The workforce you forgot you had.

Ask a chief executive to describe their workforce and you will get a picture of the people they see. In a professional services firm, consultants. In a healthcare group, clinicians. In a manufacturer, the commercial team, because that is who sits near the executive floor.

Ask payroll the same question and you frequently get a different company. There is a field operation nobody mentioned. Overnight coverage at two sites. A seasonal population that trebles in summer. Forty people at a subsidiary acquired three years ago who are still on their own pay rules because unifying them was never anyone’s priority.

The gap between those two descriptions is where platform decisions go wrong.

Why the salaried half runs the evaluation

Because that is where the buyers sit. The CFO, the head of HR and IT run the selection, and all three work regular hours in one location. They assess the system through the interface they will personally use: the org chart, the reporting, the approval workflows.

The rules engine — which decides whether a night-shift differential stacks correctly with a weekend premium on a holiday — gets a demo slot and a checkbox. Nobody in the room will ever open it. The people who will are on shift, and were not invited.

So the evaluation weights the half of the workforce that is easy to model and discounts the half where the calculations are hard and the consequences of getting them wrong are immediate.

What it costs, concretely

A rules engine that cannot express a pay rule natively does not refuse. It gets configured around — a manual adjustment each cycle, a spreadsheet that holds the real schedule, a supervisor who knows to check something the system gets wrong. Those workarounds are invisible in a business case and permanent in practice.

They also carry exposure. Overtime calculated on base rate when a non-discretionary bonus should be in the regular rate produces an underpayment every cycle, quietly, at scale. Meal-break rules configured for the headquarters state and replicated outward produce penalty pay in the states where they differ. Neither surfaces as an error. Both accumulate.

The count that settles it

Before any vendor conversation, count properly. Not headcount — composition. What proportion of people are paid hourly. How many are scheduled rather than simply arriving. How many distinct pay rules exist, including the ones that apply twice a year. In how many states and municipalities work is physically performed. Whether anyone is covered by a collective bargaining agreement.

Writing that list down is most of the work, and most organizations have never done it. The rules accumulated one exception at a time, each locally reasonable, and the complete set exists only as distributed knowledge across a few long-serving people.

The list is also the single most useful artifact you can bring to a vendor. Ask them to configure the three worst items on it, live, in your language, in the room. How much is native, how much is a workaround and how much needs a services engagement is more predictive of your next five years than any other input you will gather.

Two workforces under one roof

The hardest employers to advise are not the most complex ones. They are the ones running two genuinely different workforces and treating them as one — a clinical population on rotating shifts alongside an administrative corps on salary, a plant floor alongside a commercial office, programme staff working overnight alongside a development team on standard hours.

Each half has real requirements that conflict. A platform chosen for either alone fails the other, and the failure lands every pay period rather than at review time. The remedy is not sophisticated: count both populations, write down the harder half’s rules in full, and put someone from that half in the evaluation with authority to reject a product. It costs a few days and it prevents the most expensive category of platform regret we see.

Where the exposure concentrates

The forgotten half is not only a cost problem. It is where nearly all wage-and-hour risk lives, because that is where hours are variable and rules are jurisdictional.

Three things account for most of it. Exempt classifications set once at hire against a federal threshold and never revisited against state thresholds that have since moved — expensive because the remedy is retroactive and applies to everyone in the role. Overtime calculated on base rate where a non-discretionary bonus or shift differential should sit in the regular rate. And time records that cannot substantiate what was paid, which matters more than either, because the employer carries the burden of proof on hours worked.

Records showing identical start and stop times every day are worse than sparse ones — uniformity invites the inference that they were produced rather than captured. That single point is the strongest argument for a real time system in an hourly workforce, and it is almost never the argument anyone makes when buying one.

What the seasonal case does to pricing

One practical consequence worth flagging, because it catches people. Where headcount swings across a year, a per-employee-per-month fee and a percentage of gross payroll behave very differently — and the proposal that looks cheaper in January can be the more expensive one across twelve months.

Neither structure is wrong. But comparing them requires modelling both against your actual payroll for a full year including the peak, and that is arithmetic nobody does when the workforce in their head is the steady salaried core.

The pattern underneath

Companies describe themselves as the business they intend to be, and buy systems for that description. The hourly edge is usually the part that was acquired, or grew sideways, or exists at a site the leadership team visits twice a year. It is nobody’s identity and it is a large share of the payroll.

The correction is unglamorous and it is free: before scoping anything, ask payroll to describe the company. Then reconcile that against the version in your head, and scope for whichever is larger.

The same reconciliation is worth running against your compliance footprint, incidentally, and it finds the same category of surprise. List every state where somebody physically works, from actual addresses rather than the org chart, and set it against your registrations. Companies doing that for the first time usually find at least one gap, and it is invariably attached to the part of the workforce nobody was picturing.

Neither exercise requires a consultant, a project, or a budget line. Both take an afternoon and a willingness to be told something inconvenient by your own data — which is, in fairness, the harder of the two requirements.

The reason to do it before a platform conversation rather than during one is simply that vendors cannot do it for you. They will scope against the description you give them, accurately and in good faith. If that description is the company you picture rather than the company payroll runs, the resulting configuration will be a faithful implementation of the wrong brief — and nobody involved will have made a mistake.