Stress-testing the plan

Deep-dive lesson · about 10 minutes · short quiz at the end

ccPlanning academy · forecasting · deep dive

Stress-testing the plan

A forecast is one number. Reality is a range — so test the plan against the range.

The problem

Every plan is built on numbers that will be wrong.

You staff to a forecast of volume, handle time and shrinkage. Not one of them will come in exactly as planned. The question isn’t whether the plan is wrong — it’s whether it survives being wrong.

Stress-testing is asking, before the day arrives: if this comes in off, how much does my roster move?

Why perfection isn’t the goal

Some error never goes away.

Contacts arrive randomly, so even a perfect model lands on a different number each time the same day is lived. That irreducible scatter — the Poisson floor — means no forecast is exact.

So the useful question shifts from “how do I remove the error?” to “given the error, is my plan robust?”

The four inputs

Four numbers move the roster.

Headcount isn’t driven by volume alone. It’s driven by four inputs together:

Volume, AHT, shrinkage, and the service-level target. An error in any one of them changes how many people you need.

The key insight

They don’t move it equally.

Swing in agents from the same-sized error in each input. Volume and AHT lead; shrinkage is the quiet third; the service target trails.

Volume & AHT

The two big levers feed the same workload.

Volume and AHT usually have roughly equal pull, because they multiply together into the offered load. A 10% miss on either moves the roster about the same amount.

That’s why handle-time drift is every bit as dangerous as a volume miss — and far more often ignored.

The quiet third

Shrinkage is the lever set from a guess.

A five-point shrinkage miss can cost as many heads as a sizeable volume error. And it’s the input most likely to come from a stale policy figure rather than measured reality.

If you buffer only volume and never check shrinkage, you’re guarding the wrong door.

The one people argue about

The service target moves staffing least.

80/20 versus 80/30 feels like a big decision, but it usually shifts headcount far less than a volume or AHT miss. The loudest debate is rarely where the real money is.

Useful to know — so you spend the argument time on the inputs that actually matter.

What to do with it

Buffer the biggest exposure, not evenly.

Once you know which input your plan is most exposed to, you size your contingency against that one — and your forecasting effort too — instead of spreading worry evenly across all four.

That’s the difference between planning with numbers and planning with judgement about the numbers.

Do it yourself

One interval, four what-ifs.

Take a busy interval — say 300 contacts, 240-second AHT, 80/20, 30% shrinkage. Flex each input in turn and watch the gross agents move: about ±5–6 for a 10% volume or AHT miss, a few for a five-point shrinkage change, barely any for the service target.

The staffing sensitivity calculator does exactly this — a tornado chart of your own numbers.

The takeaway

A good plan assumes the forecast is wrong.

Error is permanent, and not all of it costs the same. Find your biggest exposure, buffer that, and stop worrying evenly. The best plans hold up when the forecast misses — because they were built expecting it to.

Now test yourself ↓

1 / 11

Slides done? Here’s the same idea in a bit more depth — the part worth keeping.

In depth: testing a plan against the error it will meet

A staffing plan is a confident set of numbers built on four inputs that are all, to some degree, guesses: how many contacts will arrive, how long each will take, how much paid time you’ll lose to shrinkage, and the service level you’re planning to. None of them will land exactly. Stress-testing accepts that up front and asks a more useful question than “is the forecast right?” — it asks “if the forecast is wrong, how badly does the roster move, and which input is doing the moving?”

Why the inputs have different leverage

Volume and average handle time multiply together into the offered load, so a percentage error in either has almost identical effect — which is why a quietly drifting AHT is as dangerous as a volume miss, and far more often overlooked. Shrinkage is the quiet third lever: it’s applied near the end of the calculation, so it flows straight through to headcount, and a five-point miss can cost as many people as a large volume error. Worse, it’s the input most likely to be carried as a stale policy figure rather than measured from reality. The service-level target, by contrast — the thing planning meetings argue about most — usually moves staffing the least. Knowing that order for your own operation tells you where to put both your buffer and your forecasting effort.

From insight to action

The payoff of a stress test is targeted contingency. Instead of padding every number a little “to be safe,” you identify the one input your plan is most exposed to and buffer that, while leaving the inputs that barely move the roster alone. A tornado view — flexing each input in turn and ranking the swing in agents — makes the exposure obvious at a glance. It also reframes accuracy work: chasing the last point of volume accuracy is wasted effort if handle time or shrinkage is your real risk. The discipline is simple to state and powerful in practice: assume the forecast is wrong, find out which way hurts most, and plan for that.

The principle to remember: not all forecast error costs the same. Volume and AHT lead, shrinkage is the quiet third, the service target trails — so buffer your biggest exposure instead of padding everything evenly.

Stress-testing the plan