Skip to content

Release History

  • Import the form your supervisors already use. Import rubric in the rubric editor reads Excel, Word, PDF, CSV or pasted text and copies each item’s wording word for word. Values the form doesn’t give are marked Value not in the document for you to check. See Importing your rubric.
  • A wording review with suggested fixes. After an import, or when you click Review wording, each criterion is checked for wording that would make the AI score inconsistently. Each finding comes with a rewrite to accept or dismiss, and any choice the rewrite makes is listed under it for you to check. Accept all suggestions accepts them all at once. See Reviewing the wording.
  • Calibrate a rubric against your supervisors. The new Calibrate tab scores up to 30 calls your supervisors already scored, twice. It shows the agreement per criterion, the scores that changed between the two passes, and each disagreement with the AI’s justification. Suggest wording fixes turns the disagreements into suggestions. Calibration runs don’t count against your plan and never change your reports. See Calibrating against your supervisors.
  • Rubric versions. Every save that changes the criteria creates a new version, and each call keeps the version it was scored with. The rubric card shows how many calls were scored with an earlier version, and combined COPC results say when they mix versions. See Versions.
  • Starter rubrics. Start from a template copies a ready-made rubric for inbound banking service or outbound sales, in Spanish, into your team. Duplicate copies one of your own.
  • A demo call project. Twenty sample calls of a fictional parcel company, already scored, with a supervisor’s scores loaded for calibration. It doesn’t use your plan’s hours or count as one of your projects. See Try the demo project first.
  • CSV import lists the rows it can’t use, with the row number and the reason, instead of skipping them silently. It also accepts si, sí and x for yes, and points written as 10.0.
  • The rubric editor asks before discarding unsaved changes when you leave it or close the page.
  • More Help links. The survey, templates, transcript and call metadata screens have a Help link too.
  • Projects have a read-only is_demo field, true for the team’s demo call project. See Projects.
  • Download any agent’s personal report. The Agent performance table on the Call Center Report has a Personal report column with a download icon for each agent. It downloads the agent’s personal PDF for the projects and dates you have selected, ready to send. The agent doesn’t need a login.
  • Clearer PDF reports. In the scorecard, a criterion that didn’t apply to any call shows — as its result instead of 100%. Audio time is shown in hours and minutes, and a section’s heading no longer ends up alone at the bottom of a page.
  • Personal PDF closing note. It now only explains how calls are scored, so it suits agents who receive the report by email as well as those with a login.
  • headline.analyzed_seconds in the report gives the exact audio length; analyzed_hours stays rounded to a tenth of an hour.
  • Help links. The analysis wizard and the Conversations, results, rubrics, Call Center Report, team and profile pages have a Help link that opens the matching documentation page in a new tab. New pages cover Team Management and Your Profile.
  • A warning when calls have no agent metadata. On call projects with automatic analysis on, Conversations shows how many calls have no agent metadata, with an Upload metadata (CSV) button. Calls keep processing, and reports use a name guessed from the conversation until you add the metadata. See Automatic Analysis.
  • The analysis wizard points call projects to the right analysis. Predefined Analysis and COPC Evaluation are marked Recommended. COPC Evaluation and Calls Analysis come first, and the other types are behind More.
  • Create a rubric from the wizard. If you choose COPC and have no rubric, Create Rubric opens the editor in a new tab, and the wizard selects the new rubric when you come back.
  • POST /api/files/ can return a warnings list in its 201 response. Uploads to call projects with automatic analysis on get call_metadata_missing when nothing says who the agent is. These uploads are still accepted and processed, and existing clients don’t need to change. To avoid the warning, see Agent metadata with automatic analysis.
  • The OpenAPI schema documents warnings on the upload response.
  • Agents can see their own results. Invite a call center agent with the new Agent role and they get a login showing only the calls where your call metadata names them: their report, scores and transcripts. They can’t see other agents, team-wide figures or settings, and they can’t change anything. See Agent Access.
  • A personal PDF for each agent. When an agent clicks Export PDF, they get a report written to them from their own figures: a summary, What you do well, Where to improve, Your best calls and Recommendations for you.
  • Agent email in call metadata. The CSV manifest has a new agent_email column, and Edit call metadata has an Agent email field. An agent_id that is an email address also counts.
  • Role picker when inviting. The invitation form in Team Settings lets you choose Member, Administrator or Agent.
  • Call metadata accepts agent_email on upload, update and the CSV manifest, and returns it in call_metadata.
  • The report and PDF accept agent for one agent’s personal report. The report has two new fields, scope and agent_name.
  • Accounts with the Agent role have no API access. API keys they already have are rejected with 403 Forbidden.
  • Closing steps are not docked when the customer hangs up first. Send who hung up first from your dialer as ended_by in the call metadata, and tick Excused if the customer hangs up before the farewell on the closing steps of your rubric. On calls where the customer hung up before the agent said goodbye, those steps get full points and are marked N/A, with a Customer hung up badge. Critical criteria are never excused. See Closing steps when the customer hangs up.
  • Suggest hang-up exemptions. The rubric editor can read your criterion descriptions and tick the closing steps for you to review. It runs on its own after importing a CSV that has no hang-up column.
  • The Edit call metadata dialog, the call detail and the Excel export of COPC results show Hung up first.
  • Call metadata accepts ended_by (customer or agent) on upload, update and the CSV manifest, and returns it in call_metadata and in the call detail.
  • A criterion excused this way carries "override": "customer_hangup" in the call detail.
  • The rubric CSV import reads an eighth column, excused_on_customer_hangup.
  • Team invitations expire after 14 days. The pending invitations list shows when each invitation was sent and marks expired ones. Resend Invitation sends a new email and starts the 14 days again. Invitations sent before this release count from the day of the update.
  • The top bar and the sidebar stay in place as you scroll.
  • Change Password is no longer shown for accounts that sign in only with Google.
  • More of the interface is translated into Spanish, including Manage Accounts and the light and dark theme options.
  • Email confirmation links open a page where you click Confirm, so an email scanner that opens the link can’t confirm the address for you.
  • Subscriptions go to the team that pays. A plan bought through checkout could be attached to a different team than the one that bought it. Plans and add-on minutes are now always credited to the paying team.
  • HTTP Basic authentication (username and password) is no longer accepted. Use an API key.
  • Month-on-month comparison. The scorecard compares whole months with the previous months (September with all of August) and month to date with the same days of last month, and shows the dates it compares against. See Quality scorecard by criterion.
  • Your time zone. Report periods start at midnight in the time zone set in your profile, not UTC.
  • Day, week or month. The compliance trend chart can be shown by day, week or month.
  • Applied to always shows a count, such as 25 of 25, instead of a blank.
  • The report accepts granularity for the trend, whose points carry period instead of week, and returns the comparison dates in date_range.previous_start and previous_end.
  • The Call Center Report counts only analyzed calls. A call uploaded with an agent or a call date in its metadata used to appear in the report before any analysis had run, so an agent could show calls with no compliance or sentiment, and Calls analyzed counted calls that were never analyzed. Those calls are now left out of every figure, the agent table, the calls list and the PDF until an analysis covers them.
  • The report says how many calls are waiting. When transcribed calls in the selected period have not been analyzed, the report shows how many and how to analyze them: run an analysis from the project, or turn on automatic analysis so new calls are analyzed as they arrive. The COPC Evaluation and Calls Analysis results pages show the same warning for the project, and the PDF prints it under the headline metrics. See Calls not analyzed yet.
  • PDF coverage counts every call received. Coverage is now the share of all calls received in the period that were scored, including those not analyzed yet, rather than the share of analyzed calls.
  • The PDF’s closing note and footer now stay together on the last page, and the contact address no longer wraps.
  • The report has a new field, unanalyzed_calls: transcribed calls in the window that no analysis has covered yet. They are not part of any other figure.
  • analyzed_calls in analysis-status now counts only calls with results. It used to include calls that had only metadata.
  • Exemplary calls. The Call Center Report now shows the period’s best calls, not only the problems: high-scoring calls with no critical errors where most of the rubric applied, with the calls that reached their goal first. Choose how many to show — 5, 10, 20 or 50 — and click any call to open it.
  • Best practices to replicate. One click turns the exemplary calls into three to five behaviors to teach other agents, each with an explanation and a quote from the call it came from. See Best practices to replicate.
  • Goal of the call. Mark the rubric criterion that records the outcome — a sale closed, a payment promised, an issue resolved — and every scored call shows whether it reached its goal. No extra analysis is needed, and calls already scored are updated straight away. See The goal of the call.
  • Exemplary only. A new filter in the report’s calls list shows only the exemplary calls, best first.
  • PDF export. The report PDF now includes the exemplary calls and, once generated, the best practices.
  • Sorting the calls list by compliance now always puts calls without a score at the end.
  • Importing rubric criteria from CSV accepts a new optional is_goal column, and checks the file the same way the editor does: a file with duplicate criterion IDs, more than one goal, or a goal that is also critical is rejected with a message instead of being saved.
  • Clearer error messages when a rubric cannot be saved.