Every call graded against your checklist, with the quote to prove it.

Most CSR coaching runs on two inputs: the manager's memory of a bad call and a five-call sample pulled on Friday afternoon. Say your booking rate slipped from 74% to 66% last month: neither one tells you why. Call scorecards grade every recorded call against a checklist you write, and they show their work.

What a graded call looks like

Take an example: a 4-minute call from Karen Baker about an AC that isn't cooling. Kim, the CSR, books it. The graded call shows:

• Overall: 86 of 100.
• Greeting (weight 1): Excellent. Quote: "Summit Heating and Air, this is Kim."
• Discovery (weight 2): Good. Quote: "Is it blowing warm air, or not blowing at all?"
• Asks for the booking (weight 3): Excellent. Quote: "I can get someone out between 10 and 12 tomorrow. Does that work?"
• Membership offer (weight 1): Missed. Not mentioned on this call.
• Outcome: Booked. Membership opportunity: yes, not offered.
• Coaching notes: two to four specific suggestions, such as mentioning the Comfort Club when the customer says the unit is over ten years old.

The quotes aren't decoration. Each one is checked against the line it cites. If the AI claims Kim said something she didn't, that score is shown as "Not counted: no matching line" and stays out of the 86.

The default scorecards, and how to change them

TradesWork starts every company with two scorecards written for the trades. They're our own rubric, not a licensed sales script, and you're expected to edit them.

Each criterion has a short key, a name, a rubric of up to 600 characters and the conversation types it applies to. The rubric is what the AI grades against, word for word, so write it the way you'd explain it to a new hire: "Offers a specific arrival window and asks the customer to confirm it. Saying 'someone will call you' does not count."

The membership offer on the call scorecard applies to human calls only, not AI receptionist calls, because you may not want the AI selling. Change that if you do.

Saving creates a new version. Calls already graded keep the version they were graded with, so a rubric change doesn't rewrite last month's numbers, and trends follow each criterion's key across versions.

Which conversations get graded

Office calls. Inbound and outbound calls that were recorded and completed. Recording only starts after your consent notice plays, so a recording is itself the proof the notice played. Calls without one are skipped. Federal law allows recording with one party's consent, but several states require everyone on the call to agree, which is why TradesWork offers an explicit announcement for those states.¹

AI receptionist calls. The AI receptionist tells callers the call is transcribed, and it keeps its own turn-by-turn transcript, so these calls can be graded even without a speech-to-text service. This is on by default once grading is on.

In-home sales conversations. Off until an owner turns it on. A technician assigned to the appointment taps record in the field app, and before any audio exists, the consent is logged: agreed or declined, how (customer tap or verbal yes confirmed by the tech), the state, and the exact wording shown. In all-party consent states, and wherever the state is unknown, the customer must tap to agree themselves. If they decline, nothing is recorded. Audio is uploaded in roughly five-minute pieces, transcribed straight away and never stored.

Skipped: voicemails, missed calls, calls under your minimum length (default 30 seconds), and calls whose transcript doesn't arrive within two hours.

The insights page

/conversations/insights is where the weekly coaching meeting starts. Filter to 7, 30 or 90 days, calls only or in-home only.

• AI-verified booking rate next to the logged booking rate. If CSRs mark calls booked that the transcript says weren't, or the other way round, the mismatch count tells you.
• Should-have-booked calls and missed membership offers: the two biggest piles of money left on the table.
• Top objections: price, timing, need to think, spouse or partner, trust, other quotes, financing, warranty.
• Why calls didn't book: price, no availability, wants to think, shopping around, decision-maker absent, outside service area, service not offered, caller ended.
• Criteria averages and a people table.

Privacy and who sees what

Phone numbers and emails are replaced with [number] and [email] before a transcript reaches the AI. The audit trail stores scores and outcomes, not transcript text. CSRs and technicians see only their own conversations and their own scorecard with an 8-week trend next to the team average. Managers with coaching access see the team and can flag moments. Viewers see scores and line numbers, but no transcripts or quotes. Accounting has no access.

Graded calls also feed two office metrics in coaching: call score and AI booking rate.

How it compares to ServiceTitan

ServiceTitan's Phones Pro lists real-time call transcription and agent scorecards.² Its AI Virtual Agent uses the same scorecards and dashboards as human CSRs.³ Field Pro auto-records every conversation on arrival "with built-in consent" and scores every call against your own field process.⁴

When it's not worth turning on

If you don't record calls and don't plan to, this has nothing to grade except AI receptionist calls. If you have one CSR and listen to most calls yourself, the insights page will tell you less than you already know. It earns its keep at two or more people answering phones, or once your techs are selling replacements in the home and you want to know why one closes at 45% and another at 20%.