Many early computers were sold as self-assembly kits (you get a box of parts and an instruction leaflet and have to solder them together yourself). For example, the Altair was priced at $439 kit, $621 preassembled.
The obvious rationale for offering this option is that if you are willing to do the assembly work, this saves the vendor the cost of labor to do it, so it's natural to pass on the saving to the customer. But it's not clear to me that this stands up so well to closer scrutiny. Some percentage of the time, the assembled computer won't work. The vendor, having no way of knowing whose fault this is, has to take the machine back and try to debug it. This quickly adds up to much more expense than it would've been to just assemble and test the machine before sending it out, particularly since if every machine will be sold preassembled, the vendor can invest in things like wave soldering equipment that can do the job much more efficiently than a user doing it once on the kitchen table. Tandy, for example, considered selling the TRS-80 in kit form, but concluded it would be more trouble than it was worth. But if Tandy didn't think it was worth the trouble, why was it that other vendors did?
- Tradition. The personal computer industry developed from roots in the electronics hobby, where magazines such as Popular Electronics would publish projects, instructions on how to build something interesting; that is exactly how the Altair was first announced to the world. Originally, you would have to source the components yourself; a kit with all the components in a box, was a step up from that. It just takes time for tradition to change.
- Market segmentation. To maximize overall profit, you want to be able to charge a higher price to customers with more money, e.g. business versus hobbyists. Kit versus preassembled was a way to do that.
- Add an egg. There is a famous, possibly apocryphal but plausible story, about instant cake mix that sold better when it was made less comprehensive, requiring the customer to add an egg; this was interpreted as a finding that people like a product better when they have to put some work into it and get to feel some pride in creation. By this way of thinking, perhaps some hobbyists actively preferred soldiering their computer together themselves, and would be less interested in a preassembled machine.
- Speed. Yes, it's ultimately more efficient to do all the assembly in the factory. But hiring workers and setting up factories takes time. Just shoving the components in a box lets the business start ramping up quickly, and that was what mattered in the early days of the industry, when companies like MITS were flooded with far more orders for the Altair than they could fulfill.
- Something else I haven't thought of.
Is there any evidence – ideally something written down at the time, but I would also be interested in any relevant experiences people have, or any logic that I have missed – to indicate which factors were the decisive ones?