Forecasting

A workforce forecast answers two questions that are usually collapsed into one. How many people, with what capabilities, will the work require. And how many will the organisation actually have, given who leaves, who moves and who can be hired in time.

Most planning exercises do the first thoroughly and the second by assumption. A business plan produces a headcount target, the target becomes the forecast, and the supply side is handled by assuming the current workforce persists and vacancies fill.

Neither assumption holds. People leave, internal moves create vacancies elsewhere, and hiring takes longer than plans allow. The gap between planned and actual headcount is almost always a supply failure rather than a demand miscalculation.

Attrition drives most of the difference between the workforce an organisation expects and the one it has, and it is usually modelled at the crudest available level.

An organisation-wide rate applied uniformly is close to useless, because attrition concentrates. It varies sharply by team, by manager, by tenure band, by location and by role type, and a forecast built on the average will be wrong in both directions at once: over-providing where attrition is low and under-providing where it is high.

• Model by segment: team, level, tenure and location, at whatever granularity your numbers support.
• Treat the first year separately, since early attrition behaves differently from established attrition.
• Distinguish regretted from unregretted, because replacing the second is not always required.
• Watch leading indicators, particularly engagement and unplanned absence by team, which move before resignations.

The last point is what turns a forecast from an extrapolation into something with predictive content. Historic attrition tells you what happened; a team whose engagement has fallen and whose absence has risen tells you something about the next two quarters.

A forecast that says a role will be needed in June is not actionable without knowing how long filling it takes.

Where a senior role takes four months from approval to joining, and the notice period adds two more, the decision date for a June requirement is December. A forecast delivered in April is a description of a gap rather than a plan to avoid one.

This is the most common practical failure in workforce planning, and it is arithmetic rather than judgement. Working backwards from the need date through joining, notice, offer, process and approval gives the date the requisition has to open, and that date is the output the business actually needs.

It also varies by role far more than organisations assume. Using one average time to hire across a volume role and a specialist one produces a plan that is early for one and hopeless for the other.

Headcount forecasts are easier and answer a narrower question. What usually matters is whether the organisation will be able to do particular things.

A team at full headcount that lost its two people with a scarce capability has no headcount gap and a real capacity gap. Nothing in a headcount forecast surfaces that.

Capability forecasting requires knowing what capabilities exist now, which most organisations do not record in any usable form, and being explicit about which ones the plan depends on. It is worth doing for a small number of critical capabilities rather than attempting it comprehensively.

• Revise on a cycle rather than annually, since a forecast set once and left is a plan that will be wrong in ways nobody checks.
• Record the assumptions, particularly attrition and time to hire, so a variance can be attributed rather than argued about.
• Compare forecast to actual afterwards, which almost nobody does and which is the only way the forecast improves.
• Give ranges rather than single numbers where the underlying variance is high, since a false precision invites plans that cannot absorb normal variation.
• Be explicit about what the forecast cannot predict, including business change, acquisitions and market shifts.

The third is the one that compounds. An organisation that never compares its forecasts to what happened repeats the same errors indefinitely, usually the same optimistic time to hire and the same understated attrition in the same teams.