# The Golden Thread

Every product decision should be traceable: from the customers who asked, through the work and the code that shipped it, to whether it worked. ProductOS is built on that idea. This page describes the method; you can run it with or without the product.

Updated 2026-09-29.

## What is the Golden Thread?

Six kinds of record, linked in both directions: feedback, themes, Work, initiatives, releases and outcomes. Walk it forward from a customer's words to see what they changed. Walk it back from a result to see who asked for it and why.

| Layer | The question it answers | Who reads it | How often |
| --- | --- | --- | --- |
| Feedback | What is one customer saying? | Product, support, sales | Continuously |
| Themes | What are many customers saying together? | Product, leadership | Weekly |
| Work | What are we going to build about it? | Product, engineering, design | Every sprint |
| Initiatives | What are we betting on, and why? | Leadership, product | Quarterly |
| Releases | What ships together, and when? | Engineering, marketing, sales | Each release |
| Outcomes | Did it work? Did the number move? | Everyone | After release |

## Why keep every layer?

Collapse two layers and you lose information. Skip themes and you build for whoever shouted last. Skip initiatives and nobody can say what the work was meant to change. Skip outcomes and you stop learning, because nothing tells you which bets paid off.

## What are the principles?

1. **Start from evidence.** Every piece of Work links the customers who asked for it, and the accounts and revenue behind them.
2. **Group before you build.** Themes turn many voices into one problem worth solving, so you build once for the pattern.
3. **State the bet.** An initiative says what you expect to change and which objective it serves, before the work starts.
4. **Keep delivery attached.** The pull requests, the coding-agent sessions and the release live on the Work item, not in a separate tool nobody links to.
5. **Close the loop.** Every initiative ends in an outcome with a baseline, a target and what actually happened, reviewed on a date.
6. **Let everyone read it.** Sales, marketing and leadership work from the same thread. Reading it should never need a paid seat.
7. **Let agents work in it.** Coding and research agents read the evidence and write back, with the same role, checks and audit trail as a person.

## How do strategy and go-to-market attach?

- Vision and OKRs sit above initiatives, so every bet names the objective it serves.
- Competitors, market trends and signals feed themes, so outside evidence counts next to customer feedback.
- Deal blockers and win/loss records count as evidence, so a lost deal can move the roadmap.
- Launch plans, messaging, channels and battle cards hang off releases, so go-to-market starts from what shipped.

## What should the thread let you answer?

- Why did we build this, and who asked for it?
- How much revenue is asking for this, and from which accounts?
- Which objective does this work move?
- What shipped this month, and for whom?
- Did it work?

## Run it in ProductOS

ProductOS keeps the thread for you: the links stay correct as work moves, AI groups feedback and drafts specs, and agents read and write it over MCP.

- [Get started free](https://productos.ca/signup)
- [The Golden Thread in the docs](https://productos.ca/documentation/guides/concepts/golden-thread)
- [Agents and MCP](https://productos.ca/agents)

## For agents

- MCP server (streamable HTTP, OAuth): https://productos.ca/api/mcp/mcp
- Clients: Claude, Claude Code, Codex, Cursor, Any MCP client
- Index of the site: https://productos.ca/llms.txt
- Documentation: https://productos.ca/documentation