ServiceTitan moved 26.8 billion dollars of contractor revenue across its platform last quarter and told investors on September 8 that the number grew 17 percent year over year. That growth rate came in roughly 200 basis points under recent quarters. The company named the cause directly: lower job growth by existing customers. Not fewer shops on the platform. Fewer jobs per shop. If your May and June felt soft and July felt closer to normal, you were not imagining it. You were a data point inside a 26.8 billion dollar sample, and that sample is now a public filing anyone can read.
Your Slow Quarter Is Now a Public Filing
ServiceTitan tied the softness to customer lead volume growing at a more moderate seasonal pace through May and June before stabilizing in July, and said the pattern matched HARDI distributor data. That is a demand story, not a software story. Homeowners called less, and when they called, they bought smaller. The mix data says the same thing from the other direction. Repair revenue as a share of home service revenue climbed from 21.6 percent in late 2021 to 31.3 percent in late 2025, and average repair revenue per job rose 47 percent over the same stretch.
Consumer surveys line up with the mix shift. Roughly 64 percent of homeowners say they are limiting spending to necessary repairs and replacements only. About 41 percent report postponing a repair that got more expensive later. Replacement quotes in the 12,000 to 15,000 dollar range are doing exactly what you would expect to a household watching its budget. They are converting a replacement conversation into a repair ticket, and they are stretching the decision across two or three visits instead of one.
Here is the operational read. A softer demand quarter does not punish every shop equally. It punishes the shop that was covering a weak booking rate with rising call volume. When calls grow 20 percent a year, a 55 percent booked rate still produces a growth chart. When calls go flat, that same 55 percent shows up as a bad month, a nervous look at payroll, and a decision to buy more leads. Buying more leads is the wrong first move. The leads you already paid for are cheaper.
The Cheapest Job You Book This Month Is Already Calling You
Run the math with real numbers before you touch a budget. HVAC Local Service Ads averaged about 51 dollars per lead earlier this year with a 44 percent book rate and a 2,110 dollar average ticket. That works out to roughly 116 dollars of ad spend per booked job. Now price the call you never answered. A shop taking 40 calls a day that misses six of them after hours, on lunch, or while the CSR is on another line is walking past roughly 2.6 booked jobs a day at a 44 percent close. At a 2,110 dollar ticket that is more than 5,000 dollars of daily revenue with zero acquisition cost attached.
The fix is not a new platform. It is coverage on the calls you already generate, and pricing has dropped far enough that this is no longer a budget question. Standalone AI answering starts around 49 dollars a month with unlimited minutes at the entry tier and around 79 dollars a month per agent where unique callers are capped. Trade-specific voice agents that triage emergency calls and write the job straight into ServiceTitan or Housecall Pro run 249 to 299 dollars a month. If you are already paying for Housecall Pro, CSR AI is bundled into existing plans. Jobber includes AI quote drafting and AI chat. ServiceTitan has its own AI voice agent and Dispatch Pro.
A five-day install that does not disrupt dispatch:
- Monday. Pull 30 days of call logs from your phone provider. Count answered, missed, and after-hours. Do not estimate. Export the file.
- Tuesday. Pick the single worst bucket. For most 3 to 30 truck shops it is 5pm to 8am weekdays, not weekends.
- Wednesday. Buy the cheapest tool that covers that bucket and writes into the CRM you already own. Do not buy a platform. Buy a phone answer.
- Thursday. Write the triage script yourself. Three questions: is there water or no heat right now, what is the address, when can a tech come. Everything else is a callback.
- Friday. Set one metric in your Monday meeting. Booked jobs from after-hours calls. Track it for four weeks before you judge the tool.
Be honest about the failure mode. These agents book the easy call well and mangle the unusual one. Route anything with a commercial account number, a warranty claim, or an angry customer straight to a human callback queue. A booked job that arrives with bad information costs a truck roll, and a truck roll costs more than the subscription.
ServiceTitan Narrowed Its Roadmap. That Is Operator News.
The same September 8 announcement carried a second item that matters more to a contractor than the revenue line. ServiceTitan said it elected to tighten focus on existing commercial trades rather than the planned expansion into new trades, to fund increased investment in Max, its AI product. Shares fell roughly 30 percent the following day. On September 16 a securities firm announced an investigation into the disclosure. None of that is your problem as an operator. The roadmap sentence is.
Read what it says in plain terms. If your trade is already core to the platform, your features get built and your AI tooling gets funded. If your trade was on the expansion list, your feature requests just moved back in the queue behind an AI product. The September 15 announcement that Canopy Services standardized its roofing brands on ServiceTitan shows where the commitment sits. Roofing at scale is core. That is useful information if you are a roofer with 20 trucks and equally useful if you are a specialty trade that has been waiting two years for a feature.
The action item is not to switch platforms. Switching costs a mid-sized shop six to eight weeks and real money, and ServiceTitan onboarding alone runs 15,000 to 25,000 dollars on top of 245 to 400 per technician per month. The action item is to find out what leaving would cost you before you ever need to know. Open a ticket with your vendor and ask three questions in writing: can you export customer records, job history, and equipment records in full, in what format, and how long does the export take. File the answer. A vendor that answers cleanly has earned another year. A vendor that stalls has told you something.
What to Decide Before October
One decision this month: cover your worst call bucket with a paid answer, and hold it to a single number. Booked jobs from previously missed calls, measured weekly, compared against the subscription cost. If four weeks of data does not clear the cost, cancel it and go back to hiring a part-time CSR. That is a real answer too, and it beats paying for software that books nothing.
The signal to watch over the next 30 days is the next round of distributor and platform commentary on lead volume heading into heating season. If October lead growth stays flat against last year, the soft May and June were not seasonal noise, and the shops that spend the fall fixing booking rate will enter January in much better shape than the shops that spend it buying leads. Price your maintenance agreements accordingly. Recurring revenue is the only line on your P and L that does not care what homeowners decided this quarter.
Atlas Unchained builds the marketing systems and automation that local operators actually run, not the ones that look good in a demo. If you want the call-capture audit above run against your own phone logs, get in touch. The AU Brief lands Friday with the week in six minutes.
