Most advice on this topic assumes your leads arrive through a browser. For a local service business that assumption is wrong, and it is why so much attribution software disappoints.
Referrals and repeat customers each account for around 59% of leads for home service businesses. Neither is visible to any analytics tool ever built. A neighbour recommending you over a fence leaves no data trail.
Which produces an unfashionable conclusion: the highest-value attribution tool available to you is a question, asked by a person.
The short answer
Ask every enquiry how they found you, write down exactly what they said, and give every lead an outcome. Do that in one place for two months and you will know more about your marketing than most competitors know about theirs. Software helps later; it does not substitute for the question.
Why accounting software cannot do this
Worth addressing early, because it is where most businesses look first.
QuickBooks and its equivalents record people who paid you. Every enquiry that never converted was never entered, which means you have the numerator of your close rate and no denominator at all.
You can add a custom field for lead source, and people do. But the tag only ever lands on customers, so a source producing forty enquiries that convert twice looks identical to one producing two enquiries that convert twice. The full reasoning is in is QuickBooks a CRM.
This is not a configuration problem. The records do not exist, and no report can create them retroactively.
The seven fields
One row per enquiry. That is the entire system.
| Field | Why |
|---|---|
| Date and time received | Reveals after-hours volume |
| Name and contact | Obvious |
| Source, verbatim | The point of the exercise |
| Job type | Sources produce different work |
| First-contact time | The most predictive field you will capture |
| Stage | New, quoted, won, lost |
| Outcome and reason | A few words |
Two rules decide whether this works:
Every enquiry goes in, including the ones that feel obviously dead. Those rows are the denominator, and skipping them is the most common way this exercise produces flattering nonsense.
Every row gets an outcome. A lead left unresolved quietly drops out of your count and inflates your close rate. Unfinished rows are the single biggest source of bad data here.
Asking the question properly
Wording matters more than it should.
Ask at intake, in the first minute, before the conversation moves to the job. Asking later means guessing later.
Record what they said, not what you concluded. "My neighbour Dave on Willow Street" is more useful than "referral," because it tells you referrals are coming from a specific job you did.
Always follow up on "Google." It means at least three different things with completely different economics: the map pack, a paid ad, or an organic result. One extra question, "was that the map, an ad, or did you find our website," captures most of the value.
Do not offer a list. Reading options out loud makes people pick the first plausible one. An open question produces a truer answer.
Watch out
The most common failure is asking at the end of the job instead of the start. By then the customer has forgotten, and they will say "Google" because it is the easy answer. Attribution collected a week late is close to worthless.
Where software genuinely helps
Three places, in order of return.
Call tracking numbers per channel. A distinct number on the website, another on the van, another in ads. This solves the "Google" problem properly and it fixes the most common blind spot, which is that a website call is indistinguishable from every other call when both use your main line. If you spend anything on advertising, this pays for itself quickly.
UTM parameters on links in ads, email and social. Free, and they tell you which campaign produced a form fill.
A CRM, once volume outgrows a spreadsheet. The trigger is usually a second person needing to see the data during a busy week, or enquiries falling through gaps.
What software cannot do is capture the offline majority. A tool that shows you your online sources beautifully while missing the 59% arriving by referral will point your budget in the wrong direction with great confidence.
Attribution only matters if enquiries are arriving in the first place. Our free check tests whether a visitor can become a lead on your site at all, whether anything catches an out-of-hours enquiry, and where the submission goes. Twenty checks, about fifteen seconds.
What to calculate
Four numbers per source, monthly. The order matters, because most businesses stop after the first.
Leads. How many arrived.
Close rate. Jobs won divided by leads. This varies far more between sources than anything else and it is the number that changes decisions.
Cost per booked job. Spend divided by jobs won.
Revenue per customer. An installation customer worth $10,000 and a tune-up customer worth $150 should not share a budget rule.
The reason to insist on the third rather than cost per lead: a $149 lead closing at 30% costs $497 per customer, while a $72 lead closing at 10% costs $720. The cheaper lead was 45% more expensive to convert, and any dashboard reporting cost per lead would have shown it winning. The full picture is in what HVAC leads actually cost.
Typical close rates by source, useful as a sanity check on your own:
| Source | Close rate |
|---|---|
| Referrals | 50%+ |
| Organic search | 25–40% |
| Local Services Ads | 20–35% |
| Paid search | 15–30% |
| Social | 10–20% |
| Shared marketplace leads | 8–15% |
Getting the team to actually do it
The system is trivial. Getting it maintained during a busy week is the entire difficulty, and it is where most attempts die.
Make it faster than not doing it. If logging an enquiry takes ninety seconds, it will not survive July. A sheet open on a pinned tab, or a form on a phone, takes fifteen. Whatever you build, time it yourself under pressure before rolling it out.
One person owns it. Shared responsibility becomes nobody's responsibility on the first busy day. Name someone, and put it in their role rather than in a group message.
Explain what it is for, once, honestly. "So we stop spending money on the ads that do not work" gets cooperation. Presenting it as record-keeping gets compliance for a fortnight and then nothing.
Review it visibly. If the sheet is filled in every week and nobody ever looks at it, people correctly conclude it does not matter. A five-minute monthly review where you actually change something based on what it shows is what keeps it alive.
Accept imperfect data. A sheet that is 80% complete is enormously more useful than no sheet, and holding out for completeness is how businesses end up with nothing. Fill the gaps you can, mark the ones you cannot, and move on.
The failure mode worth naming: businesses buy a CRM to solve this, discover nobody enters anything into it, and conclude the problem was the software. It was almost never the software. It was that entering data was slower than the way people already worked, and no tool fixes that on its own.
What the data usually reveals
Three findings recur often enough to predict, and each changes where money goes.
Referrals are producing more revenue than anyone realised, and nothing in the business is deliberately generating them. They are treated as weather rather than as a channel. Once the number is visible, asking for referrals at the point of relief, the day of service, becomes an obvious priority rather than an afterthought.
One paid channel is quietly subsidising another. The source producing the most leads is frequently not the one producing the most revenue, because close rates differ so sharply. This is invisible while you track lead counts and obvious within a month of tracking outcomes.
A meaningful share of enquiries has no recorded outcome at all. Not lost, not won, simply unresolved. Those are leads nobody ever got back to, and finding them is often worth more immediately than anything else on this page, because several are still recoverable with a phone call this week.
That third one tends to be uncomfortable and it is the best argument for starting. You cannot chase a lead you do not know exists, and until every enquiry is written down in one place, some number of them are being forgotten every month without anyone being able to say how many.
What people get wrong
Only tracking online. The majority of your work arrives through channels no pixel sees.
Tracking leads, not outcomes. Lead counts without close rates tell you nothing actionable.
One company-wide close rate. Averaging referrals at 50% with marketplace leads at 12% produces a number that describes neither and hides the whole point.
Comparing months in a seasonal trade. A shoulder month against a peak month is not a comparison. October is the busiest month in HVAC by operational load, so year-on-year comparisons are the only honest ones, which is part of the argument in the best time of year to advertise HVAC.
Buying software before establishing the habit. A CRM nobody enters data into produces worse reporting than a spreadsheet somebody maintains. Build the habit on paper, learn which fields you actually use, then specify software around your real workflow.
Attributing to the last touch. Someone who saw your van for a year, read a review, then searched your name is not a branded search lead. Recording what they said verbatim preserves this nuance that a dropdown destroys.
One number worth watching above the rest
If you track nothing else, track median time to first contact, split by source.
It is the most predictive single field you will capture, it is entirely within your control, and it explains more variation in close rate than most businesses expect. Sources that appear to convert badly frequently turn out to be sources you answer slowly, which is a completely different problem with a much cheaper fix than replacing the channel.
The trap it protects you from is real: a business concludes that marketplace leads are poor quality and cancels them, when the actual cause was a two-hour response time on enquiries that needed ten minutes. Without response time recorded alongside source, those two explanations are indistinguishable in the data, and the wrong one is the more flattering.
Record the minutes, use the median rather than the average, and review it monthly alongside close rate.
Starting this week
Monday. Create the sheet. Seven columns. Share it with whoever answers the phone.
Tuesday. Start asking, at intake, and recording verbatim. Expect it to feel awkward for two days.
Wednesday. Put a distinct call tracking number on the website, or at minimum note which enquiries arrived through the site.
Thursday. Go back over the last month from memory and add what you can. Imperfect, still useful for a baseline.
Friday. Add outcomes to everything currently open.
Month two. Calculate close rate by source. This is usually the point where something surprising appears, and it is normally that the source producing the most leads is not the one producing the most revenue.
Quarterly. Move budget on cost per booked job.
The habit is the hard part and the software is not. Most contractors have never measured close rate by source, which means every downstream number is a guess, and the guess is reliably optimistic.
Once you have the data, it makes every other fix measurable: whether the missed-call automation is working, whether faster response actually converted, whether the follow-up cadence changed anything. All of those are in how many leads is my business losing, and none of them can be proven without this.