Six Sigma

What it means

Six Sigma is a process improvement method aimed at variation in the result. It asks not what the average is but how often the process falls outside what the customer will still accept, and it counts defects per million opportunities.

It runs through five DMAIC steps: define, measure, analyse, improve, control. It differs from lean, which strips out waste and waiting, by leaning on data and statistics, and from kaizen by being a project with a team, a budget and an end.

Six Sigma is not a standard a company gets certified against: the certificates, known as green or black belts, are earned by people, not by the organisation. Nor does it replace internal control. Internal control catches errors and stops them before they travel further, whereas Six Sigma asks why the errors arise at all and how to bring their frequency down for good.

How it works in practice

In accounting and payroll the measures are the invoice error rate, the share of documents processed right first time and the length of the month-end close. The measure phase cannot run without a process map, because nobody knows which step breeds the error.

An improvement survives only if the project leaves a control behind: a check inside the system, a rewritten internal rule and a metric the process owner watches. Otherwise the process drifts back to its old behaviour within months.

Working with an accounting firm often shows that the largest variation arises not in the bookkeeping but on the client’s side: documents arrive at different times and in different quality. So we also measure the share of documents that come in after the agreed date or without supporting papers, and a change such as document digitisation with a single collection point can then be judged in numbers.

What to watch out for

Projects usually break on the definition of a defect and on the baseline measurement. If the state before the change was never measured and two people count defects differently, the claimed saving cannot be defended and never shows up in profit or loss.

In a small company with a few dozen invoices a month the statistics have nothing to draw conclusions from: the sample is too small and chance hides the variation. Training, project paperwork and measurement then cost more than they will ever return, and a plain standard operating procedure with a checklist achieves the same.