How it works
What happens between describing your idea and reading a report you can work in.
Last updated 24 October 2026 · Markdown version
1. You tell us about the idea, and about you
You describe the idea in your own words, then answer a few short questions: where you are with it, what you most need to know, who has the problem, what you already have, your budget and time, where you are launching and, if you like, the rivals you already know, each of which the research looks up. Nothing about you is assumed beyond what you enter. Before anything runs, you see what each part of the research will look at, and can change any answer. Your next report starts with your answers about yourself filled in, to keep or change.
2. Research runs in stages
- Planning. Your answers become a research brief, and every area gets a quick first look to decide how deep to go.
- Research. Problem, market, competition, feasibility, timing, prior attempts, marketing and legal areas are researched side by side.
- Building on the research. Positioning, technology choices, what to buy or build, and effort use what the first stage found.
- Money. Unit economics from real prices and costs found in the research.
- Your plan. Customer discovery, sequence, your first 100 users, and whether to raise or bootstrap, written last so every step can cite a finding.
- Writing and checking. Each group gets the one factor it turns on and its key charts, and contradictions between groups are flagged.
Search terms drawn from your idea are sent to public data sources such as code and package registries, research databases, and trademark, court and regulatory records. Your idea itself is never published.
3. Every claim is checked before you see it
- A quote must appear word for word in the source it cites, or the claim is dropped.
- A number stated as a fact must appear in the quotes behind it, or the claim is dropped.
- Confidence is capped by the quality of the weakest source behind it.
- Numbers in charts come from findings, never typed in by a model.
- Plan steps that could apply to any idea are removed.
- Claims about you must rest on an answer you gave.
The checks catch many mistakes, not all of them, so treat a report as a starting point for your own research. The full rules are in our method.
4. Six groups, every time
| Group | What it answers | Areas |
|---|---|---|
| Validation | Is the problem real, is there a market, who is already there, can you build it, and why now. | Problem, Market, Competition, Feasibility, Timing, Prior attempts |
| Marketing | How to stand out against each competitor, where your audience is, and what to say to them. | Competitor teardown, Channels, Messaging, Positioning |
| Build | What to build it with, what to buy instead, and how long it really takes. | Stack, Buy vs build, Effort |
| Legal risks | The legal issues your idea may face where you launch: licences, IP, data rules and lawsuits. | Licensing, IP, Data protection, Litigation |
| Capital | What a customer is worth, and whether to raise or bootstrap. | Unit economics, Raising, Bootstrapping |
| Roadmap | The path to your first 100 users, with every step tied to a finding. | Customer discovery, Sequence, First 100 users |
Every area has its own page. When there is little to find, it says so and shows what was checked, instead of leaving a gap.
5. A workspace, not a PDF
- Type into the report like a document: rewrite anything, highlight it, note it, add pages of your own. Your edits are marked as yours, and the original research stays in Archives.
- Add notes anywhere in the report.
- Ask SeedBasis Agent about anything in the report. It tells you when an answer rests on your own edit.
- Connect your own AI assistant, such as Claude, so it can read your report where you already work.
Reports take a while to research. While yours runs, you see the questions it will answer and a few things you could have ready. It opens whole once every check has run, and we email you when it is ready.