Skip to main content
Train Signal AI with a representative sample of calls and ground truth data — by manual scoring, spreadsheet upload, or Invoca’s human call scoring service.
Signal AI is a tool for gathering and acting on your call intelligence in real time, based on machine learning. For your Signal AI to learn to classify your calls correctly, it has to be taught. Whether you’re creating a new AI Signal or refining an existing one, follow these steps to train your Invoca account to provide the most accurate Signal AI insights possible.

​
Choosing a sample of calls

Your Signal AI insights are based on patterns within your calls, so training it requires recordings of those calls. Keep two things in mind when building your sample:
  • Represent a realistic balance of Signal outcomes. Your training model gets the best results from a sample that reflects the calls you’ll actually receive. For example, if sales typically occur in 15% of your calls, train your Signal AI using a sample that includes roughly 15% sales as well.
  • Provide a larger sample for better results. The more calls Signal AI has access to, the better it recognizes patterns. To start, we recommend at least 150 examples of the least common outcome of your insight to establish version 1 of each Signal. In the example above, if 15% of your calls result in sales, you’d need 150 sales calls out of a sample of 1,000. From there, your Signal AI automatically retrains and updates to a new version as you provide more recordings — each version requires 50% more data than the last, and higher versions produce more precise results.
This article assumes you’re training your Signal AI using call recordings already in your Invoca account. If your recordings are stored in an external system, see Signal AI Training: Recording Imports. Invoca omits words from both Invoca’s and your organization’s first IVR (interactive voice response). If a call transfers to another handler and hits a second IVR after that first one, Invoca won’t omit words from the second.

​
Providing ground truth data

Next, report the results of your Signal insights for each call in your sample — known as ground truth data. With that data, Signal AI can identify patterns in your successful and unsuccessful calls, then apply those patterns to future calls. If you haven’t already indicated which results Signal AI should look for, follow the guide on Signal AI: Advanced Conversation Insights first. You have three options for providing ground truth data:
  • Listening to your call recordings and manually scoring your Signal outcomes.
  • Uploading a spreadsheet with outcomes for all your Signals across each call.
  • Paying for Invoca’s human call scoring service.

​
If you’re listening to and scoring calls manually

Follow the guide on Signal Management: Manual Result Overrides to update Signal results for the calls you’ve identified. When you manually correct Signal results, your account gathers that information and updates your Signal AI models accordingly.

​
If you’re uploading a spreadsheet

Choose this method if you keep records of your Signal outcomes in a third-party system, such as a CRM.
  1. Log in to your Invoca account. In the sidebar menu, click Signal, then select Manage Signals.
  2. Select Signal AI Training Data, then click Download Template.
  3. Select “I have IDs for Invoca calls or transactions” to download a spreadsheet formatted to help you upload your call records accurately.
  4. Open the template. Note the first three columns — transaction_id, call_record_id, and call_start_time. These are unique identifiers used to match the call record from your third-party system with the associated record in your Invoca account. You only need to fill in one of these columns per call record:
    • transaction_id: the ID of the transaction (call leg) being reported.
    • call_record_id: the ID of the complete call being reported.
    • call_start_time: doesn’t need to be exact, since clocks aren’t perfectly synchronized — Invoca pairs this record with the closest match in your account.
    Of the three identifier columns, call_start_time is the least precise. If you fill in more than one, Invoca matches records in this order: transaction_id, then call_record_id, then call_start_time. If you only fill in call_start_time, we strongly recommend also filling in either calling_phone_number or duration_in_seconds for a more precise match.
  5. Enter your call record data into the template. Each row updates one call record with values for each Signal in your account. The template contains a column named for each Signal — enter the result of that Signal event as “true”/“false,” “yes”/“no,” or “1”/“0.” If you don’t need to include a particular Signal in any call records, you can delete that column or leave it blank.
  6. Save the spreadsheet as a Microsoft Excel file or .csv file.
  7. Return to your Invoca browser window. In the Upload New File menu, click the cloud icon and select your saved spreadsheet.
  8. Select your time zone from the dropdown menu, then click Upload and Preview.
  9. Invoca checks your spreadsheet for errors. If errors are found, see Signal File Uploads: Formatting & Data Prep. Otherwise, click Proceed to apply your uploaded Signal records.
Once you click Proceed, a status bar shows the progress of your upload. When the file finishes processing, you can view the results in the upload history table.

​
If you’re paying for human call scoring

Invoca offers a human call scoring service for customers who don’t have the time or resources to listen to and tag call recordings for AI Signal training or retraining. With this service, a dedicated team of trained call analysis agents listens to your calls and trains your Signal AI. That said, if you have the ability, we recommend listening to your own calls — nobody knows your business better than you do. To request this service:
  1. Log in to your Invoca account. In the sidebar menu, click Signal, then select Manage Signals.
  2. In your list of Signals, select the Signal you’d like to train. This Signal must have the type “AI (Custom).”
  3. Click Provide Training Data, then under “I need the data,” select Request Service.
  4. Click Request Human Call Scoring to be connected with a member of Invoca’s Customer Support team, who will follow up with more information.
Once you’ve started training your Signal AI, you can check its progress by selecting Signal in the sidebar menu, then selecting the Signal you’re training. In the Version section, you’ll see an indicator of your training progress — more progress helps improve accuracy, so keep scoring those calls.

​
Where to go next

Last modified on September 23, 2026