definition (record to demand)
In Google Ads, open the list of conversion actions and see what is counted as a conversion. Do the same in Meta.
- Pass: conversions include real pipeline stages imported from the CRM, such as "booked" or "sold", ideally with values.
- Fail: only calls over a set length, page views or button clicks.
If this test fails, the platforms are learning from activity rather than revenue. The guide to how the stack connects explains this wire in detail.
Test 6: the search-terms read (demand)
In Google Ads, open the search terms report for the last 30 days and read the first 50 terms by spend.
- Pass: almost all are searches from people who might buy what you sell, in places you serve.
- Fail: parts, manuals, DIY, jobs, wholesale, other companies' brand names, or towns you don't cover.
Add the bad ones as negatives today. It is the cheapest fix in the whole check.
Test 7: the review ratio (record to proof)
Count jobs completed last month and Google reviews received last month.
- Pass: a steady share of completed jobs produces a review, and the requests go out automatically.
- Fail: reviews arrive in bursts after someone remembers to ask, or hardly at all.
Also check: are recent reviews answered? Unanswered complaints are visible to every future customer.
Test 8: the display check (proof to capture)
Open your furnace, boiler and heat pump pages on a phone. Is there a recent review about that kind of job on each page? Is the phone number tappable at the top? Does the page load quickly on mobile data?
- Pass: yes to all three, on every main service page.
- Fail: reviews only on the homepage, or none at all.
Reading your scorecard
Map each failed test to the layer and service that fixes it.
| Failed test | Layer | Where to start |
|---|---|---|
| 1, 8 | Capture | Heating websites |
| 2, 3, 4 | Record | CRM integration |
| 5, 6 | Demand | Paid ads |
| 7 | Proof | Review generation |
If you failed tests in three or more layers, a growth stack blueprint will save you from fixing them in the wrong order.
Common patterns we see
A few combinations come up again and again in heating companies.
- Fails 2 and 5, passes everything else. The office is good and the ads are busy, but nobody can prove what works. Ad budgets drift toward cheap calls. Fix wires one and two before raising spend.
- Fails 1 and 4. Leads arrive and then evaporate. More ad spend would only make it worse. Fix the record layer first.
- Passes 1 to 6, fails 7 and 8. The machine runs but doesn't compound. A review trigger and better page display often lift conversion without extra spend.
- Fails 6 only. Easy win. A weekly search-terms review usually settles it.
Putting a number on it
Once you know where the leaks are, the stack math guide shows how to price each one in sold jobs and gross profit, using your own figures. That turns a list of problems into a ranked list of fixes.
If you'd like a second pair of eyes on your scorecard, call, text or email with your results. We'll tell you which fix we'd make first, and whether you need help to make it.
Frequently asked questions
When should I run this check?
Late summer or early autumn, before the heating rush, gives you time to fix what you find. Running it in January is still worthwhile, because that is when leaks cost the most, but you will have less time to act.
What if I fail most of the tests?
That is common and not a disaster. Start with the first failed test in the order given, because the earlier layers feed the later ones. Fixing tracking and follow-up usually comes before spending more on ads.
Do I need technical skills to run these tests?
No. You need logins to your accounts, a phone and a spreadsheet. Where a test involves a setting you can't find, note it as unknown; that is itself a finding.