Metric Definitions
Part of BI for operations management
Measuring backlog and throughput from operational data
Define open work and completed work, reconcile queue movements, and add age and case mix without confusing status events with cases.
Backlog is eligible work still open at a stated cutoff; throughput is eligible work completed during a stated period. Apply the same work unit and service boundary to both. Count entries and other exits so change in backlog can be explained.
Define what counts once
Begin with a stable case or request identifier. Decide when an item enters the measured process and which event marks completion. A status-history table may hold several events for one case, so counting rows measures updates, not cases.
Write rules for cancellations, transfers, duplicates and reopened work. A transfer may leave one team’s queue while remaining open for the wider service. A reopened case may start a new work episode or continue the original case, depending on the measure. State the rule and apply it consistently to entries, completions and open work.
Backlog vs Throughput: Key Definitions and Measurement Rules
- BacklogEligible work still open at a stated cutoff; must use consistent service boundary and work unit.
- ThroughputEligible work completed during a stated period; based on completion time, not intake.
- Cancellations & TransfersDefined rules required—transfers may be exit for one team but not for the wider service.
- Reopened WorkRule must state whether it restarts the case or continues the original episode.
Reconcile movements
For one bounded period, use this control:
Closing open work = opening open work + entries − completions − other exits.
Put defined re-entries and transfers on the correct side of the equation. A case crossing a team boundary may be an exit for one team and an entry for another, yet neither for the service as a whole.
In a hypothetical queue, 80 cases are open at the start, 35 enter, 30 complete and two are withdrawn. The expected closing count is 83. If the source shows 86, inspect case IDs, movements and cutoffs before changing the calculation.
| Measure | Count from | Check |
|---|---|---|
| Opening backlog | Eligible cases open at the start cutoff | Use the same service boundary as the closing count. |
| Entries | First qualifying entry or defined re-entry during the period | Count work units, not event rows. |
| Throughput | Approved completions during the period | Use completion time, not intake time. |
| Other exits | Withdrawals and transfers out of scope | Keep these separate from completed service. |
| Closing backlog | Eligible cases open at the end cutoff | Reconcile with movements and source records. |
Define cutoffs so each movement belongs to one period. A current-status table may not retain earlier states. Historical backlog then needs suitable event history or retained snapshots; today’s status alone cannot establish an earlier opening count.
Reconciling Backlog Movements Over a Period
- Opening backlogEligible cases open at start cutoff
- EntriesFirst qualifying entry or defined re-entry during period
- CompletionsApproved completions during period (use completion time)
- Other exitsWithdrawals and transfers out of scope
- Closing backlogOpening backlog + Entries − Completions − Other exits
Add age and work mix
A backlog total does not show whether old items are being resolved. Group open work by age since its defined entry point and by useful type or priority. State how pauses affect the age measure. A paused case can remain in the work inventory even if a separate service-time clock stops.
Report completions and arrivals for the same period and boundary. Cases completed this week often arrived earlier. A change in work mix can also make a raw completion count hard to interpret.
Little's queueing relationship connects long-run average work in a stable system, flow rate and average time in the system. It can serve as a sense check when the measures share a boundary and its assumptions are credible. Dividing one day’s backlog by that day’s completions does not predict an individual case’s wait.
Release a trustworthy series
Check the reporting time zone and latest complete period. Compare sample case IDs in each movement category with the operational record, then investigate unusually large changes and late corrections. Once the movements reconcile, a manager can distinguish a rise in entries, a change in completions and an administrative exit.
