How to Calculate DPMO and Sigma Level (Worked Examples)

Sigma level is the number the entire methodology is named after, and it's calculated from a single upstream metric: defects per million opportunities, or DPMO. If you can count defects and define what an "opportunity" for a defect actually is, you can calculate both by hand in a few minutes — no statistical software required. The part that trips people up isn't the arithmetic; it's deciding what counts as a unit, an opportunity, and a defect before you start counting.

This walkthrough covers the formula, two worked examples at different scales, how DPMO converts to a sigma level, and the definitional mistakes that quietly produce a wrong answer even when the math is right.

The three numbers you need first

Before any calculation, define three things clearly enough that someone else could recount them from your definitions and get the same answer:

• Unit — the single thing you're inspecting: one invoice, one assembled bracket, one customer call, one shipped order.
• Opportunities per unit — the number of distinct ways a single unit could have a defect. An invoice might have five: wrong customer, wrong amount, wrong date, wrong tax code, wrong line-item total.
• Defects — the actual count of things that went wrong, not the count of units that had at least one thing go wrong. A single invoice with two errors on it counts as two defects, not one.

That last distinction — defects versus defective units — is the single most common source of a wrong DPMO. A defective unit is a unit with one or more defects; a defect count adds up every individual error, even multiple ones on the same unit. DPMO is built on the defect count, not the defective-unit count.

The DPMO formula

Once you have those three numbers, DPMO is one formula:

DPMO = (Total defects ÷ (Units × Opportunities per unit)) × 1,000,000

The inner term — units multiplied by opportunities per unit — is your total opportunities for defects across the whole sample. Dividing defects by that total gives a defect rate expressed as a plain fraction; multiplying by one million just rescales it into a number that doesn't require writing out decimal places, which is the only reason the "million" is in there at all.

Worked example 1: a manufacturing line

A line inspects 1,000 units. Each unit has five defined opportunities (five weld points that could fail inspection). Inspectors log 37 total defects across the batch.

1. 01Total opportunities = 1,000 units × 5 opportunities = 5,000.
2. 02Defect rate = 37 ÷ 5,000 = 0.0074.
3. 03DPMO = 0.0074 × 1,000,000 = 7,400.

7,400 DPMO means that, scaled to a million opportunities, you'd expect about 7,400 defects. That's the number you'd plug into a sigma conversion next.

Worked example 2: a service process

A claims team processes 250 claims a week. Each claim has four opportunities for error: wrong policy number, wrong payout amount, missed required document, and wrong processing code. Over four weeks (1,000 claims total), an audit finds 18 defects.

1. 01Total opportunities = 1,000 claims × 4 opportunities = 4,000.
2. 02Defect rate = 18 ÷ 4,000 = 0.0045.
3. 03DPMO = 0.0045 × 1,000,000 = 4,500.

Same formula, same result whether the process makes physical parts or processes paperwork — which is the point of using DPMO in the first place. It gives you one comparable metric across completely different kinds of work, as long as your opportunity definition is honest and consistent from one measurement period to the next.

Converting DPMO to a sigma level

Sigma level is a conversion of DPMO onto the standard normal distribution, using the long-run 1.5-sigma shift that's conventional in industry reporting (the idea, popularized through Motorola's original Six Sigma work, is that processes drift over time, so the shift builds in a margin rather than assuming a process stays perfectly centered forever). You don't need the underlying statistics to use the conversion — a standard DPMO-to-sigma table, or a calculator built on the same formula, does it directly. As a rough sense of scale:

• About 690,000 DPMO corresponds to roughly 1 sigma.
• About 308,000 DPMO corresponds to roughly 2 sigma.
• About 66,800 DPMO corresponds to roughly 3 sigma.
• About 6,210 DPMO corresponds to roughly 4 sigma.
• About 233 DPMO corresponds to roughly 5 sigma.
• About 3.4 DPMO corresponds to roughly 6 sigma.

Using that scale, the manufacturing example above (7,400 DPMO) lands just under 4 sigma, and the claims example (4,500 DPMO) sits a bit above 4 sigma. Rather than interpolating a table by hand, our DPMO and sigma-level calculator (/tools) takes the same units, opportunities, and defects you just worked through and returns both numbers instantly — useful for checking your own arithmetic or for running the calculation repeatedly as new data comes in.

Mistakes that quietly throw off the number

• Counting defective units instead of total defects. If three invoices each have two errors, that's 6 defects, not 3 — using the smaller number understates your true defect rate.
• Changing the opportunity definition between measurement periods. If you add a sixth inspection point halfway through a quarter, your DPMO trend isn't comparable anymore unless you recalculate the earlier periods the same way.
• Inflating the opportunity count to make the sigma level look better. More opportunities per unit mechanically lowers DPMO for the same defect count, so the opportunity count needs to reflect real, distinct failure modes — not be padded to flatter the result.
• Treating DPMO and DPU (defects per unit) as interchangeable. DPU only tracks defects against units, ignoring how many ways each unit could fail; two processes with identical DPU can have very different DPMO if their opportunity counts differ.

Where this fits into a project

DPMO and sigma level typically get calculated twice in a DMAIC project: once in Measure, to establish a defensible baseline before you go looking for root causes, and again in Control, to show the sigma level actually moved after your fix. If you haven't structured a full project around this metric yet, our guide to running your first DMAIC project end to end (/blog/how-to-run-your-first-dmaic-project-end-to-end) walks through where baseline and post-improvement measurement fit against the other phases. For the definitions behind the two terms in this guide, see our glossary entries on what DPMO is (/blog/what-is-dpmo) and what sigma level means (/blog/what-is-sigma-level).

Our free White Belt (/courses/white-belt) introduces DPMO, sigma level, and the rest of the core measurement vocabulary at no cost, with the same timed, closed-book exam format used across every Averon level. Green Belt (/courses/green-belt) goes further, applying this calculation inside a full simulated improvement project where you define opportunities, collect a baseline, and report a sigma shift the way you would on a real process.