Skip to content

Latest commit

 

History

History
46 lines (30 loc) · 2.52 KB

File metadata and controls

46 lines (30 loc) · 2.52 KB

Throughput

Throughput counts how many issues your team completes per time period. It answers "how fast are we actually delivering?" without requiring story points or estimates — just the count of items that made it to done.

What counts as completed

An issue is counted in a period if:

  1. Its resolution date falls within that period's date range.
  2. It has a changelog entry showing it transitioned into the started status at some point — issues closed without ever being worked on are excluded.

The same status boundaries used for cycle time apply here (workflow.cycle_time.started and workflow.cycle_time.completed).

Aggregation periods

Throughput can be aggregated by four frequencies:

Frequency Period length Anchoring
daily 1 day Forward from --from
weekly 7 days Backward from --to
biweekly 14 days Backward from --to
monthly Calendar month Forward from --from

Weekly and biweekly periods are built backward from the end date and then reversed. This means the last bucket always ends on today (or the --to date), which is useful for "last N weeks through today" reporting. Daily and monthly periods start from the --from date and move forward.

Statistical summary

Metric Description
Total items Sum of all completions across all periods
Avg/period Total ÷ number of periods
Median/period 50th percentile of per-period counts
Min / Max Lowest and highest single-period counts

Unlike cycle time, throughput values are not outlier-filtered. Every period is included in the statistics, including slow weeks and holiday periods. This gives an honest view of actual delivery pace over time.

Relationship to forecasting

Weekly throughput is the direct input to Monte Carlo simulation. The historical distribution of "how many items did we complete each week" is sampled randomly to model future delivery pace. See monte-carlo.md for details.

Tips

  • A 90-day window (--from / --to) captures roughly 13 weeks of weekly data — enough for stable Monte Carlo sampling without pulling in work that predates a team restructure.
  • If your team has had a significant change in size or process, shorten the history window so the forecast reflects the team as it is now.
  • Zero-throughput weeks (holidays, all-hands, etc.) are included in period statistics but are excluded when Monte Carlo randomly samples past weeks — the simulation only draws from weeks where at least one item was completed.