Lead Source Tracking: Which Marketing Actually Books Jobs in 2026

TL;DR

Ask a service business owner what they spend on marketing and you get an exact figure. Ask what it returns and you get a story. That gap is not laziness — it is that the return arrives as a phone call, gets written down as a job, and the connection between the two is never recorded by anyone.

As of September 2026 you do not need an attribution platform to close it. You need one required field at booking, source attached to the job rather than the enquiry, answer rate measured before channels, a rolling window rather than calendar months, and a monthly review short enough that you will still be doing it in November. This guide walks through each, shows the arithmetic on a stated assumption rather than quoting statistics at you, and is honest about which numbers are worth trusting.

The honest starting position

Here is where most owners actually are.

You know your spend precisely, because it leaves the bank account every month with a name on it. You have a strong instinct about which channel works, formed mostly from the last three memorable customers. You have never compared booked revenue by source, because the data does not exist in a form that permits comparison. So spend decisions get made on which channel most recently produced a story you remember.

That instinct is not worthless, but it is biased toward channels that produce memorable customers rather than profitable ones. The large job from a directory listing is memorable; forty small jobs from repeat customers who found you through reviews two years ago are not.

The goal is modest: not perfect attribution, which is unavailable to anyone at any budget, but knowing within a month or two which two or three sources produce the work. That is enough to make better spending decisions, using data your scheduling software already holds.

The cheapest useful version: one required field

Start here, because it costs nothing and works within a month.

Add a source field to your booking flow and make it required. Whoever answers the phone asks in plain words — "how did you find us?" — and picks from a short list. If you have online booking, the same field appears there as a required dropdown.

Three rules make it work:

• Keep the list to five or six options. Long lists produce guessing, and guessing produces data worse than none because it looks credible. Something like: referred by someone, found us online, saw our vehicle or sign, existing customer, repeat from a previous job, other.
• Make it required, not optional. An optional field gets completed on quiet Tuesdays and skipped for a whole busy week, which biases your data toward what quiet weeks look like.
• Attach it to the customer record, not just the enquiry. When that customer returns in eight months you want to know they originally arrived from a review, because the lifetime value of a source is the number you care about.

The obvious objection is that people misremember, and they do: someone who saw your vehicle, then searched your name, then read reviews will report whichever step they recall. The field is therefore imperfect per customer and useful in aggregate. If sixty per cent of a month's bookings say referral, that is true enough to act on even if some are misattributed at the margins. Your customer database is where the field earns its keep, because it is the only place the source and the eventual revenue sit on the same record.

Calls are where the whole thing leaks

Before comparing any channel, measure how many calls you answer.

A missed call does not become a lead: it never enters the CRM, is never attributed to a source, and appears in no report you run. It is an enquiry your marketing paid for that vanished without trace, and the analysis then gets run on the calls that happened to get through.

Work the arithmetic on a stated assumption; substitute your own numbers. Suppose you receive 150 inbound calls in a month, answer 120 and miss 30. Suppose one in three answered calls becomes a booked job, so 40 jobs. If your average completed job is worth 300 dollars — use your own figure, this one only carries the arithmetic — that is 12,000 dollars of booked revenue from 120 answered calls, or 100 dollars of booked revenue per answered call.

Those 30 missed calls, at the same conversion, represent 10 jobs and 3,000 dollars. Not all are recoverable, but even half is a larger sum than most owners' monthly marketing budget, and none of it appears in any channel report.

So the order of operations matters: measure answer rate by hour of day, close the gap because it is the cheapest revenue available to you, then measure which channels produce the calls.

Closing the gap has two routes. An automatic text back to every missed call converts a portion into conversations, worked through in missed call text back. Or you answer the calls you currently miss, which for most small businesses means the after-hours and during-the-job calls nobody is available for — the honest comparison is in AI receptionist versus human receptionist, with the cost side in how much an AI answering service costs. Either way, an AI receptionist that asks the source question as part of intake solves the attribution problem and the answer-rate problem in the same step.

Call tracking numbers are genuinely good at one thing and useless at another. A distinct number per channel — one on the vehicle wrap, one on a directory listing, one on a campaign — tells you reliably which channel produced the ring, without relying on anyone's memory. What it cannot tell you is what happened next — a number that rings 90 times a month can look like your best channel while producing nothing but price shoppers. Tracking numbers count calls; only the job record counts money. Use both, and never rank channels on call volume alone.

Attribute booked and completed jobs, not leads

This change reorders the ranking, usually dramatically.

Lead-level reporting counts enquiries; job-level reporting counts work. They produce different winners because channels differ enormously in the quality of the enquiry, not just the quantity.

A price-comparison channel delivers a high volume of enquiries whose first question is the price and whose second is whether you can beat someone. A referral delivers fewer enquiries where the caller has already decided to use you and is only settling the timing. Rank those by lead count and the comparison site wins; rank them by completed revenue and it is not close.

So carry source through three stages. Enquiry — source recorded at booking, on the customer record. Booked — did it become scheduled work; this ratio is your source quality measure and it separates volume from value. Completed and paid — did the work happen and get invoiced, because cancellations are not revenue, and a source that books well but completes poorly is telling you about the customers it sends.

With source on the job record, reporting can answer the only question that matters: booked revenue by source, over a period, against what you spent on that source. Cost per booked job, not cost per lead. Everything else in marketing measurement is a proxy for that one figure.

Sources, capture and what each number is worth

Not all attribution data is equally trustworthy, and treating it as though it were is how businesses make confident wrong decisions. This is the honest ranking.

Two rows deserve a pause. Repeat business is the most trustworthy figure you own and almost nobody reports it, so the cheapest revenue in the business gets no management attention. And every paid channel looks better on call volume than on booked revenue, which is precisely why paid channels get judged on the wrong metric.

The lag problem

Marketing spend and the revenue it produces do not occur in the same calendar month, and month-by-month comparison quietly ruins good analysis.

Three lags stack up. Someone becomes aware of you and does not need you yet — a year for a roof, a day for a blocked drain. An enquiry becomes an estimate, and the estimate closes weeks later. And a completed job produces a referral months later, recorded as a referral rather than a return on the spend that created the customer.

The corrections are unglamorous. Record the enquiry date separately from the completion date on the job; without both you cannot see the lag, let alone correct for it. Compare rolling two or three month windows rather than calendar months, because a strict monthly view credits and blames the wrong periods, and it does this most severely in the months just after you change spend — which is exactly when you are looking. Give a change at least a quarter before judging it, since cutting a channel after four weeks is measuring noise.

Quote-based work has its own version of this: the estimate that closes in six weeks looks like a dead lead at week three. Follow-up is where that revenue actually lives, which is the subject of turning estimates into booked jobs.

Reviews and word of mouth: the source you cannot buy

The largest source in most small service businesses is other people's recommendations, and it is systematically under-managed because there is no invoice to look at.

Two things follow.

First, split it at booking. "Referred by someone" and "found you online through reviews" are different mechanisms with different levers, and lumping them into word of mouth hides which one is growing. A referral comes from someone who used you; a review-driven enquiry comes from a stranger who trusted a stack of ratings. You influence the first by asking customers and the second by requesting reviews after completed work.

Second, once it is a number it becomes manageable. If referral and review sources produce most of your booked revenue, a review request after every completed job is not a nicety — it is your highest-return marketing activity, and it costs a text message. The mechanics of doing that without being irritating are in getting more Google reviews.

Reviews are also largely produced by the operational things this blog spends its time on — arriving in the window, documenting the work, invoicing clearly — which means your best marketing channel is mostly downstream of your scheduling, not your advertising.

A monthly review you will actually do

Attribution fails more often from abandonment than bad data, so the review has to be short enough to survive.

Once a month, thirty minutes, four numbers per source: enquiries received, booked jobs that became scheduled work, completed revenue actually invoiced and paid, and spend including your own time if it is meaningful.

Then divide: completed revenue by spend, and spend by booked jobs. Those two ratios are the whole analysis. Write them in the same place every month so you have a series rather than a snapshot — one month tells you almost nothing and six tell you nearly everything.

Two disciplines make it stick. Look at the rolling three-month figure as well as the month, so you are not reacting to a fortnight. And change one thing at a time, or the following month is uninterpretable.

What you are looking for is not precision. It is the two sources clearly working, the one clearly not, and the ones too small to matter yet — a decision-quality answer out of a required field and half an hour.

What this costs

GetTimePad is $79/mo Starter for one staff member, which includes the customer database, online booking portal and basic reporting — enough to run the required-source-field version of this. $199/mo Pro for up to five staff adds the two-way SMS inbox, SMS and email automations, Stripe payments and deposits, and the AI receptionist, which is the tier where the answer-rate problem and the attribution problem get solved together. $499/mo Agency at unlimited staff adds multi-location plus the REST API and outbound webhooks, if you want source and revenue data flowing into something else.

Extra seats are $25/mo per seat on Starter and Pro, the SMS bundle is $19/mo, and annual billing gives you two months free. All plans include a 14-day free trial and there is a live demo with no signup required, set up in under five minutes. Full detail is on pricing.

Where to start

Add a required source field with five options to your booking flow this week and ask the question on every call. Do nothing else for a month.

Then, before you compare a single channel, count how many calls you received against how many you answered. If there is a gap, it is worth more than any reallocation of spend you were considering, and it is the first thing to fix. Only after that does channel comparison earn your attention — and when it does, rank on completed revenue per dollar spent, not on how many times the phone rang.