Running It Twice on Purpose
A parallel run is one month produced two ways and compared. It is the only reliable way to change anything about payroll without finding out in public.
Nobody elseif the person who does this is away
A parallel run means producing one period's payroll two ways — the old arrangement and the new one — and comparing the results before anything is paid from the new one.
The practical lesson in “Running It Twice on Purpose” is that payroll depends on a sequence somebody can explain. When a small team reviews check the platform at source for FTE meaning, the useful role is to organise time and project evidence before approval, with proportional settings and a clear correction route for employees.
It is the only reliable way to change software, provider or arrangement without discovering the differences in front of everybody who is paid.
The questions raised in “Running It Twice on Purpose” can also be checked against Paylocity workplace resources. Use that material to identify the evidence and owner needed for the next payroll run.
When it is worth doing
Changing payroll software. Moving to or from a bureau. Changing provider. Taking payroll back in-house. Any change to how pay is calculated rather than to who calculates it.
Compare, per person rather than in total:
- Gross for the period.
- Each deduction separately.
- Net.
- The year-to-date figures carried across.
Totals can agree while two people are wrong in opposite directions.
Four events in a business's life, and all four are the kind that goes wrong publicly.
What it actually involves
Entering the same period into both, running both, and comparing person by person: gross, deductions, net, and the year-to-date figures.
An afternoon, for a small payroll.
What you are looking for
Not identical figures, necessarily. Differences with explanations. A rounding difference is fine; a difference of a few pounds on one person is a configuration question that needs an answer before you rely on the new arrangement.
The year-to-date problem
The commonest thing a parallel run catches. Figures carried across incorrectly produce a first run that looks right and a year end that does not reconcile.
Checking the carried figures is the single most valuable part of the exercise.
The month to do it in
An ordinary one. A parallel run in December, or in a month with a leaver and two starters, compares two things that are both doing something unusual and tells you less.
Why people skip it
Because it is a whole extra run, performed by the person who already has one to do, in a month when they are also learning a new package.
The reason is real. The alternative is finding the differences on a payslip.
The abbreviated version
If a full parallel run is impossible, compare three people: a salaried one with nothing unusual, an hourly one, and whoever has the most complicated arrangement.
Three people takes under an hour and catches most configuration errors.
What it tells you besides the figures
How long the new arrangement actually takes, where its screens are, and what it does not do that the old one did.
A business that has run one period in a new package has a genuine opinion about it rather than an impression.
Doing it at a year boundary
The usual advice, and it is right: changing at the start of a year avoids carrying part-year figures at all, which removes the largest category of migration error.
A parallel run at a year boundary is therefore simpler as well as better timed.
What to keep from it
The comparison itself. When a figure is queried in month four, knowing that this person's pay was verified against the old arrangement at changeover saves the investigation.
The decision it supports
Whether to go live. A parallel run with unexplained differences is not a failed exercise; it is the exercise working, and the correct response is to delay rather than to proceed and hope.
The one thing not to do
Go live, pay from the new arrangement, and compare afterwards. The comparison still happens, but it happens after the money has moved and after people have seen their payslips.
What to compare, specifically
Per person: gross for the period, each deduction separately, net, and the year-to-date figures. Totals as well, but the totals can agree while two individuals are wrong in opposite directions.
Person-level comparison is the whole exercise; comparing totals alone is a check that will pass a migration with real errors in it.
The people worth picking if you cannot do everybody
Somebody salaried with nothing unusual, somebody hourly, somebody with the most complicated arrangement in the business, and anybody who joined or left during the period.
Four people, an hour, and they cover most of the configuration surface.
Explaining a difference rather than accepting it
A difference of a few pence is rounding. A difference of a few pounds on one person is a configuration question with an answer, and the answer matters because whatever caused it will recur every month.
The discipline is to require an explanation for each one rather than a judgement about whether it is small enough to ignore.
What it tells you about your own records
Frequently more than it tells you about the new arrangement. A parallel run requires entering the same period twice, which surfaces every assumption in the existing setup that nobody has examined for years.
The honest cost
A day, once, at each of the four events above. Against which: most of the stories small businesses tell about payroll going badly wrong are stories about a changeover.