Customer complaints up: is it product quality or better logging?

This check applies to complaint and technical service records kept by a specialty chemical supplier or distributor, counted by period. It does not decide the cause of any single complaint, and safety or regulatory complaints follow their own reporting duties whatever the trend shows.

What does a rise in customer complaints look like in the records?

A rise in customer complaints looks like a higher count in the complaint log for one period than for the one before, often reported as a single number in a quarterly review. That number mixes four things: what happened to the product, what people chose to record, which customers the records cover and how much product was shipped. A quarterly count cannot separate them on its own.

Look at the log entry by entry before reading the total. Note the complaint type, product, lot, customer, who logged it, how it arrived (call, email, visit report, sales note) and the date it was logged versus the date the problem occurred. Then put the current quarter next to the previous quarters in the same fields.

Seasonality belongs in the comparison too. An IFSQN quality blog recommends comparing a month with the same month of the previous year, because complaint levels in one season are not comparable with another (IFSQN, 2014). For specialty chemicals, temperature-sensitive products, water-based fluids and outdoor coatings can follow the same pattern. A rise against the previous quarter that disappears against the same quarter of the year before is a season, not a trend.

What are the most likely causes of a rise in customer complaints?

The most likely causes of a rise in customer complaints are a real product or process change, a change in recording habits, a change in which customers are covered, and chance variation in small counts. Each leaves a different mark in the log.

Product or process change

A new lot, raw material, production site or formulation can produce more faults. The mark is concentration: the extra complaints share a product, a lot range, a fault type or a start date.

Recording habits

A new complaint form, a new CRM field, a campaign to "log everything" or a new technical service hire can raise the count with no change in faults. The mark is spread: more minor and informal entries, many from one person or one channel.

Customer coverage

A new distributor, a large new account or a change in which sites report can add complaints. The IFSQN blog reports 10 times the complaint level from one retailer compared with another for exactly the same product (IFSQN, 2014). A change in the period covered has the same effect: in UK regulator complaints data, one firm's 2015 H1 figures covered eight months instead of the usual six after it changed its financial year (FCA complaints data, 2015 H1 note).

Chance in small counts

With a handful of complaints per quarter, a doubling can happen by chance. The mark is a rise that sits inside the limits of a counts control chart.

How can you tell a product problem from a logging change in complaint data?

You can tell a product problem from a logging change in complaint data by checking where the extra complaints concentrate: product problems cluster on product, lot and fault type, while logging changes cluster on who logged, which channel and how severe. The table below decides which of the four causes the current rise points to:

The row to read first is who logged it. If most new entries come from a source that started during the period, the rise is at least partly a recording change, and the product question has to be asked of the complaints that would have been logged anyway. When the type mix concentrates on one lot range at the same time, both can be true. Split the log in two, the entries from the new source and the rest, and run every check on each half. A product change shows in both halves; a recording change shows mostly in the new one.

Should customer complaints be counted per volume shipped rather than as raw totals?

Customer complaints should be counted per volume shipped as well as in raw totals, because a raw count rises with shipments even when the product is unchanged. The IFSQN blog recommends using sales volumes to calculate the proportion of complaints for each product, and notes that some people use complaints per tonne or per 1000 litres (IFSQN, 2014). For bulk specialty chemicals, per tonne or per delivery is usually more meaningful than per unit.

An illustrative example shows the difference. Twelve complaints on 400 t shipped is 3 complaints per 100 t; eighteen complaints on 900 t is 2 per 100 t (illustrative). The raw count rose by half while the rate fell by a third.

Choose the denominator before looking at the numbers, and keep it fixed across periods. Use shipments in the period the complaints could have come from, not the period they were logged in, since a complaint about product delivered three months earlier belongs to that earlier volume. Keep the raw count as well. A rate that holds steady while the count climbs still means more customers are affected, which matters for service load even if the fault rate has not moved. The NIST/SEMATECH e-Handbook describes counts charts that track either the total number of defects or the average number of defects per inspection unit (NIST/SEMATECH e-Handbook, 6.3.3.1), and the second form fits a per-volume rate.

Why does a customer with no complaint records not mean a customer with no problems?

A customer with no complaint records does not mean a customer with no problems, because complaints record only the faults that someone chose to report and someone chose to log. The IFSQN blog notes that not all customers will complain when they find a problem, so complaint data will not capture all product faults (IFSQN, 2014). Some customers fix problems themselves, switch suppliers quietly or mention issues only to a sales contact who never logs them.

Count a customer with no records as unknown, not as zero. This matters most when a trend compares groups. If one region has a technical service engineer who logs every visit and another has none, the second region will look better in the log whether or not it is. The same distortion appears when a large account reports through a distributor whose complaints never reach your system.

Complaints are one input to customer satisfaction, not a measure of it. ISO 10004:2018, last reviewed and confirmed in 2023, gives guidelines for monitoring and measuring customer satisfaction, in a separate standard from ISO 10002 for complaints handling (ISO 10004:2018). Where coverage is uneven, mark the gaps in the trend report and turn them into requests to confirm with the customers concerned, rather than reading them as good news.

How few customer complaints are too few to call a trend?

Customer complaints are too few to call a trend when the change between periods sits inside the limits that chance alone would produce for that count. The NIST/SEMATECH e-Handbook models counts of defects with a Poisson distribution and sets counts chart limits at c plus or minus k times the square root of c, with k usually set to 3 and no lower limit when it comes out negative (NIST/SEMATECH e-Handbook, 6.3.3.1).

An illustrative calculation shows how wide those limits are. If a product averaged 4 complaints per quarter, the upper limit is 4 plus 3 times 2, which is 10 complaints (illustrative). A quarter with 8 complaints is twice the average and still inside the limit. On that count, the right report is that the data cannot show a change.

This does not mean small counts can be ignored. A single complaint about a safety hazard or a regulatory failure needs action on its own merits. It means that a small count cannot support a statement that quality has changed. Report the count, the limit and the period, and say plainly when nothing has changed beyond chance. A run of quarters above the average, or a cluster on one lot, can still be a signal worth checking.

When should a rise in customer complaints go to a quality specialist?

A rise in customer complaints should go to a quality specialist or the product's manufacturer when the extra complaints concentrate on one product, lot range or fault type after the logging and coverage checks, or when the count exceeds the counts chart limit. Send the evidence with it: the complaint list with type, lot, customer and logger, the shipped volume per period, the denominator you chose and the chart.

ISO 10002:2018 addresses analysing and evaluating complaints in order to improve the quality of products and services, and it was last reviewed and confirmed in 2023 (ISO 10002:2018). A specialist adds most when the product question needs retained-sample testing or a review of production records for the affected lots. Keep the recording and coverage findings separate from the product findings, so a logging change is not reported as a quality failure and a real fault is not hidden by one.

Lubinpla works from your own complaint records and site history: it keeps the scope and gaps of the records visible, interprets only the patterns the records can support, turns the rest into confirmation requests, and says so when nothing has changed.