What Engineers Can Learn From People Who Repair Things

New products are wonderfully persuasive. Everything is clean, aligned, charged and accompanied by a photograph of somebody smiling beside a houseplant. At that stage, almost any design can appear competent.

Repair begins after the flattering lighting has gone.

The person opening a laptop, appliance, wheelchair, sensor node or old radio is dealing with the product as it actually exists: dusty, strained, intermittently faulty, missing a screw and carrying the consequences of choices made years earlier. They see which connector fatigues, which seal traps moisture, which battery swells, and which tiny plastic clip was apparently designed to defend the Crown Jewels.

Engineers should pay more attention to them. Not as an afterthought in a warranty process, but as people with a particularly clear view of whether a design deserves to survive.

Failure is a better teacher than a launch event

A product rarely fails in the neat, single-variable way imagined in a design review. It may be dropped, left in a hot car, cleaned with the wrong thing, charged from an indifferent power supply, or used daily by someone who has neither the time nor inclination to read page 47 of a manual.

Repairers encounter the accumulated evidence. A cracked solder joint may point to repeated flexing. Corrosion around a connector may reveal an inadequate gasket, poor drainage, or a material pairing that was fine in a CAD model but less charming after two winters. A failed power regulator may be the visible casualty of a marginal thermal design upstream.

That feedback is valuable because it connects a fault to a mechanism. “Units are failing” is a management problem. “The cable exits the enclosure without strain relief, transfers bending into the PCB, and fractures the copper near the connector” is an engineering problem with a path to a solution.

A sensible failure-analysis loop should include repair data: common replaced parts, time to diagnose, time to access, repeat failures, unavailable spares and failures that make economic repair impossible. Warranty returns alone are not enough. Many people simply replace an item, put it in a drawer of electrical regret, or live with the fault because the repair process looks like a small hostage negotiation.

Modularity is not just a nice diagram

Engineers often talk about modularity in terms of development speed and product variants. Those matter. But modularity has a more human benefit: it lets a fault remain local.

If a battery, display, cable, pump, sensor or control board can be replaced independently, one failed part need not turn an otherwise useful product into waste. That requires more than dividing a block diagram into coloured rectangles. Physical access, connector cycles, calibration requirements, sealing, software pairing and spare-part availability all determine whether a module is genuinely replaceable.

Take a temperature-monitoring device. If the sensor is likely to be damaged or drift outside specification, a replaceable probe can be a sound decision. But only if the system can recognise the new probe, apply the correct calibration data where needed, and clearly record the change for traceability. A detachable part that requires factory-only tools, unpublished procedures and a mystical sequence of button presses is modular in the same way a locked door is technically an entrance.

There are real trade-offs. Sealed assemblies can improve ingress protection, reduce size and simplify manufacturing. Connectors add cost, take space and can become failure points themselves. The answer is not that every product should be built like a Victorian steam engine, with every component visible and invitingly greasy. It is that engineers should decide, explicitly, which parts will fail first and what a reasonable recovery looks like.

Diagnostics are part of the product

A repairer’s first job is often not fixing the thing. It is working out what is wrong without replacing half of it in hope.

Products can make this dramatically easier. Provide meaningful error codes, test points where appropriate, service documentation, logs that survive a reboot, and a way to distinguish a power fault from a communications fault or a sensor fault. In embedded systems, a watchdog reset counter and a recorded brownout event can save hours of guesswork. So can a clear indication that a measurement is implausible because the probe is open-circuit, rather than merely presenting a number with enormous confidence.

Good diagnostics should serve ordinary users too. “Device error 34” is not helpful unless error 34 has the decency to explain itself. A message such as “battery not charging: check cable and port for debris” gives somebody a useful first step. It also avoids the traditional support workflow in which everyone restarts the device three times and hopes morale improves.

This is particularly important when access to repair is difficult. Disability has made me alert to how much a supposedly small maintenance task can depend on grip, reach, vision, dexterity, transport and energy. A design that needs a strong hand, a hidden latch and twenty minutes on the floor may be serviceable on paper while being effectively unserviceable for plenty of people. Repairability is partly about parts and tools, but it is also about who can realistically use them.

Design for the person holding the screwdriver

Repairers work under constraints engineers can easily overlook. They may not have the original jig, the ideal lighting, a pristine workbench or a spare afternoon. They need screws that are visible and standard, cables with enough slack to disconnect safely, components that can be reached without destroying the case, and instructions that do not assume prior knowledge of the designer’s private mythology.

This does not mean treating users as careless or repairers as heroic scavengers. It means accepting that products have lives beyond the factory. They are carried, cleaned, shared, adapted, knocked about and occasionally confronted with a butter knife because somebody has misplaced the proper tool. We can disapprove, certainly. We can also design with enough grace that the product survives the encounter.

The best repair-friendly designs are not necessarily the ones with the most screws or the fewest adhesives. They are the ones whose likely failures are understandable, contained and recoverable. That is good engineering, good economics and, increasingly, basic respect for the people who bought the thing.

Repairers do not get to judge a product at its launch. They judge it after ordinary life has had a go at it. Their verdict is worth hearing, because it tells us whether we designed an object to be owned for years or merely admired until the first inconvenient fault.