Hard limits survive every draw
The algorithm builds a capacity network between people and tasks. A person-to-task connection has capacity one, an excluded connection is absent, and each side has its own entered limit. A maximum-flow calculation fills as many task places as possible. No successful edge violates these constraints, including in a partial result. Re-generating can change which legal allocation is selected, but cannot make an impossible configuration complete. If a job needs more different people than remain eligible, its headcount cannot be filled by repeatedly assigning one helper.
Randomized allocation is not a workload guarantee
Secure browser randomness provides randomized connection costs when multiple legal allocations exist. This is not a uniform lottery over every possible full assignment, and people with different exclusions or limits do not have equal chances for each task. The tool prioritizes coverage, not equal task counts, equal minutes, difficulty, skills or preferences. Two short tasks can be easier than one long task. Agree on sensible limits and task definitions before drawing, and review the actual workload with participants afterward rather than treating task count as measured effort.
Keep the complete result
Copy includes task assignments, unfilled headcounts, eligible-person totals and each person’s used and remaining task capacity. CSV exports one row for every required task place, including blank-person rows explicitly labeled unfilled. Printable output includes every task and the person summary, even when the on-screen lists scroll. The CSV uses quoted fields and protects formula-leading names for spreadsheet use. These are downloadable snapshots; exporting does not send messages, create project-management tasks, synchronize another application or track whether the work was completed.