Instrument

The Unread Pile

What the part you do not review has been hiding

A percentage gets reviewed and the rest is assumed to look like it.

Sampling assumes what you are looking for is evenly spread. It usually is not, which is why the first full-coverage cycle on a delivered contract-verification system surfaced 2.3 times the discrepancy rate the sample had been reporting.

This calculator needs JavaScript to run its arithmetic. The inputs below are exactly the questions a discovery call opens with, so you can send us the numbers instead.

What you have

Items arriving each month that should be checked

Calls, settlements, invoices, inspections, submissions - whatever is currently sampled.

Share you actually review today, as a percentage

Two delivered cases sat at 2% and at about 15%. If you do not know, the real figure is usually lower than the policy says.

Discrepancy rate your sample reports, as a percentage

How often the reviewed items turn out to have something wrong with them.

Cost of one discrepancy that nobody catches

Leakage, a credit issued late, a compliance finding, rework. An order of magnitude is enough.

What it implies

Items a month nobody looks at

Discrepancies a year in that remainder, at the observed multiple

Annual cost of the ones nobody catches: -
For comparison: the same figure if the rate really were flat: -

The figure above, before and after the modelled change

The basis for this. Two delivered cases, and they measure different things. Quality review on calls moved from a 2% sample to 100% of calls with compliance-phrase detection recall at 94%. Separately, settlement checking moved from about a 15% sample to 100% and the first full cycle found 2.3 times the discrepancy rate the sample had reported. The 2.3× belongs to the second case, not the first, and this calculator applies it as an illustration of what uneven distribution does - not as a claim that your remainder behaves the same way. The flat figure is shown beside it so the difference between the two assumptions is visible.

Each asks you for a number you are estimating. The audit counts it instead.

Send us your numbers The service line behind it

You get a reply within one working day, from the engineer who would do the work.

Start here

A calculator estimates. A diagnostic counts.

Send us what you put into any of these four and we will tell you which of the numbers is worth measuring properly first.

Book the audit Run the numbers yourself

You get a reply within one working day, from the engineer who would do the work.