The company

Then the rest of it, which took about as long as the desk did.

A brand and logo system, built around a mark that is the desk seen from above with two ports in it. That was deliberate. The identity had to be the interface, because the interface is the only part of this product that can never change.

A cost model from the bottom up: per desk, per order, and fixed startup. An MDF core with a laminate or veneer surface and edge banding, which is what the machines I was quoting could realistically produce, costed down to glue and electricity. The figures in it are draft assumptions rather than quotes I could hold anyone to, so I am not putting them on this page.

And a storefront. A custom Shopify theme carrying a desk configurator I wrote by hand in plain JavaScript: a multi-step wizard holding size, finish and leg state, compositing a layered preview image as you go, mapping the combination to a real purchasable variant and totalling the price with add-ons live. A landing page under it, and preorder capture rather than checkout, because there was nothing to ship.

This half is the reason the whole project is filed under what I got wrong.

None of the commercial work was wasted exactly. The mark is good, the cost model taught me what the surface finish actually costs, and writing the configurator forced the option matrix to be finite, which a CAD model never does. A configurator will not let you offer a leg colour you have not decided how to price.

But it was all built on top of a promise no one had tested. The store could take a preorder for a desk with three ports in it before anybody had put three ports in anything. When a top finally was cut, on totes in a garage with no shop, the tolerances held. That was luck as much as drawing, and I found out afterwards rather than before.

The order I should have used is the one I would use now: cut one port, drop one module into it, then decide whether there is a business.