Assembly instructions are part of the design
Why a design that cannot be explained cannot be built twice.
Assembly instructions are not paperwork produced after a design is finished, they are one of the outputs of the design itself, because a designer who has not worked out and recorded a valid order of operations has not actually finished specifying how the thing gets built, only what it looks like once it is.
Why a design that lives only in one person's head cannot be repeated
Every assembly of more than a few parts has an order buried inside it, some steps that have to happen before others because a later part physically blocks access to an earlier one, a fastener that becomes impossible to reach once a panel goes on top of it, a wire that has to be threaded through a channel before the bracket on the far end of that channel is bolted down. A designer builds the first prototype with that order already worked out, usually without ever writing it down, because they discovered it by trial while putting the pieces together and simply remember which piece came next. That order is every bit as much a part of the design as any dimension on the drawing, since a set of parts that can only be assembled correctly in one sequence out of several possible ones is not really specified until that sequence itself is recorded somewhere a second builder can find it.
The tent-pole comparison
A four-pole camping tent looks, laid out on the grass, like it could go together in almost any order, since the poles are interchangeable lengths of the same material and none of them announces which sleeve it belongs in first. In practice only one or two of the several plausible sequences actually work without straining the fabric, because threading the last pole through its sleeve while an earlier pole is already flexed into its full arch means fighting the tent's own tension the entire way, sometimes hard enough to tear a seam or snap a section near a joint. Someone who has pitched that exact tent before knows the right order without thinking about it, the same way a designer knows the right order for their own prototype, and a first-time camper without a picture instruction sheet has to discover the same sequence by trial, at the cost of a strained seam or two along the way. The tent was never redesigned to fix this; the actual design already had one workable order hidden inside its geometry, and the instructions either capture that order or leave every new user to rediscover it themselves.
Why the order hides inside the geometry rather than announcing itself
The reason assembly order so often escapes the drawing entirely is that a drawing shows a finished, static arrangement of parts in their final positions, and nothing about a finished arrangement reveals which parts had to arrive first to make that arrangement reachable at all. Two completely different assembly sequences can produce the identical finished object, one of them straightforward and the other requiring a part to be flexed, angled, or partially disassembled again later to get everything into place, and a drawing alone cannot tell a builder which of those sequences they are looking at. Working out the valid sequence, or the several valid sequences where more than one genuinely exists, is a distinct piece of design work from working out the part geometry, one that has to be done deliberately rather than assumed to be obvious just because the finished result looks simple. It is also work that gets harder, not easier, as a design grows more constrained, since a machine designed tightly to save weight or space tends to leave fewer routes into its own interior once mostly assembled, and the very compactness a designer was aiming for can quietly narrow a dozen workable sequences down to exactly one, without anyone deciding that on purpose.
The one number worth remembering
A tent that goes up in under two minutes once the pole order is known can take a first-time camper the better part of twenty minutes of trial and error to get right without instructions, and at least one pole or panel typically has to be unclipped and redone somewhere in the middle, even though every single part being handled is completely identical to the one in the fast, correctly ordered build.
What this changes in practice
Treating assembly order as part of the design rather than as an afterthought changes when the instructions get written, moving the task from the end of a project, once a customer or a service manual finally demands them, to somewhere in the middle, while the original builder still remembers exactly where the trouble spots were. Written early enough, the instructions can also feed back into the design itself, since a sequence dependency that turns out to be awkward to describe is very often a sign that the underlying part arrangement could be improved rather than merely explained more carefully. A bracket that has to be threaded in at an odd angle before a neighbouring panel goes on is not just hard to write instructions for, it is also hard to build correctly under time pressure or poor lighting, and catching that difficulty while writing the instructions is considerably cheaper than catching it after fifty units have already gone together the awkward way.
Where this stops being true
Not every assembly actually has a hidden order worth documenting; plenty of designs genuinely can be put together correctly in more than one sequence, and writing an artificially rigid instruction sheet for one of those designs adds paperwork without adding any real protection against a mistaken build. The distinction worth making before writing instructions at all is whether the design has a genuine sequence dependency, something that physically fails or becomes unreachable if done out of order, or whether it merely has a sequence the original builder happened to use out of habit. Documenting the second kind as though it were the first only trains people to follow a recipe rather than to understand the actual constraint, and the first time a real variation is needed, a rigid habit dressed up as a rule tends to produce more confusion than a design that was honest about which order actually mattered and which did not.