Somewhere in most QMS document trees sits a customer satisfaction survey, sent annually, answered by almost nobody, summarized into a bar chart for management review, and forgotten. Average score: 4.2 out of 5, same as last year, same as the year before. Meanwhile, in the same twelve months, the company’s largest customer quietly moved 30% of their volume to a competitor, filed nine complaints, and put the company on a supplier watch list. The survey didn’t notice. The survey wasn’t built to notice.
Clause 9.1.2 is one sentence, and it’s worth reading as written: monitor customers’ perceptions of the degree to which their needs and expectations have been fulfilled, and determine the methods for obtaining, monitoring, and reviewing this information. Two things stand out. It says perceptions, plural and ongoing, not “conduct a survey.” And it hands you the choice of methods, which means it also hands you the obligation to choose methods that actually work.
Perception is the requirement, and perception is earned at the transaction
The clause’s note offers examples: surveys, customer feedback on delivered products and services, meetings with customers, market-share analysis, compliments, warranty claims, and dealer reports. The list is a hint that satisfaction data is mostly exhaust from things you already do, not a new data collection project.
Think about where a customer’s perception of you actually forms. It forms when the shipment arrives complete and on time, or doesn’t. When their quality engineer emails about a dimensional issue and gets a containment answer in four hours, or in four days. When your response to their corrective action request reads like an investigation, or like a form. No annual questionnaire outweighs fifty of those transactions, and the transactions are all generating records already: on-time delivery rates, complaint counts and closure times, returns, warranty claims, credit memos, scorecard grades from the customers big enough to issue them.
A monitoring method built from those signals has a property the survey never will: it updates while there’s still time to act.
The signals you already own
Complaints, counted honestly. The complaint log is the highest-grade satisfaction data you have, and its biggest threat is definitional. If “complaint” only means “formal written complaint on our form,” the phone call that ended with “just fix it this once” never enters the record, and the log reads calmer than the customer feels. Log the informal ones. A rising count after you widen the definition isn’t deteriorating satisfaction, it’s improving eyesight, and you should say exactly that at management review.
Customer scorecards. If your customers grade you, their scorecard is clause 9.1.2 monitoring that someone else pays for. Treat a slipping grade as an input requiring analysis and response, not a quarterly insult to be filed. Auditors increasingly ask for scorecards directly, because they know the scores arrive unfiltered.
Repeat business and share of wallet. Market-share analysis sounds corporate, but at small-company scale it’s one question: are existing customers giving us more work or less? Volume drifting away is the most expensive form of feedback, and it never fills out a survey on the way out.
Compliments. The standard lists them, and they’re not decoration. Knowing what customers value tells you which capabilities to protect when cost pressure comes hunting.
The survey can stay, but demote it to what it’s good at: an occasional structured check on relationship-level questions the transactional data can’t answer, like how easy you are to do business with, or whether the customer would recommend you.
Making it monitoring instead of collecting
The clause’s verbs are obtaining, monitoring, and reviewing, and the middle one is where systems fail. Data that’s collected but never trended isn’t monitoring, it’s archiving. Monitoring means the complaint rate has a threshold that triggers something, the OTD trend appears where someone who can act on it will see it, and a slipping scorecard opens an action with an owner rather than a sigh.
This is also where satisfaction data has to connect to the rest of the QMS instead of living in a folder of its own. A customer complaint that’s genuine nonconformance belongs in the NCR process, subject to clause 8.7 and, when warranted, a 10.2 corrective action. Satisfaction trends are a mandatory input to management review under 9.3.2, and “customer satisfaction: 4.2/5, no action” recurring across three consecutive reviews is the kind of entry that makes an auditor start pulling threads. If the signals live in the same system as the NCRs, CAPAs, and review records they should be triggering, the connections happen by reference instead of by heroics.
What an auditor actually checks
Auditors rarely challenge your choice of methods; the clause gives you that freedom. What they check is whether the loop closes. Show me your satisfaction data. Show me where it was reviewed. Show me a case where it was bad, and what happened next. The company with a modest complaint log, a visible trend, and two corrective actions traceable to customer feedback is in far better shape than the company with a glossy survey report and no evidence anything ever changed because of it.
There’s a one-afternoon self-audit here too. List your top five customers by revenue. For each, write down what you actually know about their current perception of you, and the date of the evidence. If the freshest thing you have on a top-five customer is a survey response from last spring, clause 9.1.2 isn’t the real problem. The real problem is that your largest sources of revenue can cool off faster than your instruments can detect, and the standard is just the second person to point it out.