Real-Time Adherence - mental model and key terms
Objective. Build the mental model Real-Time Adherence is built on - adherence as a comparison of actual state against the schedule, layered into three distinct questions - and define the key terms plainly so a Sales or CSM person uses them precisely.
The one idea to hold: adherence is a comparison, not a performance score. It asks 'did the actual state match the plan?' RTA layers that one comparison into three - schedule, work and output - because presence, effort and results are three different questions. Get that model right and the glossary and the guardrails fall out of it.
See it in the demo
The core concept: adherence is a comparison against the plan
Adherence is the comparison of what actually happened against what the schedule said should happen. That is the whole idea. It is not a measure of how hard someone worked, how much they produced, or whether they are good at their job - it is 'did the actual state match the planned state'.
Two consequences follow immediately and drive everything else in RTA:
- Because adherence is a comparison against a plan, it is only as correct as the plan (the published schedule) and the rules that define a match (grace windows, state mappings). Change the plan or the rules and the same behaviour changes verdict.
- Because 'matching the plan' says nothing about results, presence can be perfect while the work is not getting done. That gap is exactly why RTA adds the work and output layers on top of classic schedule adherence.
RTA is real-time because the comparison is run live, while the shift is still fixable - its value is intraday, not retrospective.
The three layers - the heart of the model
RTA runs the adherence comparison at three levels on the same roster, answering three different questions:
- Schedule adherence - is the agent in the planned state at the planned time? The classic view. Answers 'are they on plan'.
- Work adherence - is that scheduled time actually being spent on work (drawn from activity capture)? Answers 'is the present time turning into work'.
- Output adherence - is the work that is happening meeting the expected output standard for the interval? Answers 'is the work meeting standard'.
They stack: an agent passes schedule adherence by being present and on-state, passes work adherence if that time becomes work, and passes output adherence if the work meets standard. They must never be collapsed into one score, because the fix for each is different. A presence gap is a scheduling or attendance matter; an output gap is about the work itself - complexity, process, the standard, Work Output.
Key terms defined plainly
- Schedule adherence - actual state matched the planned state at the planned time. Presence and state only.
- Work adherence - scheduled time was actually spent on work, not merely logged in.
- Output adherence - the work produced met the expected standard for the interval.
- Published schedule - the currently live, published version of the roster RTA measures against (not a draft or a stale import).
- Grace window - the tolerance before a deviation counts as a violation (for example, a short break overrun is absorbed, not flagged).
- State mapping - how the customer's real aux and work states are mapped to RTA's state model so 'on work' and 'off-agent' mean what the customer means.
- Exception - a signal raised when something crosses a threshold and a lead should decide on it. A decision-pending item, not a confirmed problem.
- Disposition - the decision a lead records on an exception: action it, or dismiss it with a reason.
- Coverage - required versus actually-active staffing for a specific interval and skill (not roster headcount).
- Output standard - the expected output level for an interval/work type that output adherence compares against; only as trustworthy as how it was derived.
- Response time - minutes from a raised signal to a lead acting on it; RTA's headline value measure.
How the concepts chain together
The model runs as a chain, and each link depends on the one before it:
- A published schedule sets the plan. State mappings and grace windows (the rules) define what counts as a match.
- Live activity capture is compared against that plan to produce schedule adherence and work adherence.
- An output signal compared against the output standard produces output adherence.
- Where any layer crosses a threshold, an exception is raised; a lead works the queue and records a disposition.
- Coverage reads the same live picture per interval to show where the floor is thin while it is still fixable.
- Response time measures how fast signals become action - the behaviour that makes RTA worth adopting.
Break an early link - a stale schedule, a wrong state mapping, a bad time zone - and every later link is wrong. That is why validating the rules and feeds comes before trusting any number.
The one concept people most often get wrong
The near-universal error is treating adherence as a productivity or performance score. It is not. Adherence is a comparison against a plan - a red state can be a perfectly correct reassignment to protect a hotter queue, and a schedule-perfect agent can be well below the output standard.
Say it in one line to a customer: green on the schedule board means present and on-plan, not that the work is getting done - that is what the work and output layers are for. Everything else in the model exists to keep those three questions separate and to keep a live signal a prompt to look, not a verdict to act on blindly.
Interpretation check
Adherence is a comparison of actual state against the planned schedule, not a performance score. RTA layers that comparison into three separate questions - schedule (present on plan), work (time spent on work), output (work meets standard) - and each is only as correct as the published schedule and rules behind it.
In the field
OVO (contact centre): the effort-to-outcome POC is the cleanest illustration of the chain - presence (schedule) and activity (work) are not the goal; work meeting standard (output) is. KnackGlobal (RCM, multi-site): the 'published schedule' and 'time zone' links in the chain are where a multi-site operation most often breaks the model, making the comparison wrong before any layer is read.
Practical task
Write the glossary from memory: define schedule adherence, work adherence, output adherence, published schedule, grace window, state mapping, exception, disposition, coverage and output standard in one plain line each. Then draw the six-step chain (schedule+rules -> capture -> adherence layers -> output standard -> exceptions/disposition -> coverage/response time) and mark the one broken link that would make every downstream number wrong.
Self-check
In one sentence, what is adherence and what is it not?
Adherence is the comparison of actual state against what the schedule planned - 'did the actual match the plan' - and it is not a measure of how hard someone worked or how much they produced.
Define schedule, work and output adherence and say why they must stay separate.
Schedule adherence: present and in the planned state at the planned time. Work adherence: that scheduled time was actually spent on work. Output adherence: the work met the expected standard. They stay separate because the fix for each differs - a presence gap is scheduling/attendance, an output gap is about the work itself.
What is a grace window and why does it mean adherence is not an absolute truth?
A grace window is the tolerance before a deviation counts as a violation. Because adherence is computed from rules like grace windows and state mappings, changing the grace window flips the same behaviour between violation and non-event - so adherence is the output of configurable rules, not an absolute measurement.
Why is 'coverage' not the same as roster headcount?
Coverage is required-versus-actually-active staffing for a specific interval and skill. A rostered agent who is off-plan is not covering the interval, and a day with enough total people can still have a starved interval or an uncovered skill.
What is the single concept people most often get wrong about RTA, and the one-line correction?
That adherence is a productivity score. Correction: green on the schedule board means present and on-plan, not that the work is getting done - that is what the work and output layers are for; and a red state can be a legitimate reassignment.
Verified vs demo. The three-layer model, rules and coverage are demo-shown at rta.demo.timechamp.io on seeded data - the ideal place to make the model concrete. Exact labels, the output-standard derivation, figures and refresh latency are SME-validation-pending; confirm specifics with product. The concepts here are about the model, not about asserting any particular UI control.