Docs

The Heartbeat platform documentation

Everything you need to run call QA on The Heartbeat — from connecting your first recording to reading agent scorecards and administering workspaces.

What the platform does

The Heartbeat audits contact-centre calls. Recordings arrive from your storage or a manual upload, and the platform transcribes them, separates agent and customer speech, scores sentiment and emotion, flags risky and non-compliant moments, and rolls all of it up into agent-level and workspace-level reporting. Trained analysts verify flags before they reach your reports, so what supervisors act on is real rather than algorithmic noise.

Two things use the same engine. The weekly QA report is the zero-effort option — a PDF in your inbox every Monday, nothing to log into. The web application documented here is for teams that want to work the calls themselves: browse and filter every processed call, listen alongside the transcript, comment and tag colleagues, build charts, and track each agent week over week.

Scope of these docs

These pages cover the web application: what each screen does, what every metric means, and how to administer it. If you only receive the weekly report, the FAQ and How It Works pages are the better starting point.

How a call flows through

  1. Ingest A recording lands in the platform either from a connected source (Amazon S3, Google Drive, Dropbox, Box, GCP Bucket, Wasabi) on a schedule, or from a direct upload in the app.
    Diagram: recordings arrive from Amazon S3, Google Drive, Dropbox, Box, GCP Bucket or Wasabi on a schedule, or by direct upload, and land in The Heartbeat with their call metadata
    Connect & collect — recordings and call metadata come from the systems you already run.
  2. Process The call is transcribed, split into agent and customer channels, and scored — text sentiment, audio emotion, talk and silence ratios, dead air, cross talk, flagged keywords, and the call score built from your workspace's opening / handling / closing weightage.
    Diagram: one recording is split into agent and customer channels, transcribed, then measured for sentiment, emotion, talk ratio, dead air and flagged keywords, producing a call score
    Speech-to-text and NLP models transcribe the call and measure how it sounded.
  3. Verify Analysts review flagged moments in the language they were spoken in, so false positives don't reach your supervisors.
  4. Review The call appears on the dashboard with its sentiment, duration and score. Open it to hear the audio against the transcript, read the summary, and leave comments for your team.
  5. Report Everything aggregates into the dashboard KPIs, your saved analytics charts, and the agent performance scorecards — plus the weekly PDF if your plan includes it.
    Diagram: every processed call aggregates into dashboard KPIs, analytics charts, agent scorecards and the weekly PDF report
    Insights are distilled across thousands of calls, not the handful a team can sample by hand.
Diagram: the five stages a call passes through — ingest, process, verify, review and report — with coaching feeding back into the next week
The same loop end to end. Nothing here is a sampling step — every call that arrives goes through all five, and what the last one produces is what shapes the next week's coaching.

The app at a glance

The application is a left sidebar plus a slim top bar. The sidebar is grouped into three sections, and only shows destinations your account can actually reach:

Sidebar groupDestinationWhat it's for
Overview Dashboard KPIs, sentiment and emotion charts, and the recent-calls table for the active workspace.
Agents Performance Per-agent scorecards, parameter and overall scores.
Upload Call Drag-and-drop upload into a chosen workspace, or import from a connected source.
Analytics Your own saved bar, line and quadrant charts across agents and flags.
Manage Workspaces Create workspaces and set goals, PII masking, language, dead air and processing limits.
All Users Invite admins and users, assign workspaces, resend or revoke invites.
All Agents The agent roster and what each agent is allowed to see.
Integrations Connected storage sources, scan schedules and on-demand runs.
Account Usage Wallet balance, remaining minutes, processing status counts and the daily ledger.
Account Profile, password, plan and billing, and webhook preferences.

The top bar carries the page title, the workspace switcher (everything on the dashboard, analytics and call screens is scoped to the workspace selected here), an Upload Audio shortcut, the notification bell for comment mentions, and your profile.

Who sees what

Every screen is permission-gated, so two people signed in to the same account can see different sidebars. There are three roles:

  • Admin Account owner Full reach — all workspaces, user and agent management, integrations, plan and billing.
  • User Team member Scoped to the workspaces they were invited to. Typically QA leads and supervisors who review calls and coach agents.
  • Agent Frontline agent Sees only their own calls and performance, and only if an admin grants access. Agents have no management or billing screens.

Full detail, including the four agent access levels, is in Users, roles & access.