the craft · first draw

The four questions before you fix anything

Diagnosis is the oldest tool, and it is not made of steel.

A method for thinking, not instruction in any trade. Nothing on this page is a repair manual, and no procedure here is offered as code or standards guidance, because none is claimed. Unplug anything before you inspect its cord or plug, never work a live circuit, and leave damaged cords, wall wiring, gas appliances and fuel systems to a licensed electrician, plumber or mechanic — not to a first attempt. Nothing on this page is sponsored.

Study II — The Craft.

the mechanism

Never fight the whole machine at once.

There is a man who has replaced every part of a machine except the one that was broken. He is tired, he is poorer, and the machine still does not run. Every one of us has been him, and the cure was never a better tool. It is the oldest tool in every trade and it is not made of steel: diagnosis, which is nothing more impressive than four questions asked in the right order before anything is taken apart.

Under all four sits one habit, and it is worth naming before the questions themselves, because everything else is a way of practising it: isolate the stage. Do not ask what is wrong with the machine. Ask which part of the machine can be proved innocent, then prove it, then go to the next one. Each test you can devise should cut the field of suspects roughly in half rather than confirming a hunch, and a fault with nowhere left to hide will show itself without being hunted.

That is why the second question below does the most work. Most things that do anything at all can be read as three parts: something feeds it, something does the work, and something tells it when — supply, mechanism, control. This is a decomposition and not a law of physics; a hammer has no control stage and needs none. But wherever it fits, it converts a vague and infuriating problem into three suspects who can each be questioned alone, which is the difference between an afternoon and a weekend.

the four questions

In this order, before a tool comes out.

The order is not decoration. Each question narrows the field the next one has to search, and the last one gives you permission to stop.

  1. What changed?

    Things do not break on their own — that is the great myth of the frustrated. Something was moved, drained, worn, updated, or done. Usually last, and usually by someone. So the last event before the failure is your prime suspect: the lamp went dark after the furniture was moved, the car stumbled after the fuel stop. Begin at the last footprint.

    Hold it as a hypothesis rather than a verdict, because coincidence does happen and a part that fails the same week you serviced something else will happily take the blame. And ask the question honestly, because often enough the answer is: I did. That is not shame. That is your best evidence, and the man who cannot say it spends the afternoon investigating a machine instead of his own Tuesday.

  2. Is the trouble in the supply, the mechanism, or the control?

    Name the three parts before you touch any of them, and the suspects are already standing in a line. Take the dead lamp: supply is the wall outlet, mechanism is the bulb and its socket, control is the switch. Three questions, and each can be asked alone.

    Unplug the lamp first — a suspect stays out of the circuit while you work around it. Now plug something else into that outlet; not if the outlet is scorched or warm, because that one is an electrician’s and not a test point. If the second thing fails too, the trouble is upstream of the lamp entirely. If it works, move the lamp’s bulb into another lamp. Two small tests, no tools, and half the problem is already behind you.

  3. Can you make it fail on purpose?

    That sounds like madness and it is in fact mastery — where it is safe to summon. A fault you can produce at will is a fault you understand, and the intermittent fault that will not appear while you are watching is the one that eats whole weekends.

    Take the flashlight that dies in your pocket and works on the bench: grip it and twist where the barrel threads meet, and if it flickers you have caught the loose contact in the act. For the fault that comes and goes, change one condition at a time — heat, cold, vibration, load — and never more than one, or you will learn nothing from the result. The safety question comes before the test and not after it: a cord or a plug you suspect gets unplugged and looked at, never held and worked while the current is in it, and anything involving fuel, gas or a live circuit is not yours to reproduce at all.

  4. What does the failure cost, and what does the fix cost?

    Count the money, but count the hours and the risk as well. Not everything broken deserves repair, and the question is not whether you could fix it.

    An old rule of thumb serves here, and it is a rule of thumb rather than anybody’s standard: when the repair costs more than half of replacement, and the thing is past half its expected life, let it go. A one-dollar washer in a good faucet — always. A new compressor in a dying refrigerator — think hard. Knowing when not to repair is also craft; the man who fixes everything owns a museum of half-dead machines and no free Saturdays.

three machines, one method

The same four questions, three times.

  • 01 The lamp that will not light

    The cleanest case, because all three stages are visible and none of them needs a tool. Unplug it. Question the outlet with something else, question the bulb in another lamp, and whatever survives both is where the switch and the cord live. If the answer turns out to be the cord or the plug, the job stops being a repair and becomes a replacement or a trip to an electrician — a damaged mains conductor is not a thing to nurse along.

  • 02 The car that will not start

    One question splits the world in two before any of the four: does it crank? If the starter will not even turn the engine, no amount of fuel will help you — begin at the battery and its terminals, then look past them, because a strong battery still has to get its current through the heavy cable and the ground strap, and past a starter, a switch that must be satisfied, and an engine that can itself be seized. If it cranks and will not fire, an engine is asking four things: air, fuel, spark, and compression — the squeeze that makes the other three matter. Work them one at a time and each answer closes a door. (This branch assumes a keyed, petrol, audible-pump car. Diesel has no spark; push-button start, a silent pump or an immobiliser will defeat the particular tests while leaving the four questions untouched.)

  • 03 The faucet that will not stop

    Same three parts. Supply is the shutoff beneath the sink, control is the handle, and a steady drip from the spout with the handle closed points almost always to the seal inside — a washer or a cartridge, or the seat it closes against. It is the mechanism, not the pipes, which is why the whole job is usually a part that costs less than the drive to fetch it. Turn the supply off before anything is opened, and if the drip is coming from a joint rather than the spout, the stage has changed and so has the trade.

the two myths

Where it goes wrong.

The first is the parts cannon: replacing pieces one by one, on guesses, until the problem leaves. It is the most expensive tool in any house and it is not for sale. Every unneeded part is a wager, and every swap is a fresh chance to break something that was working — which is how a repair becomes two repairs. Mechanics watch it every week: hundreds of dollars of ignition parts fired at a car that needed a battery cable. The parts cannon does not aim. Questions aim.

The second is the sentence that ends all diagnosis: it just broke. Said aloud, it closes the case before anyone has opened it, and the fault settles in to stay. There is always a cause — and wear is one of them, which is the honest version of the objection. A bearing that ran for nine good years did not break for no reason; it broke for the most ordinary reason there is, and ordinary reasons are findable.

the limits

What four questions cannot do.

They do not promise you the fault. They narrow it, which is a different and more honest claim. Two faults at once will defeat a clean bisection, because clearing one stage no longer clears the symptom; a fault that only shows under a meter will not show under your hand at all; and a machine that has already been worked on by somebody with a parts cannon is a machine whose history is now lying to you.

The three-part reading is a decomposition, not a law. It fits a lamp, a pump, a furnace and a car, and it does not fit a hammer, a rope or a beam — and where it does not fit, forcing it is its own kind of guessing. Take what it gives on the machines it suits and put it down on the ones it does not.

And the method has no opinion about what is yours to touch. Four good answers will still hand you a job that belongs to a licensed trade, and the correct response to that is satisfaction rather than disappointment: you now know what to tell them, which is most of what you were paying them to find out. Damaged cords and plugs, anything on the wall side of a socket, gas appliances and fuel systems are in that category before the first question is asked.

the rep

Do this once, on something small.

Choose one broken thing in your keeping this week — the small annoying one, not the expensive frightening one — and before you touch a tool, write the four answers down. What changed. Which stage. How you would summon it. What the fix is worth. Four lines on the back of an envelope, and then, and only then, reach for the steel. The second time you do it you will stop needing the envelope.

Guessing with parts is gambling. Questions aim.

drink deep

The rest of the bucket.

Nine domains, and a draw from each when it meets the standard and not before — leave an address and you will hear when the next one does.

No spam. One note when it’s live, and nothing you didn’t ask for.