Primary Metric vs Secondary Metric: What to Fix and What Not to Break

Every DMAIC project carries one number it exists to move and a short list of numbers it must not damage on the way. The first is the primary metric. The second group is the secondary metrics. Teams get the first one wrong by choosing a dollar figure instead of a process output, and get the second group wrong by not choosing at all. Both mistakes show up at the same place: the Improve tollgate, when the sponsor asks what else changed.

This post sets the selection rules for each, gives worked pairings by industry, and covers why a program office cares that every project in the portfolio defines its metrics the same way.

The one metric the project exists to move

The primary metric is a direct output of the process under study, expressed as a number the team can track on a regular cadence. iSixSigma's definition puts it plainly: there can only be one, and it must match the problem statement. If the problem statement says invoices take 11.4 days against a 5-day standard, the primary metric is invoice cycle time in days. Not "AP efficiency." Not "late payment cost."

The goal statement is built from it. iSixSigma's guidance on metric selection gives the template: improve the primary metric from the baseline to the target by a date. If you cannot fill that sentence with one measure, one baseline, and one target, the primary metric is not yet chosen.

Three checks before you commit to it:

1. It is a process output, not a financial result. Savings follow from moving the process; they are recorded as benefits, not as the goal. See Six Sigma financial tracking for where the money lives.
2. It appears in the problem statement and the goal statement in the same unit over a comparable window. Days in one and percent in the other means two different projects.
3. It can be measured at least weekly during the project without a special study. A metric you can only collect once a quarter gives you two data points before close-out.

Secondary metrics catch what the primary metric hides

A process has more than one output. Push on one and others move. Cut cycle time and error rate often climbs. Cut cost and customer wait rises. The primary metric records the win; it says nothing about what paid for it.

Secondary metrics are the measures the project must hold or improve while the primary moves. iSixSigma's dictionary calls them consequential metrics: "the metrics that can be used to spot and measure potentially negative side effects resulting from process improvement efforts." They are tracked before, during, and after the project so the team can prove no collateral damage was caused. The same source notes that one is common but three or four complementary to the primary is better, and that outside life-or-death settings these metrics are always present. A project that claims it has none has not looked.

Two distinctions keep the list honest.

First, a secondary metric is a guardrail, not a bonus. Some practitioner sources split the two, reserving "secondary" for welcome side effects and "consequential" for damage to be prevented. The split is not needed on a charter. What matters is that each entry carries a constraint: hold at or above, hold at or below, do not exceed. A metric with no constraint is a wish, and iSixSigma warns against a secondary list that becomes the champion's extended wish list.

Second, a secondary metric is not the financial metric. Savings are calculated from the primary and reported separately as hard or soft benefits. Putting "annual savings" on the secondary list double-counts the goal. The distinction between the two kinds of benefit is covered in hard savings vs soft savings.

How to choose secondary metrics

Start from the primary metric and ask one question: what is the cheapest way to move this number that the business would refuse? Every answer is a secondary metric.

Cycle time falls fastest by skipping steps, so the guardrail is the quality check those steps protected: defect rate, rework, first-pass yield. Cost falls fastest by cutting service, so the guardrail is customer wait, backlog, or satisfaction. Yield rises fastest by loosening the inspection standard, so the guardrail is escapes to the customer or downstream returns. Throughput rises fastest by running unsafe or by building inventory, so the guardrails are recordable incidents and work-in-process.

Then write each one down with two numbers: the current level and the constraint. "First-pass yield 94.1 percent, hold at or above 94 percent." "Customer satisfaction 4.3 of 5, hold at or above 4.2." Without the constraint the team will argue at close-out over whether a half-point drop counts.

Stop at two to four. If the list runs to six, the problem statement upstream is unfocused, and the fix is to narrow the project, not to track more. The Six Sigma project charter, field by field covers the same discipline for the rest of the charter.

Pairings by industry

The same pattern repeats across settings. In each pair below, the second column is what the cheapest fix to the first would break.

• Manufacturing. Primary: first-pass yield on a line. Secondary: cycle time per unit, scrap cost, and recordable safety incidents. A yield gain bought by slowing the line and adding inspection is not the improvement the charter promised.
• Healthcare. Primary: emergency department door-to-provider time in minutes. Secondary: left-without-being-seen rate, 72-hour return visits, and staff overtime hours. Faster triage that pushes patients back out the door shows up in the second and third numbers.
• Transactional, such as accounts payable. Primary: invoice cycle time in days. Secondary: payment error rate, early-payment discounts captured, and duplicate payments. Speed bought by skipping three-way matching costs more than it saves.
• Software support. Primary: ticket resolution time in hours. Secondary: reopen rate, customer satisfaction score, and escalations to engineering. A closed ticket that reopens was never resolved.
• Logistics. Primary: on-time delivery percentage. Secondary: cost per shipment, damage claims, and driver hours. Meeting the window by expediting every load is a cost problem wearing a delivery badge.

None of these pairings is exotic. The work is choosing them in Define, recording the current level of each, and measuring them on the same cadence as the primary through Improve and into Control.

Recording the numbers so they survive to close-out

The primary metric needs three values: the baseline established in Measure, the current value updated as the project runs, and the target committed in the goal statement. The baseline is the one that gets lost. Teams overwrite it as data improves, and at close-out cannot show where they started. Refine the baseline during Measure, keep the refinement visible, and lock it before the Measure tollgate.

In SixGrid, the Charter tab holds the Primary Metric as its own card with baseline, current, and target fields and a progress bar between them. Secondary Metrics sit in a list directly below it, each with its own current value, so the tollgate reviewer reads the primary result and the guardrails on one screen without opening a spreadsheet. The end-to-end DMAIC guide covers where each phase's work lands in the project.

Why every project should define metrics the same way

A single project can get by with whatever metric the team likes. A program cannot. The iSixSigma metric-selection guidance makes the case for consistency across projects: when every charter defines one primary metric with a baseline and target, a list of secondaries with constraints, and a separate financial figure, the impact of the portfolio becomes legible without a survey.

Three things become possible that are not possible otherwise. Projects can be compared at selection time on the size of the gap they attack, in the same terms. A tollgate reviewer who has seen ten charters in the standard form can find the weak one in the eleventh in a minute. And leadership can ask "what did we break this year" and get an answer, because every project measured its guardrails rather than the ones that happened to think of it.

The form does the work. A blank template does not enforce it; a program office that rejects charters missing a secondary list does. Hold the standard from the first project and the portfolio inherits it. Start from the DMAIC hub for the phase-by-phase structure the metrics live inside.