Headcount Forecast

The plan line assumes nobody is ever missing

A headcount plan is usually drawn as a starting number plus the new roles, climbing in a straight line. That line quietly assumes every person who leaves is replaced the day they go. Nobody is. A role that opens today is filled in six or eight weeks, and in the meantime somebody else has resigned.

So at any moment there is roughly one time-to-fill’s worth of openings standing empty. That is Little’s Law, and it gives the size of the gap without any modelling: open roles on hand equal openings per month times months to fill. A team losing ten people a month with a two month hiring cycle is always twenty short. The gap does not close by waiting. It is the permanent price of not being able to hire instantly.

The forecast here is tuned so that its gap settles exactly where that law says it should, and the tool shows you that every seat in the gap is an open requisition rather than an abstraction.

New hires leave faster than everybody else

A growth plan fills its new roles with new hires, and in most organizations people leave far more often in their first year than after it. A forecast that applies the company’s average turnover to them undercounts the departures, and the requisitions needed to replace them.

So the file is measured twice: turnover among people who have been there a year or more, and the share of new hires gone within their first year. The second one needs care. Somebody hired four months ago has not failed to leave within a year; they simply have not been there for one yet. Counting them as stayers makes new hires look far more loyal than they are, often by a factor of two or three. The tool uses Kaplan-Meier, which counts each person for exactly as long as they have been observed.

A range, because who leaves when is random

Even with the rates exactly right, the number of people who happen to resign in any given month varies. So the forecast is simulated six hundred times and the likely range is drawn around the expected path: the band eight in ten simulated years land inside.

That range covers the randomness only. It does not cover the rates themselves changing, which over a year is usually the bigger uncertainty, and the tool says so rather than presenting a band as though it were a guarantee. The simulation uses a fixed seed, so the same file always gives the same range.

The things a roster can tell you, and the one it cannot

Turnover and first-year loss come from the file, with the working shown and either one replaceable with your own figure. People with a leaving date after the forecast starts are treated as known departures on those dates rather than as odds. Time to fill cannot be read from a roster; the Time to Fill Calculator measures yours from a requisitions export, and the Turnover Analyzer goes deeper on who is leaving and why.

What this does not do

It projects a workforce from its own recent history. It is not a budget, and it knows nothing about reorganizations, seasonal hiring, a market turning, or anything else the file does not contain. It is the honest version of the line in the deck, and the judgment about what to do with it stays with you.