Last updated: September 12, 2026
What a CSR scorecard is for
CSR means customer service representative here, the person who answers the phone at an HVAC, plumbing, or electrical company and turns that call into a scheduled job. The scorecard below runs to 100 points across seven sections, from the greeting through the arrival window, and every scored call carries the booking outcome next to its total.
Owners usually go looking for one because of a spread. Two people answer the same phone off the same ad spend, one books eight jobs out of ten calls and the other books five, and nobody can say out loud what the first one does differently. Scoring the recordings is how that question gets answered.
The 100 point CSR scorecard
Twenty three criteria across seven sections. The weights follow what a miss costs in booked work rather than what is easiest to hear on a recording, which is why the appointment section carries almost a quarter of the card by itself. Section totals are in parentheses.
| Section | Criterion | Points |
|---|---|---|
| Greeting and contact capture (16) | Answered with the company name as the greeting standard writes it | 5 |
| Greeting and contact capture | Took the caller's first and last name and used it again later in the call | 4 |
| Greeting and contact capture | Took a callback number and read it back digit by digit | 7 |
| Urgency and safety triage (14) | Asked the triage questions on the company's intake sheet for that trade before moving to scheduling | 6 |
| Urgency and safety triage | Classified the call against the company's own emergency definitions | 4 |
| Urgency and safety triage | Followed the company's documented escalation procedure when the call met one of those conditions | 4 |
| Problem captured (14) | Recorded what the caller is experiencing in the caller's own words rather than a category alone | 5 |
| Problem captured | Captured the equipment or fixture involved, with make and approximate age where the caller knows them | 4 |
| Problem captured | Asked whether anyone has been out for this before and pulled the prior job | 3 |
| Problem captured | Noted what changes the truck load or the tech assignment, such as a crawl space or a rooftop unit | 2 |
| Address and access verified (12) | Read the full service address back and confirmed the city and zip | 5 |
| Address and access verified | Captured access details, including gate code, parking, pets, and which door the tech should use | 4 |
| Address and access verified | Checked the address against the service area before offering a time | 3 |
| Appointment offered and window booked (24) | Asked for the booking once the problem was captured, without waiting for the caller to ask | 8 |
| Appointment offered and window booked | Offered two specific arrival windows by day and time rather than asking when the caller is free | 8 |
| Appointment offered and window booked | Entered the job in the scheduling system while the caller was still on the line | 5 |
| Appointment offered and window booked | Offered a third option or a callback slot when both windows were declined, before ending the call | 3 |
| Fees and plan explained (10) | Stated the diagnostic or trip fee exactly as the current price list has it | 5 |
| Fees and plan explained | Answered the cost question with the company's approved wording and gave no repair price sight unseen | 3 |
| Fees and plan explained | Offered the membership or service plan where company policy calls for it | 2 |
| Close and confirmation (10) | Confirmed the date, the arrival window, the address, and the fee the caller will owe | 5 |
| Close and confirmation | Told the caller what happens next and when they will hear from someone | 3 |
| Close and confirmation | Asked whether the caller needed anything else and closed with the company's approved wording | 2 |
| Total | 23 criteria | 100 |
Grab the CSV and edit the wording against your own intake sheet before anybody scores a call with it.
Auto fail items
Four failures void the card no matter how many points the rest of the call earned. A critical fail should reach whoever runs dispatch the same day, ahead of any monthly average.
| Auto fail | Why it is absolute |
|---|---|
| Booked a routine slot on a call that met the company's own emergency definition | The escalation procedure existed and was not followed |
| Dispatched a job to an address that was never read back and confirmed | The truck may never reach the customer |
| Quoted a price, a repair, or a timeframe the company does not offer | The company is now committed to something it cannot deliver |
| Recorded a callback number that does not match what the caller said | The lead cannot be recovered once the call ends |
The safety and urgency lines
Every trade keeps a short list of calls that stop being bookings and turn into escalations. A furnace out on a January night with an elderly customer in the house, a building with no water, a gas odor, sewage coming up through a floor drain.
What the CSR is supposed to tell that caller to do while they wait comes out of your own written escalation procedure, drafted by the people who carry your license and your insurance. The rubric grades whether the CSR followed it and stops there. A QA vendor has no business writing safety instructions for gas or water or live electrical, and none appear here.
We went back and forth on the weight. Fourteen points out of 100 looks light for the section with the word safety in it, and the case for taking it to twenty five is easy enough to make. It stayed at fourteen because a missed emergency condition is already an auto fail, and stacking points on top of an auto fail counts the same failure twice.
Address verification, which looks trivial until it is not
A truck sent to the wrong address costs the round trip plus the slot a paying job could have used, and the customer who sat through the arrival window is the one who writes about it afterward.
Say your office takes 900 calls a month and books 600 of them. One percent of those bookings going out on an address nobody read back is six wasted truck rolls, and at two hours each that is about a week of one tech's time over a quarter.
Access details do the same work. Gate codes, which door the tech should use, the dog, whether the equipment sits in a crawl space with a 30 inch opening. A tech who arrives at the right house and cannot reach the unit has burned the window anyway.
Download the scorecard
The rubric downloads as a CSV with no form in front of it. Add one column to your scoring log that the card itself does not carry, a booked or not booked flag recorded per call. That column is what the last two sections of this page run on.
The card stays generic on purpose, so how the weights were chosen is worth reading before you move points around. For a CSR who also works a mixed support queue, the general call center scorecard is the version without the trade lines.
Where this sits next to a missed call tool
Missed call text back services and AI receptionists work on the calls nobody picked up, and that is a real hole worth plugging. This rubric grades the calls your people did pick up. An office can have voicemail handled and after hours handled and still be losing jobs inside the calls a human answered at 10 a.m., and those two problems get fixed in different places. QA for home services sits downstream of whatever catches the calls you miss.
Testing whether the rubric predicts the booking
Score fifty calls. Split them into the ones that booked and the ones that did not, then compare the section averages between the two piles. Sections where the booked pile scores clearly higher are the ones carrying weight. A section that comes out flat across both piles is measuring something other than what makes a customer say yes, and it should lose points or leave the card.
Fifty calls is not a statistical proof, and two of your CSRs will be over represented in any fifty you pull by hand. Spread the sample across at least two weeks and across everybody who answers, then treat the result as a direction to look rather than a finding to act on.
Our guess is that the appointment section separates hardest, because asking for the booking with two specific windows is the behavior most often missing from a call that did not book. If your own fifty say otherwise, the weights are wrong and the card should move rather than the calls.
Scoring fifty calls by hand runs four or five hours, which is why most offices do this once and never again. Automated scoring is reliable on criteria written as rules, like the greeting string or the fee quoted against the price list, and Aircall and StackAI both report it getting shakier on anything that depends on tone. That is why almost nothing about tone appears on the card above. Rolling scores up per CSR over a month works best on the rule based lines for the same reason.
What the card can say about your booking rate
Booking rate is the number owners in this trade actually watch, and it is the only reason to spend an afternoon scoring phone calls. ServiceTitan publishes benchmark content on it and ships a CSR scorecard inside its own product, and their figures put the average residential CSR booking rate at 65 to 75 percent, top performers at 85 percent and above, and under 50 percent where things are going badly. Those are third party numbers we have not traced to a primary source, so treat them as a rough shape rather than a target.
The mix matters as much as the person does. A week heavy with no heat calls and a week heavy with price shoppers will not produce the same booking rate out of identical skill, so ranking two CSRs on the raw number without looking at what each of them answered is how an office ends up coaching the wrong one.
The card can tell you why a booking was lost when the reason is audible, whether the ask never came, the windows were vague, the address went unverified, or the fee got quoted wrong. It cannot tell you whether your price was simply too high for that customer on that day. Somebody who heard a clean greeting and a correct fee and two real windows and still said no may have been calling three companies and taking the cheapest, and no rubric is going to surface that. Strong scores sitting next to a flat booking rate usually mean the problem has moved off the phone and into the price list.
Send over a handful of your own recordings and we will score them against this card, then put the section averages next to what actually booked, which is all a review of your own recordings is.
Frequently asked questions
What is a good CSR booking rate?
ServiceTitan's published figures put the average for residential CSRs at 65 to 75 percent, with top performers above 85 percent. Those are third party numbers and we have not traced them to a primary source. Your own mix of emergency calls and price shoppers moves the number enough that last quarter's result is a better baseline than anybody's benchmark.
How do you improve CSR booking rate?
Score fifty calls, split them by whether the job booked, and compare the section averages between the two groups. The sections where the booked calls score higher are the behaviors worth coaching first. Spending equal hours on every section means spending some of them on lines that had no effect on the outcome.
What should a CSR scorecard measure?
The card above runs seven sections, opening at greeting and contact capture and closing at confirmation, with urgency triage, the problem captured, address and access, the appointment ask, and fees in between. Four auto fails sit on top of the points, covering a missed emergency condition, an unverified address, a price the company does not offer, and a callback number that does not match what the caller said.
Should we use an AI receptionist or score our CSRs?
Those answer different problems. An AI receptionist or a missed call service picks up what your people did not get to, and a scorecard grades what your people did get to. Jobs lost after hours point at the first one. If the phone gets answered and the work still does not book, the second one is where the money sits.
What is the difference between a CSR scorecard and a call center scorecard?
Two things. A CSR card records the booking outcome next to the score, because booked or not booked is the result the whole call exists to produce. It also carries urgency and safety triage lines that a general card has no reason to include, since a support queue rarely has to recognize a no heat call in January.
Sources
- ServiceTitan on residential CSR booking rate bands and its own CSR scorecard feature. Third party, unverified. Figures captured 2026-09-08 and pending a source URL and date.
- Aircall and StackAI on rule based criteria scoring more reliably than tone. Third party, unverified.
- Trinsic feature sheet and call notes, 2026-08-27 and 2026-09-03. Verified internally.
- The 100 point weighting, the section balance, and the four auto fails. Editorial judgment for this template.