Most quality programs stop at the score. A reviewer grades a call, the number lands on a dashboard, and the process assumes the agent will somehow get better standing next to a lower bar. That assumption is where improvement programs die. Devotel Orbit's coaching-plan workspace fills the missing half: a supervisor authors an explicit multi-week plan against a named agent, the agent works it, the plan closes with an outcome, and the platform reports how the agent's handle time, first-contact resolution, and CSAT actually moved between the before and after windows.
Why score-based coaching decays
Point-in-time QA scores are necessary and insufficient at the same time. Necessary, because without a score there is no signal that something needs work. Insufficient, because a score is a snapshot — it names a gap at a moment, and then the organization hands that gap to the weakest tool it has: a coaching conversation that produces no artifact.
The decay pattern is predictable. The supervisor holds a session, the agent nods, the conversation is not attached to a measurable goal, and two weeks later the same scorecard produces the same score on the same weakness. The QA review did its detection job and the remediation leg dissolved into memory. Repeat this for two quarters and the program settles into the steady-state line the industry calls a QA plateau: measurement stays high, quality stays flat.
What breaks the plateau is not more measurement — it is attaching the measurement to a longitudinal container. The finding needs to become a plan with a goal, a window, an owner, a close-out, and a measured result, or it stays a comment on a scorecard.
The coaching-plan model
Devotel Orbit grows the point-in-time score into a four-stage lifecycle, and each stage maps to an action the supervisor takes in the workspace.
Author. The supervisor creates a plan against one agent: the agent, a free-text goal ("Reduce average handle time by 20% over 4 weeks"), and an optional target date. A plan can also arrive on its own — the coaching rule engine reads scorecard outcomes and auto-assigns a plan when a threshold trips, and those land in the same list marked as auto-assigned. Either way the plan is a named, dated commitment, not a chat note.
Execute. The plan sits in active status while the agent works the goal. It is editable during this stage — the goal, the target date, the addressed agent — because a four-week plan that cannot react to reality is a plan that dies on contact with reality.
Close. When the window ends, the supervisor closes the plan with a terminal status (completed or cancelled) and an outcome judgment: goal met, partially met, or missed, plus free-form notes. The close is permanent — it is the human verdict on the plan, and it is what separates a program that records outcomes from a program that records activity.
Measure. The effectiveness report joins a comparison window to the plan and reports the before/after movement on three KPIs: average handle time, first-contact resolution, and CSAT. Each metric gets a direction (improved, regressed, flat, or insufficient data) and a percentage change. The report is the metric complement to the supervisor's outcome judgment — the two read alongside each other, not instead of each other.
The result of all four is that a coaching program stops being a collection of undocumented conversations and becomes a ledger of plans whose outcomes the platform can count.
Walking the workspace
Open Voice → Quality & Analytics → Coaching Plans. The header lists every plan, filterable by status (active, completed, cancelled, or all), with the agent, the goal, the target date, the lifecycle stage, and the outcome badge once the plan is closed.
Creating a plan opens a dialog with three fields: the agent's user ID, the goal text, and an optional target date. The Create Plan button sits in the header; an empty list offers the same action as a first-run empty state, so the workspace reads as an invitation rather than a blank page.
An active plan carries an Edit action (revise the goal or the target date) and a Close action. Closing opens a dialog that records the terminal status, the outcome judgment, and optional notes; the plan moves out of active and cannot be edited again.
Every plan — active or closed — carries an Impact action. That opens the effectiveness report: one tile per KPI shows the before value, an arrow, the after value, and a direction label with the percentage change. For a plan that is still active, the report reads as in-flight progress measured from the plan start, not a final verdict. Below the tiles, the report names the comparison window and the number of calls joined into the measurement, so the supervisor reads the direction in the context of its sample size rather than as a naked up-or-down arrow.
The API surface
The workspace is a thin surface over four endpoints on the Voice API. POST and GET on /voice/coaching-plans author a plan and list plans (the list filters by agent or by status). POST /voice/coaching-plans/:id/close performs the terminal transition with the status, the outcome judgment, and the notes. GET /voice/coaching-plans/:id/effectiveness returns the before/after KPI report — the metric set, the before and after values per metric, the direction, the percentage change, the window in days, and the call counts on each side of the window.
A customer building on the API gets the same lifecycle the dashboard drives: author, list, judge on close, measure per plan. The report is tenant-scoped like the rest of the Voice surface, and the supervisor's outcome judgment and the measured KPI movement travel as separate fields — the API does not fuse a verdict and a measurement into a single number you cannot audit.
Honest framing
The effectiveness report measures movement; it does not promise causation. A plan that closes with goal met and an AHT that dropped 14% in the after window is evidence that the coaching and the movement coincided — not proof the plan caused the drop. Volume mix shifts, queue composition changes, seasonality, and the agent's own growth all move the same metrics, and no before/after window separates coaching from that background.
That said, the honest version of the measurement is also the useful one. A program that measures direction per plan can do two things a score-only program cannot: it can stack the plans that moved the metric against the ones that did not and refine the coaching content from the difference, and it can hold the supervisor's judgment and the data-side movement next to each other — a goal-marked-met that moved nothing invites a different conversation than a goal-marked-met that moved the number. The report is the discipline of writing down what actually happened, per plan, so improvement compounds instead of resetting every quarter.
Where coaching plans sit in the Quality suite
Coaching plans close the loop the other Quality surfaces open. The scorecard finds the gap on /quality/evaluations; the remediation leg rehearses the failure mode on /quality/practice (covered in the practice-studio walkthrough); quality scoring itself names what to measure; and the leaderboard keeps the ranking honest enough to credit development rather than raw output. The coaching-plan workspace is the longitudinal container that joins those point-in-time surfaces into an improvement loop with a written outcome — the missing half a QA review by itself will never produce.