The chair took 9 seconds to rise, and the pump that caused it carried the same 200 kg nameplate as the good ones

The order, and the check that decided the case

A European distributor of salon equipment was outfitting eighteen barbershops and placed a single order for 90 hydraulic barber chairs with a factory in Foshan. Cast-iron base, chrome-plated hydraulic pump worked by foot pedal with a locking lever, reclining backrest, adjustable footrest. The unit price was $186 and the order value was $16,740. Thirty per cent was paid as a deposit, and the balance of $11,718 was held against the pre-shipment inspection.

The hydraulic part of the buyer's specification ran to five lines. Pump: cast iron body, chrome plated, foot operated, with locking valve. Rated load 200 kg. Seat height range 550 to 750 mm. Pump: as the approved sample. Cycle life 100,000 cycles.

Three of those five lines could be checked in the factory in under a minute. The rated load was on the nameplate. The seat height range came off a tape measure against the pump in the free state and the pump under load. The pump was compared to the approved sample the buyer held on site.

The check the buyer had used for two years was one line, and it was on the incoming sheet at each shop: raise the chair to full height, lock it, wait thirty seconds, confirm it is still there.

It is a good check. It is the wrong check, and the whole case sits in the gap between those two paragraphs.

What a thirty-second hold test cannot see

The buyer's hold test passed on all ninety chairs. Nothing about a hydraulic pump can be wrong in a way that shows up in thirty seconds and then in three hundred thousand cycles.

Two things had changed on the shop floor, and neither of them had changed on the incoming sheet. The pump casting number on the body was 1180 on the first forty-eight chairs and 1120 on the last forty-two. The second pump was a smaller-displacement unit. It carried the same 200 kg nameplate rating, the same chrome, the same foot pedal, and roughly a third of the oil per stroke. Its rated cycle life was 50,000 rather than 100,000.

The mechanism is worth stating plainly, because it defeats a check most buyers would call thorough. A pump's nameplate carries its load rating, and both pumps carried the same one. The value that decides how fast a chair moves is the volume of oil delivered per foot stroke, and that is a catalogue number rather than a nameplate number. The buyer's check read the nameplate. The factory bought the catalogue.

Nothing about the substitution was hidden. The casting number is cast into the pump body in raised letters and is plainly visible when the chair is turned over. It simply was not a line on any sheet that anyone was using.

Measuring the rise, with a warm pump and a fixed definition

The measurement is simple and it has two traps, and both of them change the answer.

The first trap is temperature. Hydraulic oil thins as it heats, so a pump with a leak path can hold on a cold pump and fail once it is warm. Every chair was cycled thirty times up and down with no load before anything was measured. Without that step the results would have been a mixture of cold and warm readings and would not have meant anything.

The second trap is the definition of the stroke. Rise time means nothing unless the start and the end are fixed. It was defined as the elapsed time from the first movement of the seat at its lowest to the seat reaching its highest position, with the pump pedalled at a steady cadence and nothing on the seat.

The pedalling cadence was held constant across the lot, and the operator was the same person throughout. That matters: at a fixed cadence, rise time is a direct readout of how much oil each stroke delivers.

All ninety chairs were measured on this check, because it is quick, it does not damage anything, and it returns a yes or no answer. The attribute sampling under ISO 2859-1, general inspection level II, code letter D for a lot of ninety, sample size eight, was used for the slower checks and for the ones needing a load on the seat.

This is the part of a barber chair inspection that costs nothing and settles the argument fastest, because the split was not gradual and it was not random.

Two different failures, both of which get called the chair sinking

A chair that loses height has at least two causes, and they need different checks and lead to different conclusions.

The first is external. Oil passes the rod seal and appears on the chrome rod and on the base plate below it. This one is visible and it is checked with a piece of white tissue. Wipe the rod after fifty cycles. A mark on the tissue is a failure, and a film you can see without a tissue is a worse one.

The second is internal. Oil bypasses the piston seal inside the cylinder and returns to the reservoir. Nothing comes out of the chair, and the chair simply goes down. No wetness, no drip, no mark on the floor. It can only be found by measuring height under a stated load for a stated time.

The measurement used was this. With the chair at full height and the lock engaged, a calibrated 150 kg load was placed on the seat, and the seat height was recorded at zero minutes and again at ten minutes. The difference is the decay, and it is a number that can be argued about.

The descent figure in the table came out of the same session, and it is the one finding with a consequence that reaches past the factory. A chair on the approved pump took nearly four and a half seconds to come down with 150 kg on the seat. A chair on the cheaper pump took between 1.1 and 1.4 seconds for the same 200 mm of travel.

The faster descent was not caused by the smaller pump. It was caused by the release valve that came with it. The approved sample had a metered valve that restricted the returning oil. The second pump had a cheaper one that did not meter it, and the same 200 mm that took four seconds on one chair was completed in about a second and a quarter on the other.

What the report recorded, and what it did not

The report recorded the pump casting numbers present in the lot and the point at which the changeover fell. It recorded measured rise times, counted pedal strokes, metered oil delivery per stroke, measured height loss under a stated static load, the rod wipe results, and measured descent times. Forty-two of ninety chairs did not match the approved sample, and the changeover fell exactly between chair 48 and chair 49.

It did not establish the cycle life of either pump. That cannot be measured on site; it needs a rig and a lot of time, and nobody at the factory had one. It did not establish that every one of the forty-two pumps is identical in every respect, because eight of them were measured. It did not establish that the chairs will fail in service, and it did not certify conformity to any standard.

That last point came up because the factory offered a pump test report. It was a genuine report, and it was also not evidence for this order. It was written against EN 1494, whose scope is mobile or movable jacks and associated lifting equipment, and which explicitly excludes the lifting of persons. A chair pump is fixed into the base and it lifts a person, so the report was out of scope for the application however well the pump had been tested.

The report also did not state the four numbers the buyer needed: the number of cycles run, the load in kilograms, the dwell time, and the measured height loss at the end of the test. It stated the pump's rated load and the standard it was tested against, and it left the question of endurance entirely open. A report that is honest about what it does not cover is still useful. It is not proof of anything else.

The report was issued as an order verification statement against the buyer's documents, not as a test report, and it says so on its face.

Why the factory did it, and why nobody stopped it

The factory was not concealing anything, and on every line the buyer had written down, the delivered pump complied. It carried a 200 kg rating. It was chrome plated. It was foot operated with a lock. The thirty-second hold test passed on all ninety chairs.

The economics were ordinary. A pump of this type is a catalogue item, and the lower-displacement unit cost roughly half as much as the one in the approved sample. Across forty-two chairs that is a real number on the order.

And the buyer's own specification had set the trap, in the way it does. The order named the pump as the approved sample, which is a physical object on the buyer's desk and not a part number. That sample had been approved on its appearance and on the two values a shop could read off it: the nameplate rating and the finish. Nobody had ever asked the sample how much oil it moved per stroke, because there was no line in the order that made that a question.

The load rating and the height range were never at risk. The pump was always going to raise a ninety kilogram client, and it was always going to stop at 750 mm. The two things that actually mattered to the person using the chair every day were the two things nobody had written down.

What the specification should have said

Eight lines. Seven of them are numbers, and the eighth is a part number. Together they replace a check that says the chair still works with a set of values a report can actually defend.

• Name the pump by casting and nameplate rather than by appearance. Hydraulic pump: casting number 1180, nameplate rated load 200 kg. Any change of pump casting requires written approval before production continues, including a change made after the first article has been approved.
• Write the delivery volume as a number. The pump shall deliver not less than 12 cm3 of oil per foot stroke at rated pressure, metered over ten consecutive strokes at 10 bar into a graduated container. The sample pump measured 14.0 to 14.5 cm3; the unit supplied in its place measured 4.6 to 5.0 cm3.
• Write the rise time as an acceptance number with its measurement condition attached. Full travel from 550 to 750 mm seat height, unloaded, after thirty warm-up cycles at a steady pedalling cadence, shall be achieved in not more than 4.0 seconds measured from the first movement of the seat.
• Define a cycle and require the rating to be evidenced. One cycle is one full rise and one full descent through the full travel. The pump shall be rated for not less than 100,000 cycles at rated load, evidenced by a test report that states the number of cycles run, the load in kilograms, the dwell time and the measured height loss at the end of the test. A catalogue rating is not evidence.
• Write the pressure decay limit. With the seat at 750 mm, the lock engaged and a static 150 kg load applied to the seat, the seat height shall not fall more than 2 mm over ten minutes, measured on a warm pump after the warm-up cycles.
• Write the leakage criterion as an observation rather than a judgement. After fifty full cycles the chrome piston rod shall show no oil film to a white tissue wipe, and no oil shall be present on the base plate or on the floor beneath the rod.
• Write the descent rate. With 150 kg on the seat and the lock released, the seat shall descend over not less than 3 seconds. The release valve shall be of the metered type fitted to the approved sample, and shall not be substituted without written approval.
• Make the sample and the run the same part, and make the paperwork show it. The pump fitted to the production run shall carry the same casting number as the pump on the approved sample. The casting number shall be recorded per unit on the packing list, so that a changeover part way through a run is visible in the documents and not only on the pump body.

What the buyer did with the finding

The forty-two chairs from 49 to 90 were held and the other forty-eight were released against the balance. The factory refitted the approved pump to all forty-two, and the work took nine working days. A re-check on eight of the reworked chairs, on the same day, with the same operator at the same cadence, gave rise times between 2.4 and 2.9 seconds and height losses of 1 mm or less over ten minutes at 150 kg. The balance was released four days after the re-check report.

The specification was amended the same week, on the eight lines above. The new cycle line required a rig report, and the report that eventually arrived came back with the cycle count, the load in kilograms, the dwell time and the measured height loss written into it, because the amended order asked for those four things by name.

The casting number line went to the two other factories the distributor worked with. One of them asked which part number was cast into the body. It was the first question anyone had ever asked them about a hydraulic pump.