Six Sigma is often associated with manufacturing floors, but its structured approach to reducing defects and variation applies equally well to finance operations — from invoice processing and journal entry accuracy to month-end close cycle times.
The DMAIC framework — Define, Measure, Analyse, Improve, Control — provides a disciplined sequence for tackling process problems. In a finance context, the Define phase begins with a clear problem statement: for example, “Invoice processing errors currently run at four percent, resulting in supplier disputes that delay payment and damage vendor relationships.” This statement anchors the entire project; without it, teams drift into scope creep or solve the wrong problem.
During the Measure phase, the team collects baseline data. For invoice errors, this means categorising each error type — wrong amount, incorrect GL code, missing purchase order reference, duplicate entry — and plotting the frequency on a Pareto chart. In most cases, two or three error types account for seventy to eighty percent of all defects. The Pareto chart makes this visible to stakeholders who might otherwise insist that “everything needs fixing.” It does not; a small number of root causes drive the majority of pain.
The Analyse phase uses a fishbone diagram and process walk-throughs to identify why those dominant error types occur. In a recent engagement, we found that sixty percent of GL-coding errors traced to a single root cause: the chart of accounts had been restructured six months earlier, but the ERP’s default coding rules had not been updated to match. The Improve phase was straightforward — update the default rules, add a validation check at data entry, and retrain the three staff members who processed the highest volume of invoices.
The Control phase is where most improvement efforts fail. Without ongoing monitoring, the process drifts back to its old performance within weeks. A simple individuals control chart, updated weekly with the current error rate, provides an early-warning system. If a data point breaches the upper control limit, the team investigates immediately rather than waiting for the next quarterly review. Over six months, the invoice error rate in our example dropped from four percent to under one percent — a measurable, sustained gain achieved with no new software and no additional headcount.