Operations Blueprint: a plan and a starter repo for the small manufacturers software forgot

Most writing about operational software, the smart factory, AI in manufacturing, quietly assumes you are big. It assumes a budget for six-figure systems, an IT team to run them, and clean data already sitting in a warehouse. A small manufacturer has none of that. And small manufacturers are not a niche: about 99% of US manufacturers are small, most with fewer than twenty people.

I came to this from inside one, trying to build the kind of operation that is supposed to require a team and a budget I did not have. What came out of it is a method, and now an open tool.

What it does

Operations Blueprint is a small, deterministic engine (no LLM, no network) that takes ten answers about a shop and produces two things:

• A foundation-first plan: the relational data model, the process and SOPs, an ordered build list, and a self-hostable tooling list.
• A runnable starter repo: a tailored project you can hand to a coding agent, with a real Postgres schema, a FastAPI skeleton where the cost engine and quote endpoint are stubbed, AGENTS.md, TASKS.md, and a compose file.

Both are tailored to the answers. A configurable metal-fab shop gets a different schema and build order than a catalog sign shop. You can try it in the browser at damiankao.com/blueprint.

The idea

Once you decompose a shop's product into honest, relational data, down to every material, labor second, and overhead dollar, the things above it (pricing, quoting, scheduling, sales) become functions of that data instead of work you do by hand. Lean, Six Sigma, and continuous improvement made large-scale manufacturing reliable, but they rest on a measurement foundation small shops have never had. This is an attempt to supply that foundation, and to close the estimate-versus-actual loop automatically.

A worked example

The repo includes a full one: a fictional seven-person steel fabrication shop, run end to end from its (synthetic, deliberately messy) spreadsheets to a working slice. Loaded with its own data, it surfaced two problems that were invisible in the spreadsheets:

• A quote priced by hand that was quietly under the real cost.
• A job whose welding ran over, dropping the real margin well below the quote.

Neither needed new data; it was all already there, just not joinable. Full write-up: Running the Blueprint on a Shop That Doesn't Exist.

It is early

Scoped to custom and made-to-order manufacturing, MIT-licensed, and honestly still finding its edges. If you have run a small shop, I would like to hear where it is wrong.

• Repo: github.com/DamianFKao/operations-blueprint
• Try it: damiankao.com/blueprint