Most post-market quality teams find out about a problem when it becomes a corrective and preventive action (CAPA). By then the issue has a file number, an owner and a due date, which means it also has a history. However, that history was almost always visible months earlier, sitting in complaint records, nonconformance logs and receiving inspection data.
A leading indicator is a measurement that moves before the outcome it predicts. In post-market quality, that means a metric that changes while a problem is still small enough to correct without a formal investigation. Lagging indicators tell a team what already happened. Open CAPA count, number of reportable events, audit findings from the last inspection: all of them are accurate, and all of them describe a period that has already closed.
The distinction matters more after launch than before it. Pre-market, the data set is small and the team knows every unit intimately. Once a device is commercialized, volume climbs and the installed base spreads across distributors and countries. Failure modes start showing up in places the design team never considered. Post-market quality work accumulates whether or not anyone has built a structure to catch it.
Lagging indicators are the ones most teams already report
Ask a quality manager at an early post-market company what they report to leadership, and the list is usually short. Open CAPAs, overdue CAPAs, complaints received, complaints closed, and whatever the last audit turned up. Those numbers answer a compliance question well and a prediction question badly.
Counting complaints is a good example. A monthly complaint count goes up when sales go up, and it goes down when a team gets slower at logging. Neither movement says anything about whether the device is behaving differently in the field. The count is a volume measurement wearing the costume of a quality measurement.
Regulators have been reasonably explicit about wanting more than counts. ISO 13485:2016 clause 8.4 requires organizations to determine, collect and analyze data on the characteristics and trends of processes and product, including data from feedback, suppliers, audits and service reports. Under the Quality Management System Regulation (QMSR), which took effect Feb. 2, 2026, that clause carries the force of FDA regulation for devices marketed in the United States. In Europe, Article 88 of the EU Medical Device Regulation (MDR) goes further and requires manufacturers to report statistically significant increases in the frequency or severity of incidents that do not meet the serious incident threshold. Annex III requires the post-market surveillance plan to name the indicators and threshold values used to make that call.
Both frameworks assume a manufacturer is watching movement, not totals.
Complaint rate trends, normalized and segmented
The first metric worth building is complaint rate rather than complaint volume, normalized to units shipped or to installed base, and tracked as a rolling three- or six-month figure rather than a monthly snapshot.
Normalization removes the growth signal. If complaints rise 40% while shipments rise 50%, the rate improved. Segmentation is what makes the number predictive. A complaint rate calculated across an entire product family will hide a problem confined to one lot, one distributor or one failure mode, because the good population dilutes the bad one. The same rate broken out by failure code, manufacturing lot, software version and geography will show a single failure code climbing while everything else holds flat.
Also worth tracking alongside the rate:
- Time from complaint receipt to investigation decision, tracked as an aging distribution rather than an average. A lengthening tail usually means the team is triaging by capacity instead of by risk.
- The share of complaints closed without an identified root cause. When unexplained closures climb, investigations are being resolved administratively rather than technically.
- Repeat complaints from the same account or the same clinical site, which often precede a formal report by weeks.
Service and returns data belongs in the same view. A device that comes back for repair or gets swapped under warranty rarely generates a complaint record, because the customer never framed the issue as dissatisfaction. Field service visits go uncounted for the same reason. Yet ISO 13485 clause 8.2.1 treats feedback from production and post-production activities as a required input to risk management, and service reports appear by name in clause 8.4. Teams that keep the service log in a separate system from the complaint log are leaving one of the sharpest available signals out of the analysis. Our guide to effective customer complaint handling covers how to structure intake so the records are usable for this kind of trending.
Nonconformances that never reach a CAPA
The most underused early warning data in a post-market quality system is the set of nonconformances that were correctly closed without escalation.
Every quality system has a threshold below which a nonconformance gets dispositioned, documented and closed. The decision is usually right on a case-by-case basis. Read as a series, though, those closures carry information that no single record does. Six rework events against the same part number across four months is a supplier or design signal, even if each individual event was minor enough to handle at the line. The same is true of concessions granted, deviations approved and use-related issues resolved at the point of care.
Two metrics turn that pile into a leading indicator. Recurrence rate by nonconformance code shows whether the same problem keeps reappearing after being closed. Escalation rate, the share of nonconformances that get promoted to a formal investigation, shows whether the threshold itself is drifting; a falling escalation rate alongside a rising recurrence rate usually means the bar for opening a CAPA has moved.
Getting this right depends on having a defensible view of what should really trigger a CAPA in the first place. A team that cannot articulate its threshold has no way to detect when the threshold has moved.
Supplier performance drift
Supplier problems announce themselves gradually. By the time a nonconforming lot reaches final inspection, the supplier's process has usually been degrading for several deliveries.
Purchasing controls under ISO 13485 clause 7.4 require monitoring and re-evaluation of suppliers, but the evaluation is often an annual scorecard exercise that captures a point in time. Drift is a rate-of-change question. Four measurements answer it:
- Lot acceptance rate by supplier, tracked over consecutive deliveries rather than annually. A supplier moving from 100% to 94% acceptance across six lots is a different situation from one that sits steadily at 94%.
- Certificate of analysis discrepancies and documentation errors, which tend to rise before physical quality falls, because paperwork is where an overloaded supplier cuts first.
- Supplier-initiated change notifications, counted and categorized. A sudden increase in notifications frequently signals a facility move, a sub-tier supplier change or a process revision the manufacturer has not yet assessed.
- On-time delivery variance, which correlates with capacity strain more reliably than the on-time percentage itself.
None of these require new data collection. Receiving inspection records, incoming material reviews and purchasing correspondence already contain all four. The work is in pulling them into a view somebody reviews on a schedule.
Setting thresholds before the data forces the issue
A metric without a threshold is a chart. Deciding in advance what level of movement triggers a review is what converts a chart into an early warning system, and it also produces the documented rationale that EU MDR Annex III asks for in the post-market surveillance plan.
Thresholds do not need statistical sophistication to be useful, though they do need to be written down before the number moves. A reasonable starting point for an early post-market team is a percentage change from a rolling baseline, set per failure mode, with a defined review action attached. Any complaint rate that rises more than 25% above its trailing six-month baseline for a given failure code goes to the quality review board that month. Any supplier whose lot acceptance rate falls two consecutive periods gets a supplier corrective action request. The specific numbers matter less than the fact that somebody committed to them while the data was quiet.
Cadence and ownership decide whether any of it holds. Quarterly management review is too slow to catch a trend while it is still correctable, and an unowned dashboard gets checked in the weeks nothing is happening and ignored in the weeks something is. A monthly review with a named owner for each metric, and a standing agenda item for anything that crossed a threshold, is usually enough at early post-market volumes. Leadership involvement tends to be the deciding factor, because the response to a crossed threshold often costs money before it saves any.
Feeding the result back into risk management closes the loop. ISO 14971 expects production and post-production information to be reviewed for relevance to safety and for effects on previously estimated risk. Trending data that never reaches the risk management file has done half its job. Our post on using post-market data in risk management walks through how that input actually gets documented.
Track post-market leading indicators in Greenlight Guru
Every metric described here depends on complaints, nonconformances, CAPAs, supplier records and risk files living in the same system, coded consistently enough that a trend can be pulled without a manual export.
Most early post-market teams have the records but not the structure. Producing a segmented complaint rate by failure code and lot turns into a two-week project, so it happens once a year for the management review and never in the month it would have mattered.
Greenlight Guru holds all of it in one connected quality system, with failure coding and traceability built into the records rather than added afterward. Trending becomes a report instead of a data assembly project.
Ready to see what post-market trending looks like when complaints, nonconformances and supplier data share a system? Get a demo of Greenlight Guru and we will walk through the reporting with your own process in mind.
Keep reading
If you are building out your post-market quality processes, these related guides go deeper on the specific components:
CAPA Report Template
