A practical startup strategy system built around Amazon's Working Backwards method.
This repo is not just a copy of Amazon's PR/FAQ process. It separates:
- the core Working Backwards artifacts I think are foundational
- the strategic extensions I added because a startup eventually needs more than a customer-facing narrative
The goal is not to produce a stack of pretty documents. The goal is to pressure-test startup ideas before they absorb too much time, money, and attention.
These are the documents closest to the original spirit of Working Backwards:
core/pr-faq.mdcore/mrd.mdcore/prd.md
They answer three different questions:
- PR/FAQ: What is the customer promise?
- MRD: What does the market need?
- PRD: What are we actually building?
These are not canonical Amazon artifacts. They are extensions I use when the idea survives the first stage and needs deeper pressure-testing:
extensions/market-analysis.mdextensions/competitive-analysis.mdextensions/investor-landscape.mdextensions/ma-analysis.md
They answer questions like:
- Is the market large and attractive enough?
- Where do we actually win?
- Who would fund this?
- Who would buy this and why?
Start with:
core/pr-faq.md
Do not skip this step.
If the PR/FAQ is weak, vague, incremental, or internally framed, stop there. Most ideas do not deserve the full stack.
If the PR/FAQ survives:
extensions/market-analysis.mdextensions/competitive-analysis.mdcore/mrd.mdcore/prd.md
This is where customer promise, market reality, positioning, and product requirements start checking each other.
Only after the idea still looks coherent:
extensions/investor-landscape.mdextensions/ma-analysis.md
These are optional early and much more useful once the product and market story are sharper.
PR/FAQ alone is excellent for customer clarity, but it is not enough for full company strategy.
A startup idea can survive the PR/FAQ and still fail because:
- the market is too small
- the buyer is wrong
- the positioning is weak
- the product requirements do not support the promise
- the investor case is unconvincing
- the likely exit logic is poor
That is why this repo is structured as a system instead of a single template.
These templates are designed to work well with LLMs, but the LLM is not the strategist.
Use an LLM for:
- first drafts
- critique
- alternative framings
- consistency checks across documents
- fast iteration after new evidence
Do not use an LLM to:
- invent customer truth
- replace user research
- skip judgment
- fill weak sections with generic confidence
Documents like these go stale quickly unless they stay coupled to reality.
At a minimum, each working doc should have:
- an owner
- a status
- a last updated date
- a reason it changed
Good update triggers include:
- new customer interview patterns
- a major competitor launch
- a product-direction shift
- pricing changes
- new evidence that invalidates a prior assumption
- fundraising timing changes
The point is not to update constantly. The point is to keep the documents alive enough that dead assumptions do not masquerade as current strategy.
Start here:
core/pr-faq.mdcore/mrd.mdcore/prd.mddocs/template-index.md
Then use the extension docs as needed.
The extension documents are my adaptation, not a claim about what Amazon formally does.
That distinction matters.
The repo is best understood as:
- Working Backwards as the spine
- additional strategy documents as the surrounding system