Overview
Signals are what the Field Pro v2 AI evaluates on every recording. A signal describes anything you want the AI to detect from the transcript of the conversation, and tells the AI how to judge it. Most signals are behaviors, such as asking permission to enter or handling objections. A signal can also capture information, such as a list of needs the customer mentioned. The recording summary itself is a signal, and it's also posted to the Job's history in ServiceTitan.
Think of each signal as a question you want answered about every Call. "Did the technician ask permission to enter?" "How well did they handle objections?" "What needs did the customer mention?" Field Pro asks every question on every recording and records the answer. The signal's instructions are simply how you phrase the question and what a good answer looks like.
Use the Signals tab to see what Field Pro evaluates, create your own signals, copy a system signal so you can change it, write instructions that tell the AI when a signal applies, and adjust signals when their scores aren't what you expect.
Who uses this feature
Administrators, business owners, and sales managers
To view or change signals, you need the View Field Pro Settings and Edit Field Pro Settings permissions. These aren't limited to admins. Anyone whose role includes them can manage Field Pro settings.
Feature configuration
Signals are in ServiceTitan under Settings > Field Pro > Settings, on the Signals tab. Field Pro v2 comes with a library of system signals already in place, so you can review scores before changing anything.
Things to know
System signals are read-only, apart from the Default tag. To change one, copy it and edit the copy.
Choose a signal's type carefully. You can change the type only until the signal is in use, meaning it's added to a recording profile or a signal watch, or evaluated on a recording.
You can delete a custom signal only if it has never been used. After that, deactivate it instead.
Editing a signal affects recordings processed afterward. Recordings already processed keep their results unless you re-analyze them. Scorecard changes are different: they recalculate past scores. See Recording Profiles and Scorecards in Field Pro V2.
A signal does nothing on its own. A recording profile decides which signals are evaluated and which are scored.
Basics
Signals are the building blocks of recording profiles and scorecards. To see how they fit together, read Recording Profiles and Scorecards in Field Pro V2.
Find your signals
Go to Settings
> Field Pro > Settings.Select the Signals tab.
The list shows every signal in your account. System signals have a System badge. Each row shows the signal's type (YES/NO, SCALE, CUSTOM or METRIC), its tags and its version. Use the Status filter to see inactive signals.
Click a signal to open it. Each signal has:
Name: how it appears on scorecards, filters and recordings.
Description: what the signal measures, in plain language.
Tags: labels for grouping and filtering. Three tags (Default, Bookmark and Objection) also change what the product does. See Tags that change behavior under Advanced.
Other tags, such as Process or Skills, are just for filtering. Use New Tag to create your own.
Signal Type: how the signal is evaluated. See Signal types below.
Instructions: what the AI should look for (not used by Metric signals).
Treat Absence as "No Score": what happens when the signal doesn't apply (Yes/No and Grade Scale only).
Additional Output (optional): extra written output, with a checkbox to show it on the scorecard.
Signals that come with Field Pro
Field Pro v2 includes about 60 system signals:
Process signals (Yes/No): moments in your visit, such as Greeting, Permission to Enter, Shoe Covers, Agenda Set, Financing Mention, Membership Plan Mention, Review Ask and Referral Ask.
Skills signals (Grade Scale): such as Building Rapport, Needs Assessment, Value Framing, Objection Handling, Closing Attempt and Compelling Need Identified. A few cover specific kinds of work, such as Energy Efficiency Selling and Indoor Air Quality Discussion.
Objection signals (Yes/No): common objections, such as "I need to think about it" and "I need a lower price".
Speech Proficiency signals (Metric): measurements such as talk-time balance and pacing. They're tracked on the default profile but not scored.
Customer Difficulty Rating: how hard the Appointment was to win, tracked but not scored.
Recording summary: the Custom Output signal that writes the summary.
The default profile tracks all of them and scores about 20.
Signal types
The signal type determines how a signal is evaluated. Three types can produce a score. Custom Output doesn't.
Yes/No
A Yes/No signal asks the AI a single question: did this happen or not?
On a scorecard, a Yes/No signal scores 0 or 100, nothing in between. Use it for discrete moments in your process that either happened or didn't.
Examples of behaviors suited to Yes/No: asking permission to enter, using shoe coverings, setting the agenda.
Grade Scale
A Grade Scale signal asks the AI how well the technician demonstrated the behavior, and scores it on a 0 to 100 scale.
Use it for skills, where performance is a matter of degree. For example, Compelling Need Identified rates how well the technician surfaced the customer's real reason for needing the work, not merely whether they tried.
Custom Output
A Custom Output signal produces written output and no score, so it can't go on a scorecard. Its main use is the recording summary. Each profile uses one Custom Output signal to write its summary.
Signals don't have to be scored to be useful. Any signal a profile tracks but doesn't score appears on the recording's Signals tab, next to the Scorecard tab.
A fourth type, Metric, scores speech measurements that Field Pro calculates, such as talk time. See Metric signals under Advanced.
Choosing a type
If you want to know… | Use |
|---|---|
Whether a step in your process happened | Yes/No |
How well a skill was performed | Grade Scale |
What was said or present, without scoring it | Custom Output |
A speech measurement Field Pro calculates, such as how much the technician talked | Metric |
You can change a signal's type only until it's in use (added to a recording profile or signal watch, or evaluated on a recording). If you need a different type after that, copy the signal and change the type on the copy.
Copy a system signal to change it
On the Signals tab, open the menu (three dots) on the system signal you want to change.
Select Copy. Field Pro creates an editable custom signal named "[signal name] Copy", with the same type, tags, instructions and settings.
Open the copy, rename it if you like, edit it and save.
In your recording profile, swap the copy in for the original. If it was scored, put it in the same scorecard category.
The system signal stays as it was.
Tip: Copying is safer than building from scratch, because you start from instructions ServiceTitan has already tested.
Edit a signal's instructions
To change what a signal looks for, open your copy or custom signal, edit its Instructions in plain language, and save. The change applies to recordings processed afterward. For how to write instructions that score well, see Write instructions that define relevance under Advanced.
Advanced
Most teams can stop at Basics. Read on when you want to fine-tune how signals are written and scored.
Write instructions that define relevance
Instructions tell the AI how to evaluate the signal and, just as importantly, when it doesn't apply. If a customer raises no objections, the technician had nothing to handle. A signal that doesn't apply shouldn't count against the technician. Your instructions can tell the AI when that's the case.
Good instructions cover three things:
What good looks like. The behavior you want the AI to recognize and reward.
When the signal is relevant. The conditions under which the behavior should be expected at all.
When the signal is not relevant. The conditions under which the AI should skip the signal instead of scoring it.
The Instructions field is required for Yes/No, Grade Scale and Custom Output signals. You don't need to write scoring rules, because the score scale is built in. For example:
What good looks like: "Evaluate whether the technician greeted the homeowner warmly and introduced themselves by name."
When it doesn't apply: "If the customer raised no objections, don't score this signal."
Tune a signal when the scores are not what you expected
Signals are meant to be adjusted. Review scores after a week or two of recordings and change signals in response to what you see.
What you are seeing | What to check |
|---|---|
A signal scores zero on conversations that went well | Whether the signal applies to every conversation. If not, turn on Treat Absence as "No Score" and add a line to the instructions saying when not to score it. |
A Grade Scale signal clusters at the top or bottom for everyone | Whether the instructions describe what good looks like specifically enough for the AI to distinguish degrees of performance. |
A signal scores against behaviors you did not intend to measure | Whether the description and instructions match the behavior you actually want, and whether the signal type fits the question. |
A behavior is either present or absent, but the signal is grading it | Whether the type should be Yes/No rather than Grade Scale. |
When you change a signal:
Open the signal on the Signals tab. If it's a system signal, copy it first and swap the copy into your profile.
Adjust the description, instructions or Treat Absence as "No Score". You can change the type only if the signal isn't in use yet.
Save. Review new recordings before you change anything else.
Change one thing at a time, so you know which change moved the scores.
Each time you save a signal, its version number goes up. You can see it in the Version column on the Signals tab.
Use Treat Absence as "No Score"
Treat Absence as "No Score" is a checkbox under Instructions on Yes/No and Grade Scale signals. It's off by default. It matters only when the AI decides the signal didn't apply to the call. A real score of 0 always counts.
Off: a signal that didn't apply scores 0, and that 0 counts toward the category and overall score.
On: a signal that didn't apply is left out of that recording's score. If every signal in a category is left out, the category is left out too, and the other categories are re-weighted.
Turn it on for signals that only apply to some conversations, such as handling objections when there may be none. Some system signals already have it on, such as Permission to Enter and Agenda Set.
The setting and your instructions do different jobs. Your instructions tell the AI when a signal doesn't apply. The setting decides what that means for the score. If you turn the setting on without saying in the instructions when the signal applies, the AI may still score the behavior, often low, instead of skipping it. Set up both.
Leave it off for signals that should apply to every conversation. If a step is part of your process on every job, a missing step is a real miss, and 0 is the correct score.
Metric signals
A Metric signal scores a speech measurement that Field Pro calculates from every recording. The AI doesn't judge it. You pick the measurement in Speech Analytics Field, set the ideal range with Optimal Min and Optimal Max, and choose a Scoring Function that turns the measurement into a 0 to 100 score. Field Pro includes a set of Speech Proficiency metric signals you can track or add to a scorecard.
Speech Analytics Field | What it measures |
|---|---|
Rep Speaking Ratio | How much of the conversation the technician did the talking |
Rep Words/Min | The technician's speaking pace |
Rep Filler Words/Min | Filler words (such as "um" and "uh") per minute of the technician's speech |
Rep Question Count | How many questions the technician asked |
Customer Question Count | How many questions the customer asked |
Exchange Count | How many times the conversation went back and forth |
Longest Monologue (ms) | The longest stretch the technician talked without the customer speaking, in milliseconds |
Longest Customer Story (ms) | The longest stretch the customer talked, in milliseconds |
Dead Air (ms) | Total silence in the conversation, in milliseconds |
Scoring Function | How it scores | Uses |
|---|---|---|
Linear Falloff | 100 anywhere inside the ideal range. Outside it, the score drops the further the call falls from the range, reaching 0 when it's a full range-width away. For example, with a range of 120 to 160 words per minute, a call at 100 scores 50. | Optimal Min and Optimal Max |
Threshold | 100 when the value is at or above Optimal Min, otherwise 0. Use it for "at least this much." | Optimal Min |
Inverse | Lower is better. 100 at zero, dropping evenly to 0 at Optimal Max. Use it for things like filler words or dead air. | Optimal Max |
Tags that change behavior
Most tags are just for filtering. Three tags also change what Field Pro does:
Default adds the signal to new recording profiles automatically (it won't auto add it to existing profiles though).
Bookmark marks the moment on the recording's transcript.
Objection includes the signal in objection tracking and the Objections report.
Apply a change to past recordings
Signal changes apply to recordings processed afterward. Past recordings keep their results. To apply a change to a specific recording, open it, open its menu (three dots), and select Reprocess > Re-analyze. That re-scores the recording from its existing transcript. Full Reprocess starts again from the audio. The recording's earlier results are replaced and can't be recovered.
Retire a signal
You can delete a custom signal only if it has never been used. Once it has been used, deactivate it instead:
Remove it from every recording profile that uses it.
On the Signals tab, open its menu and select Deactivate.
Deactivated signals aren't evaluated or scored. To find one, set the Status filter to Inactive. Select Activate to bring it back. System signals can't be deleted or deactivated; remove them from your profiles instead.
When to bring in your Pro Product Specialist
You can write and tune signals yourself. Contact your Pro Product Specialist when:
You are translating a formal sales process or scorecard methodology into signals for the first time.
A signal is not producing sensible scores after you have revised its instructions more than once.
You are unsure whether a behavior belongs in one signal or should be split across several.
You want to review whether your signal set still reflects how your team actually sells.
Bring specifics: the signal, the score you expected, the score it produced, and an example recording.
If you don't know who your Pro Product Specialist is, ask your Customer Success Manager.