Scoring the conversations according to the scorecards
Conversation Scoring
You can edit the scorecard associated with the conversation from the grid. You can also share and download the conversation.
You can save the filters for this page by selecting the filters of your choice and clicking on save filters. Whenever you visit this page after saving the filters, the grid will, by default, show the data according to the saved filters.
If you click on Reset to default, these filters will be removed.
- Admin and Standard user roles can act as Managers/Selectors and view this page. Also, users with role-type scorers can access the page and score conversations based on permissions.
- Read-Only and Identified-Only users act as agents.
- Users can view the conversations on the Conversations scoring page depending on their role type and permissions.
- You can apply filters by clicking on the ‘Filters’ button.
- Conversations can now be scored as many times as they like but can be rescored only twice after a rescore request.
- The outcome can be true, false, or N/A, and the Outcome will be an optional component of the Scorecard and can be based on AI insights.
- By default, the correct user/agent gets assigned to the conversation using the AI model, but it will also be user-editable on the UI.
- If no agent is found by the CAI agent extraction model/insights, it will be left blank.
- After clicking on conversation, you will be directed to the scoring page below, where you can score all the insights, Manual and AI.
- You can click on Custom insights/ AI insights to view the insights to be scored.
- Classification, Agent Name, and Scorecard are the mandatory fields.
- You can also add tags or comments for the conversation on this page.
- Click on Save Scoring to save the score.
- Select the outcome label and hit save to save the score.
- You can edit the Agent name, Scorecard, and Tags with the inline edit option on this page. You can edit Scorecard only for the unscored conversations.
- The user can turn the toggle to show help text off or on for scoring. It is turned on by default.
- Default view with Help Text toggled on:
Default view with Help Text toggled off: - Coaching View with help text on:
Coaching View with help text off:
This view shows the section name and individual insights for the section, and individual scoring for the child insights in the sections. The help text is not displayed. Tags and comments are shown next to the transcriptions/Insights History (If applicable) - If the scorecard does not have sections, then the same view is shown without section names.
Agents can also add notification alerts to receive for the calls that are scored for them. To do this, click on the Notifications button in the right-hand corner, and it will open the pop-up as shown below. The notifications can be received if a call has been scored for me (Agent) and a call I scored has been reviewed. They can be obtained by email or SMS. If notifications are enabled for scored or reviewed conversations, the designated recipients will receive emails in the specified format mentioned below.
If notifications are enabled for scored or reviewed conversations, the designated recipients will receive emails in the specified format mentioned below.
The notification provides the option to go to the conversation directly. The user must log in to CAI to access the page. The Audio file & Scorecard PDF of the conversation will be attached to the email.
Scoring instructions:
- There are three types of Scorecards.
- AI (Scorecards with all AI insights)
- Manual (Scorecards with all manual insights)
- Hybrid (Scorecards with the combination of AI and manual insights)
- The conversations are scored under the Conversations scoring page. To score a conversation, go to the conversation scoring tab, click on the conversation you wish to score, Select the agent, choose Scorecard (If different from the default Scorecard for the group), click on custom insights, assign a score for every insight, and save the score.
- If the Scorecard assigned to the conversations has all the AI insights, then the conversation is scored automatically.
- All or some insights need to be scored manually for the conversations with a Manual or hybrid Scorecard assigned to them.
- To score the insights, the scorer needs to listen to the call and assign a score for each insight between 0-100. If the score is between 0-50, it is considered a Miss; if it is between 51-100, it is considered a Hit. They also need to consider the instructions mentioned by the client to score the insight.
- The scorer can review and change the score as often as they like.
- If the Scorecard has different sections, the total score is calculated based on the section weightage and the importance of every insight.
- If the Scorecard does not have sections, the total score is calculated based on the importance of every insight.
- The conversations showing in CAI have the status ‘Unscored.’ Once the conversation is scored by the scorer or automatically by the system, the status changes to ‘Scored.’
- The Admins can send the request to rescore the conversation again. To do this, click on a particular conversation with the ‘Scored’ status on the conversations page. You will find the ‘Request rescore’ button at the bottom of the page. When the 1st rescore request is sent, the status of the conversation changes to ‘Rescore request 1’, and when it is rescored for the first time, the status changes to ‘Scored 1.’
- When the 2nd rescore request is sent, the status of the conversation changes to ‘Rescore request 2’, and when it is rescored for the second time, the status changes to ‘Scored 2’. To do this, click on a particular conversation with the ‘Scored 1’ status on the conversations page. You will find the ‘Request rescore’ button at the bottom of the page.
- The conversation can be rescored only twice. Once it is marked as reviewed, the score can not be changed.
FAQ:
Q1. How are the filters defined?
A1. Filers are defined as the current period. Today is the day you are looking at the data. The week is the current week. The month is the current Month, the Quarter is the current Quarter, and the Year is the current year. You can also customize it with the Date Range option.
Q2. Is there a way to know which conversations were scored from the tiles?
A2. No, tiles just give the overview.
Q3. Is there a way to identify the conversations that reached VM? Are actual Robo calls that reached VM or Actual VM?
A3. Robo calls will always get tagged as Classification: "Robo calls," so they won’t indicate as Reached VM
Q4. If the tracking number is Vanity toll-free, all will be routed through the same number. Is there any way to identify it?
A4. The conversations can be manually tagged after listening to them, and there is no automated process to tag them yet. However, there will be keyword filter enhancement in the future that will allow the filtering of conversations based on the keywords matched from the conversation transcriptions.
Q5. Why aren’t more Cloud/Keywords should be represented - More useful/important words should be shown or given a tab or search
A5. This is currently limited due to the data coming from the 3rd party source.
Q6. If two agents have the same name, how does AI identify them separately?
A6. Currently, AI looks up the first name of an agent available in the existing Agent list, Two agents with the same first name can exist in different groups, but in the case where two agents with the same name are in the same group, then the name of the second agent is altered to a nickname.
Q7. Can specific insights be removed from the account?
A7. Yes, it is possible to remove specific insights for the account with the help of the Dev team.
Q8. Is it possible to download the data on Recordings in PDF format?
A8. No, Only the CSV option is available. You can download Audio, Summary, Transcript, or all three.
Q9. Where can the user find the whole transcript of the conversation?
A9. Transcripts for the conversation can be found in the downloaded CSV and on the Conversation page.
Q10. How are the agents assigned to the conversations in CAI?
A10. Agents are assigned to the conversations with the help of the Agent Extraction Model. All users must be added to the CAI account to assign them to conversations.