Which Join to Build First

Workforce management projects are usually sequenced by module. Sequencing by join produces a working system earlier and fails more cheaply.

The usual order

Buy the suite, implement scheduling, then time, then absence, then integrate payroll.

Which means the payroll join — the one that matters — is built last, under time pressure, after the budget is spent.

And the first three months produce nothing usable, because a schedule with no time capture is a plan with no feedback.

The order that works

Identity first. Nothing else functions until one person is one record everywhere.

Then reference data.

Then the time-to-payroll join, because it is the one with consequences and the one that must be right.

Then schedule-to-time.

Then absence.

Then forecasting, which is an optimisation and can wait.

Why payroll first among the joins

It is the most regulated, the most consequential and the least forgiving.

Building it early leaves time to run it in parallel for several periods.

And it forces the rule ownership question immediately, rather than discovering at month nine that two systems calculate overtime.

A project that gets this join right and nothing else has still delivered something.

Parallel running

Two or three periods, compared per person rather than by total.

Totals match while individual figures are wrong in both directions, which is the trap.

Include the awkward cases: someone with corrections, someone at two sites, a shift crossing midnight, a mid-period leaver.

Sign off in writing before the old process stops.

What to do in the meantime

Keep the existing process running.

Do not switch off anything until its replacement has reconciled for a full period.

And do not cut over before a peak, which combines a new system's first-month errors with the month that has the least tolerance for them.

The measure of progress

Not modules live.

Joins reconciling: how many of the four produce a clean comparison, month after month.

That number tells you whether the system works, and it is the one a project plan never contains.

Compare parallel runs per person

Not by total.

Totals match while individual figures are wrong in both directions, which is the trap that makes a bad migration look successful.

Two or three periods, person by person.

Include the awkward cases deliberately: corrections, two sites, a shift crossing midnight, a mid-period leaver.

Sign off in writing before the old process stops.

Turn the boundary into a test

For a concrete product reference, this implementation example can help turn this boundary into a test. Verify the current behaviour with representative fields and record what reaches the destination.

Independent reference

For a thematic point of reference, see the Project Management Institute. Consult the source directly because technical, operational and legal details change.