Two Definitions of Complete on a Contractual Dashboard
A monthly service dashboard is assembled from three regions. Region A records an inspection as complete when the field technician submits the digital form; Region B records it when a QA reviewer signs off. Last quarter the combined figure was 1,412 inspections against a target of 1,350, and an incentive payment of $86,000 was calculated on it. A sample of 60 records drawn from Region A found 22 with no QA review at the time of reporting, and 4 of those were later rejected by QA and returned for rework. The client contract defines a completed inspection only in an appendix, as all required checks performed and verified. The dashboard is due on the fifth business day; the program manager argues both regions are directionally right and that changing the method mid-contract would create a reporting discontinuity.
1. Identify the governing issue and explain how the definition of complete should be settled: who decides, and on what evidence.
Read the worked answer
The governing issue is not dashboard arithmetic. It is that two different operational definitions are being aggregated into a single contractual measure, and the measure is being labelled with a trailing-indicator name while Region A is actually reporting a leading indicator (submission).
The contract appendix supplies the controlling definition: performed and verified. Verification is the QA sign-off, not the technician submission, so Region A is measuring something real but calling it by the wrong name. That is a definitional ruling rather than a negotiation between regions. The contract owner - program manager with contract and quality concurrence - issues a written measurement definition specifying the population (which record types count), the event timestamp (QA verification date, not submission date), exclusion rules, and the evidence required to support a count.
Evidence needed before the ruling is published: a 100% re-basing of Region A's last-quarter records against QA status, not just the 60-record sample. The sample establishes that the problem exists (22/60 unmet at reporting, 4/60 later rejected) but it cannot size the correction. The client should acknowledge the interpretation in writing, because a documented correction is defensible and a silent change of method is exactly the discontinuity the program manager fears.
2. What should be published this cycle, and how should the $86,000 incentive question be handled?
Read the worked answer
Publish the reconciled figure and disclose the restatement separately from the run-rate. The dashboard should carry three lines: (a) inspections verified under the contract definition, (b) the previously reported count and the reason for the difference, and (c) the count still in the QA queue and therefore not yet countable either way.
The incentive should be flagged as provisional pending restatement, not presented as earned. The reason is not that payment is automatically wrong; it is that payment on a measure the program now knows to be mis-specified becomes a disclosure problem the moment it is not flagged. Of the 60 sampled records, 4 were ultimately rejected after reporting, which means some reported output was not merely unverified but non-conforming.
The five-day deadline is a delivery constraint, not a licence to publish a number the program cannot defend. If the full re-basing cannot be completed in time, publish with an explicit qualification and a committed date for the restated figure. A qualified number delivered on time is recoverable in the next cycle; an unqualified wrong number that is later corrected on someone else's initiative is not, because it makes every prior month suspect.
3. What structural change prevents recurrence, and what is the risk of deferring it one cycle?
Read the worked answer
Move the definition out of a document appendix and into the system. There should be exactly one state that counts as complete, and it should be a system-enforced, timestamped transition (QA verified) with submission and verification both visible but only one feeding the measure. Anything that depends on a human remembering which definition applies will drift again the moment a region changes its workflow.
Add a reconciliation control: each month, sample at least 30 records per region and compare reported status against underlying evidence, with the result published next to the number rather than buried. Assign a single named owner for the measure and place it under change control so that a future redefinition is a governed act with a restatement note, not someone's local improvement.
Deferring one cycle means at least one more incentive payment and at least one more client-facing report built on a number the program knows is wrong. The cost is not linear: if the deferral is discovered later, the question shifts from is this month right to how long have you known, and the entire historical series - not just the current month - has to be re-examined.