Setting up Brytebuild: what to do first

The order you load things in is not a matter of taste. Each stage below produces the data the next stage needs, and skipping ahead is the usual reason a rollout stalls halfway with everyone quietly back on the spreadsheet.

A realistic timeline for a small manufacturer is four to six weeks, running alongside normal work. Faster is possible; it usually means data that has to be redone.

Load what you buy, make and sell. No quantities, no recipes, no locations. Just the definitions.

Why first: everything else references an item. A BOM is a list of items, a stock figure is a quantity of an item, an order line points at an item. Load anything else first and you will be creating items on the fly to satisfy it, which is how you end up with three records for the same bracket.

The decision that matters here: your part-numbering scheme. It is genuinely hard to change later, because every other record and every label refers to it. Whatever you pick, make it something a person can read out over the phone.

Do not load your entire catalogue. Start with what moved in the last twelve months. Dead parts can wait, and loading them first is a way of spending three weeks with nothing to show.

Now say what goes into what. Work top-down from your best-selling assemblies rather than trying to be complete.

Why second: a BOM can only reference items that exist. It is also the stage that finds errors in stage 1 — you discover the same washer entered twice under different numbers precisely when you try to use it in a recipe.

Expect this stage to be the slow one. Most shops find their BOMs were never fully written down, and the real quantities differ from the drawing because the shop improved the design years ago. That reconciliation is valuable work, but it is work, and it is worth telling whoever asked for the timeline.

Count what you have and enter it. Then start recording every movement in and out from that moment.

Why it must be a single day: an opening stock figure is only true at an instant. Load half your stock on Monday and half on Thursday, keep trading in between, and neither half is correct — you will spend longer reconciling than the count would have taken.

This is the point of no return, and the right kind of pressure. From here the system is the record. If material moves without being recorded, the numbers drift, and once people stop trusting them they stop using the system. Every failed implementation we have seen failed here, not at stage 1.

Pick a quiet day. The end of a week is better than the start of one.

Only now add the method: how each thing is built, which work centres it passes through, how long operations take, quality checks, integrations.

Why last: none of it is needed to run the material loop, and all of it is easier to write once you have a month of real jobs to look at. Written up front, from memory, work instructions tend to describe how things are supposed to happen rather than how they do.

Write these for your newest employee, not your most experienced one. If the instruction only makes sense to someone who already knows the job, it is a reminder rather than an instruction.

It is tempting to keep the spreadsheet going in parallel "just until we trust it". Keep it as a read-only reference by all means, but do not keep updating it.

Two live records means two sources of truth, twice the entry, and a guarantee that when they disagree nobody knows which is right. In practice the old system wins, because it is familiar — and the rollout quietly dies. Better to commit at stage 3 and fix problems in the new system than to hedge.

What's next