Find the work an agent never reached

The workflow reports a 98% successful completion rate. Every completed run has a useful output. Then a manager finds an eligible account that never entered the queue. It has no failed run, so it is absent from the success report.

Measure coverage from the work that should have been handled, not only from the work the agent attempted. Define the eligible population independently, reconcile each unit to a disposition and keep unprocessed work visible. Execution success and business coverage are separate questions. A reliable process needs both.

Define eligibility before looking at the output log

Start with the business obligation. For example, the workflow may need to prepare a brief for each eligible account meeting in an agreed calendar window. The eligible set comes from the authoritative meeting population and exclusions, not from the list of briefs already generated.

Write the unit and cutoff precisely. One account with three meetings may require one shared brief or three separate preparations, depending on the service agreement. Cancelled meetings, internal tests and unsupported account types need explicit treatment. Otherwise two teams can produce different denominators while both believe they followed the rule.

An exclusion is a business decision, not a way to improve a metric after the result is known. Preserve the reason and applicable rule. If the source cannot establish whether a unit qualifies, put it in an eligibility-unknown group instead of quietly excluding it.

Reconcile a fictional preparation cohort

Suppose a fictional weekly source list contains 120 meetings. Ten are confirmed cancellations and ten are internal meetings outside the agreed scope. That leaves 100 eligible preparation units under this example's one-brief-per-meeting rule.

At the review cutoff, 70 have accepted briefs, ten have drafts awaiting review, eight are blocked by missing identity, seven failed during processing and five never entered processing. These dispositions sum to 100. The team can see both the work in progress and the otherwise invisible five.

If the workflow's execution report includes only 87 completed attempts and 80 of those attempts succeeded technically, it answers a different question. It may also count retries. Do not divide that numerator by the business denominator without reconciling attempts to unique preparation units. A successful generation is not necessarily an accepted, timely brief.

The first management action may concern the five unprocessed units, but not because five is universally unacceptable. Their consequence matters: are they tomorrow's strategic meetings or events that no longer need preparation? Coverage makes the question visible; the owner still decides priority.

Distinguish late arrivals from missed work

Some eligible records appear after the planned extraction. A meeting created Tuesday for Wednesday could be genuinely new work, while a Monday meeting imported late may reveal a source delay. Preserve when the event became eligible and when the workflow could first observe it.

Agree how late arrivals enter the service. A supplemental run, a manual preparation route or a narrower timing promise may be appropriate. Do not rewrite the original cutoff to make the initial run look complete. Show the original cohort and the subsequent additions separately, then reconcile the final disposition.

This is also why a source's recent synchronization timestamp is insufficient. A fresh sync can contain old or incomplete business events. Coverage depends on the actual eligible population and the source's known limits, not the freshness of a status badge.

Keep exclusions from becoming a blind spot

A workflow can report perfect coverage by narrowing its scope until only easy work remains. Narrow scope may be sensible, but leadership should see the business work left outside it and how people handle that work. Managed coverage and total organizational coverage are related measures, not synonyms.

When eligibility changes, identify which version applies to each cohort. Do not compare last month's broad population with this month's narrow one and attribute the higher completion rate entirely to improved performance. Show the population change and the operational result separately.

Sample the edges. Inspect a few included cases and a few exclusions, especially records near a date boundary or with ambiguous identity. The aim is to check that the denominator represents the intended work, not merely that every row has a status.

Make the reconciliation an operated responsibility

Assign someone to own the population definition and someone to operate the reconciliation. They may be the same person in a small team. The receiving business owner should confirm that the exclusions and deadlines still fit the work people depend on.

RevTech inspects and prepares work within agreed scope. The current RevOps description does not claim universal inspection of every record. Ask which source establishes eligibility, how unsupported or unobserved work is surfaced and what evidence the review receives. A provider's healthy run history should be one input to that conversation.

Start with one recent workflow period and independently reconstruct its eligible units. Match them to accepted, pending, blocked, failed and unprocessed states, adapting the categories to your process. The most useful finding may be a record the agent never knew it was supposed to handle.

Sources and scope

Original coverage-reconciliation method and fictional cohort, reviewed October 30, 2026. Current RevTech RevOps and Sales Agent descriptions establish in-scope inspection boundaries. The broader agent-metrics article remains the canonical overview; this article develops completeness specifically.