The numbers: lead sources

Cost per Booked Job and Lead Source ROI: The Arithmetic, the Three Traps, and a Worked Example in Invented Numbers

A lead is a phone call that lasted a minute. A booked job is a truck in a driveway. Most marketing reports stop at the first and price everything by it, which is how a source that rings the phone all day can quietly lose money. Our post on a marketing dashboard built on ServiceTitan data puts two campaigns side by side where cost per lead picks the wrong winner.

November 13, 2026 · 10 min read

dash.acmeplumbingco.com · lead sources, last quarter
Paid search, checked for counting traps
$73 per booked job, once members were taken out
Sample
  • PassExisting customers removed from new demand14 of 96
  • PassTechnician-generated jobs excluded9 jobs
  • FailBooked jobs with no campaign21%
  • WarnManual calls outside marketing analytics11 calls
  • PassRevenue window written on the report60 days
  • WarnMost recent month still filling inmarked
Illustrative: a sample lead source check for a fictional plumbing company

The short answer

Cost per booked job is what you spent on a source in a period divided by the jobs booked from that source's leads in the same period. Lead source ROI is the completed revenue from those jobs minus the spend, divided by the spend. Count new demand only, tie each job to its tracked call, write the window down, and judge a source over two or three months.

On this page
  1. Five ratios that get used as if they were one
  2. One source, start to finish, with made-up numbers
  3. Trap one: billing the ad for a customer you already had
  4. Trap two: the calls that never got a source
  5. Trap three: closing the window too early, or never
  6. No par score, and the clause that keeps it that way
  7. When the spreadsheet gets old

Five ratios that get used as if they were one

Ask three people in a shop what a lead source is worth and you will get cost per lead from the agency, booked jobs from the office and revenue from the owner. Each is a fair number. They answer different questions, and the trouble starts when one is quoted as if it answered all of them.

ServiceTitan defines one of them precisely. Its Marketing Scorecards article, read on the ServiceTitan help center in November 2026, says: "Cost Per Lead: Cost of the campaign divided by the number of call leads," where a call lead is any call "with a call duration over 60 seconds or that was manually classified as a lead." The same article lists Inbound Booked and Sold columns. We did not find a cost per booked job formula there, so the one below is ours, and we label it that way.

Five ratios for judging a lead source, whose definition each one is, and what it actually grades
RatioFormulaWhose definitionWhat it grades
Cost per leadCampaign cost / call leadsServiceTitan, Marketing ScorecardsThe channel's ability to make the phone ring
Booking rateBooked jobs / lead callsOurs for this page; ServiceTitan's versions vary by screenThe office, more than the channel
Cost per booked jobSpend / jobs booked from that source's leadsOursChannel and office together
Revenue per dollarCompleted revenue / spendMatches the first ROI sentence in ServiceTitan's scorecard articleGross pull of the channel, before costs
Revenue ROI(Completed revenue minus spend) / spendMatches the formula in ServiceTitan's scorecard articleReturn beyond the money spent, still before job costs

The last two rows come from one paragraph. ServiceTitan's article describes ROI as "the total revenue divided by the total campaign cost," then calculates it as "( (Revenue - Cost) / Cost) × 100)". Our marketing ROI dashboard page explains which one a tile should carry. The short version: call the first one revenue per dollar, call the second one ROI, and never let a break-even campaign show 100 percent.

One source, start to finish, with made-up numbers

Here is a single paid search source for an invented plumbing company, worked from the invoice back to the ad. Every figure is an illustration chosen to make the arithmetic visible, not a result anyone achieved and not a suggestion of what yours should be.

  1. 1

    Spend: $3,000

    The month's cost for the source, taken from the daily cost entered in campaign settings. If the agency fee is part of what this channel costs, add it here, and say so.

  2. 2

    Lead calls: 96, then 82

    The tracked number took 140 inbound calls. By ServiceTitan's 60-second or Is Lead rule, 96 were leads. Fourteen came from existing customers who searched the company name, leaving 82 leads of new demand.

  3. 3

    Booked jobs: 41

    Jobs whose lead call came from this source's 82 new-demand leads. Booking rate is 41 divided by 82, or 50 percent.

  4. 4

    Revenue: $22,200 from 37 completed jobs

    Four booked jobs canceled or are still open. Revenue counts invoices on completed jobs within the window the company chose, here 60 days from the call.

  5. 5

    The ratios

    Cost per lead on all 96 leads: $31.25. Cost per new-demand lead: $36.59. Cost per booked job: $73.17. Revenue per dollar: $7.40. Revenue ROI: ($22,200 minus $3,000) divided by $3,000, or 640 percent.

Notice how far apart the ratios sit. Quoted alone, $31.25 a lead flatters the source. Quoted alone, 640 percent flatters it more, because none of that revenue has had the cost of parts and labor taken out. The honest summary is a sentence: we paid $73 to put a truck in a new customer's driveway, and those jobs brought in $7.40 for every dollar. Whether that is good depends on margin, which our gross margin per job guide works through.

Trap one: billing the ad for a customer you already had

A member's furnace quits. She cannot find the magnet on the fridge, so she searches the company name, clicks the top result, which is your ad, and calls. Your paid search campaign just took credit for a customer you won three years ago. Repeat it forty times a month and the channel looks far better than it is.

Dark green chalkboard with chalk arrows looping from a doodled house back to itself, a black-and-white halftone cutout of a brass compass in front
Some arrows on the board loop back to customers you already had, and a fair count leaves those out.

ServiceTitan gives you two handles. Its glossary, read November 2026, defines a Marketing Lead as "any job booked through the Call Booking screen that is not a Technician-Generated Lead," and a technician-generated lead as work created "when a technician convinces a customer to schedule another job." Technician-generated jobs belong to nobody's ad budget. The second handle is the customer's history: a caller with a completed job on file is not new demand, whatever number they dialed.

Watch one label in particular. The Marketing Scorecards article says the "New Customer" label in the Calls Generated table "is not connected to the New Customers KPI" on the same screen. Two places, one phrase, possibly two counts. We define new customer once, by the date of the customer's first completed job, and show cost per booked job both ways: all bookings, and new-customer bookings only. When the two drift apart, branded searches are doing the work.

Trap two: the calls that never got a source

The second trap runs the other way. Untracked calls make a source look worse, because its booked jobs leak into a pile with no campaign, and the spend gets divided by too few jobs. ServiceTitan's own help center names where the leaks start.

  • Manual calls. The CSR metrics article, read November 2026, says "manual calls aren't included in marketing analytics." A job booked from a callback the CSR dialed by hand may carry no campaign at all.
  • Short calls. ServiceTitan's call classifications article calls an answered call that ends "in less than 60 seconds" an excused call, and the dashboard metrics article counts a call lead only when it "lasts at least 60 seconds or is marked Is Lead." A brisk call from a real prospect can drop out of cost per lead.
  • Abandoned calls. The glossary says an abandoned call "does not count towards CSR conversion rates or call-related reporting metrics." A missed call from a costly ad simply vanishes from cost per lead.
  • Call reasons left unset. The classifications article warns that without them "any call longer than 60 seconds will be classified as a lead," so the vendor who calls about your copier counts as a lead for whatever campaign owns that number.
  • Forms and bookings typed in by hand. These arrive without the web session that would tie them to a source, as our ServiceTitan marketing attribution guide explains in detail.

The fix is not a cleverer formula. It is a smaller unknown pile. Count the jobs with no campaign every month and show that count beside the ratios. If a fifth of booked jobs have no source, every cost per booked job on the screen could be off by roughly that much, in a direction nobody knows. Tracked numbers per channel are covered in our ServiceTitan call tracking guide, and the office side of the leak in our call center dashboard page.

Trap three: closing the window too early, or never

Every source ROI has an attribution window, written down or not. ServiceTitan writes one into a rule of its own: its Google Ads integration article, read October 2026, counts as leads "Abandoned or excused calls followed by a booking within seven days." Seven days is ServiceTitan's choice for that integration. Your report needs its own, for revenue.

Repairs settle in days. Replacements do not. ServiceTitan's Technician Performance template article gives the example of an estimate built on a job completed January 1 and sold February 1: January's "Close Rate is 0%" while "Close Rate As Of is 100%." The same lag hits lead sources. Judge March's ads on April 1 and you miss the system sold in May from a March call. Our close rate explainer covers the As Of versions.

Then there is the question of which call gets credit. The jobs record in ServiceTitan's API carries a leadCallId, the "ID of the Call that resulted in this job." Read literally, that points at the call that produced the booking. If a customer first rang the postcard number, hung up, and booked a week later from a search ad, the job follows the later call unless someone changes it. That is a reading of one field, not a ServiceTitan policy, but it is worth knowing before blaming the postcards.

No par score, and the clause that keeps it that way

Owners always ask what a good cost per booked job is. We do not print one. None in our research came from a sourced study, and a number without a source is a rumor with a dollar sign. Your own trend, month over month, defined the same way, is the only comparison that holds up.

There is a second reason. ServiceTitan's API Terms of Use, last updated April 15, 2026, bar anyone using the API from deriving "any benchmarks, industry reports," "trend data, statistical insights, or similar derivative works from Content or data obtained through the APIs." We read that as a reason not to build cross-company comparisons from client data, and we do not offer them. Confirm what your own agreement permits.

What you can do on Monday morning is narrower and more useful. Take last quarter, pick one source, and run the five ratios above with the window written down. Then do the same for your second-largest source. The gap between the two will tell you more than any industry chart, and it costs nothing but an afternoon with the reports you already have. If the reports disagree with each other, our guide to reading ServiceTitan reports explains why.

When the spreadsheet gets old

Doing this by hand once is instructive. Doing it every month is a chore that slowly stops happening. A custom dashboard does the counting the same way each time: it reads campaign costs, calls, jobs and invoices through ServiceTitan's API, applies the window and the new-customer rule, and puts the unknown pile on screen. Our custom ServiceTitan dashboards start at $575 to build, depending on complexity, as of November 2026, plus hosting from $55 a month, and the API connection goes live the same day.

If the counting is fine and the meaning is the problem, our monthly numbers review, from $575 a month with no long contract, sits with the owner and walks through what moved. Tracking that has to be fixed first is the work of our conversion tracking setup service. DataCram is a web studio in San Antonio, Texas, building for the web since 1999 (about us) and not affiliated with ServiceTitan; see our work or start at the ServiceTitan dashboards hub. Or call (210) 346-0848 and tell us which source you suspect.

Questions owners ask about ServiceTitan dashboards and numbers

Divide what a source cost in a period by the jobs booked from that source's leads in the same period. Use new-demand leads where you can, leaving out existing customers and technician-generated work. ServiceTitan's Marketing Scorecards article defines cost per lead, not cost per booked job, so write your own definition on the report and keep it.

Cost per booked job, because a lead in ServiceTitan's scorecard is any call over 60 seconds or one marked as a lead, and plenty of those never become work. Cost per lead still helps when booking rates are steady. When they move, cost per lead hides whether the channel or the office changed.

Not in the main number. A member who searches your name and clicks your ad was won years ago. Exclude customers with a completed job on file and technician-generated leads, then show the all-bookings version beside it. A widening gap usually means branded searches are carrying the channel.

Long enough for its jobs to finish. Repairs settle within days; replacements sold from estimates can land weeks after the first call. Pick a window per kind of work, credit revenue to the month of the lead call, and mark recent months incomplete. Two or three months of data is a fairer basis than one.

We do not publish one, because none we found came from a sourced study, and ServiceTitan's API Terms bar deriving benchmarks from data obtained through its APIs. Compare a source with its own past months and with your other sources, using the same definitions each time.

Usually because the agency divides by leads or form fills, and you divide by booked jobs. Untracked calls also push your number up, since booked jobs without a campaign leave spend divided by too few jobs. Count the unknown pile every month and shrink it with tracked numbers.

Keep reading

Free · No obligation

Find out what each source really costs per job

Tell us your sources and how you count them today. A person at DataCram replies with the traps we would check first and what a dashboard that counts it monthly would cost.

Which of these sounds like you? (pick any)

Rather talk now? Call (210) 346-0848.

CallGet a quote