Where a workforce plan breaks first

A workforce plan looks like a headcount number. It's really a chain of estimates that produces that number, and the number is only as sound as its weakest link. Somewhere in the model sits one assumption the whole thing rests on. When that one is wrong, the plan fails, and it usually fails around the point where you've already committed budget and told the board a date.

So the useful audit asks which line in the model breaks first, and what happens to everything downstream when it does. That's a question you can answer by walking the plan the way you'd walk a set of accounts: line by line, asking of each input where the number came from and how confident you actually are in it.

In practice three inputs do most of the breaking. The attrition rate you subtracted. The ramp time you assumed before a new hire is productive. The supply of the talent you plan to hire. Each is easy to set optimistically, because the optimistic version makes the plan look cheaper and faster. Each is testable. This is a method for testing them, and for reading the order in which their failures spread.

Most plans carry a single blended attrition figure, often last year's company average, applied evenly across every team. That figure is comfortable because it came from your own data. It's also the assumption most likely to be quietly wrong, for two reasons. It's usually a lagging number in a market that has moved, and it hides the variance that actually matters.

Attrition concentrates. It runs higher in the first 18 months of tenure, higher in the hottest skill bands, higher on teams that lost a manager. A blended 12% can sit on top of a critical team running at twice that, and the plan will never show it, because the average absorbed the spike. You budgeted backfills for the company. The hole opens in one function.

To pressure-test it, stop using one rate. Segment attrition by tenure band, by team, and by skill scarcity, then look at the segments that feed your most important roles rather than the blended top line. Ask a second question the average can't answer: is this rate stable, or is it drifting? A supportive external signal helps here. India's Net Employment Outlook was +68% for Q2 2026, the strongest reading since 2008, per ManpowerGroup's Employment Outlook Survey. When hiring intent is that high across the market, the poaching pressure on your scarce teams is rising whether or not last year's average has caught up to it yet.

Ask a plan when a new hire starts contributing and it'll often answer “when they start.” Two silent optimisms hide inside that. The first is that a filled seat equals productive output, which is never true on day one and rarely true by month three for a senior role. The second is that the seat fills quickly at all.

Both cost time the plan already spent. If the model needs 30 people delivering by the end of Q3, and each takes a real 4 months to reach full productivity, and the roles themselves take months to fill before the clock even starts, then the hiring had to begin in the previous quarter. Plans that assume instant ramp are describing a different quarter than the one they'll actually land in.

Pressure-test it with your own recent hires. Pull the last handful of people hired into the same role and ask their managers, honestly, when each became fully productive. That's your ramp curve. Then put the true time-to-fill in front of it, benchmarked by how scarce the role is rather than a blended average. The sum of those two is the real lead time between deciding to hire and having capacity, and it's almost always longer than the plan allows.

There's a compounding cost here that the ramp number alone misses. Every ramping hire consumes the time of an already-productive one. Onboard 10 people into a team of 15 and the 15 slow down while they teach. A plan that treats ramp as free for everyone except the new hire has understated the dip twice.

The third assumption is the one that gets waved through, because the numbers behind it look reassuring. India has one of the deepest technical talent pools in the world. A plan that needs to hire from it reads that depth and assumes the roles are fillable. The gap is between how large the pool is and how much of it is actually reachable for your specific role, at your seniority, in your city, at your price.

The aggregate market makes the trap easy to fall into. Deloitte and Nasscom estimate India's AI-talent demand will exceed 1.25 million by 2027, against a shortfall of around 50% in 2024, with only about 16% of IT professionals AI-skilled. A pool measured in the hundreds of thousands is real. But once you filter for the seniority you need, the sub-skill you need, the location, and the slice of that pool that's actually open to moving rather than merely employed, the reachable supply can be a small fraction of the headline. The plan sized the ocean. You're fishing one bay of it.

To pressure-test supply, narrow before you count. Take the pool down to your exact seniority band, sub-skill and geography, then separate the total supply from the active supply, and set both against the competing demand for the same people. That last step is what most plans skip. A pool of 10,000 with 50 employers hiring hard against it behaves like a much smaller one, and 82% of employers already report difficulty finding the skills they need, per ManpowerGroup. Supply you can't reach is just a comforting number in a column.

Auditing each assumption alone is only half the method. The reason a single broken input sinks a plan is that these three are wired together, and a failure in one loads the next.

Trace it in order. Attrition runs hotter than assumed on a scarce team, so you need more hires than the plan called for. Those hires ramp slower than assumed, so the productive capacity you were counting on arrives late even after the seats fill. And the extra hires draw on a supply pool thinner than the headline suggested, so time-to-fill stretches at exactly the moment you needed to move faster. Each failure makes the next one worse. A plan that would have survived any one of these in isolation breaks when the first knocks into the second.

This is why the audit reads the order, as well as the lines. Attrition tends to break first because it's the assumption you drew from your own comfortable average. Ramp breaks next, because the extra backfills you now need all have to climb the same curve. Supply breaks last and hardest, because it was the most optimistic to begin with and it's now carrying more weight than it was ever sized for. Find the first domino and you can usually see the rest of the row from there.

None of this requires rebuilding the model. It requires reading the one you have with a red pen and three questions. Which line is carrying the most weight? Which line came from the most comfortable assumption? And if that line is off by half, what moves downstream?

Do that at the plan stage and the corrections are cheap. You segment the attrition rate, add the real lead time in front of ramp, narrow the supply pool to what you can actually reach, and re-run the number. The headcount that comes out is less flattering and far more likely to survive contact with the market. The planners who close their hardest roles on time pressure-tested the load-bearing assumption while it was still a line in a spreadsheet, back when moving it cost nothing.