> ## Documentation Index
> Fetch the complete documentation index at: https://docs.startupsuperpowers.io/llms.txt
> Use this file to discover all available pages before exploring further.
# surveys
> Quantitative signal at scale — honest question sets, optionally deployed live as a Tally form
`/surveys` helps you test signals beyond your first handful of interviews: a short, bias-checked question set tied to specific hypotheses, a realistic distribution plan, and — if you want a live link — a deployed Tally form.
## When to reach for it
* You've done a few interviews and want to confirm the pattern at scale
* You need to prioritize among competing hypotheses
* You can reach a targeted audience you can't interview one-on-one
* "Is a survey even the right tool right now?"
## First, an honest applicability check
Before building anything, the skill gives you a frank 2–3 sentence read on whether now is the moment. Good signs: interviews already done, untested hypotheses, a concrete channel to reach respondents. Caution signs it will raise — but not block on: no interviews yet (surveys *confirm* patterns; they rarely discover them), no distribution channel (a great survey nobody sees is wasted effort), or an idea still in flux. You decide; it helps either way.
## Drafting the questions
One question at a time, the flow settles: which hypotheses to test (capped at 4–5, prioritizing problem and willingness-to-pay bets), where the survey will actually be posted — vague answers like "social media" get challenged until there's a concrete community — and what respondents get out of it. The question budget follows the audience: 3–7 questions for cold audiences, up to 10 for warm ones.
Every question passes a silent bias check before you ever see it: anything leading, compound, solution-hinting, or asking about future intentions instead of past behavior gets rewritten first. A final open-ended question has to be *earned* — the advisor presents the tradeoff instead of adding one by default.
You end with a question set you can paste into any tool — or deploy directly:
## Going live with Tally
Say "deploy this" and the skill turns your question set into a live [Tally](https://tally.so) form and hands back a shareable link, updating the survey file with the form ID and URL. Afterwards, "how's my survey going?" fetches the current submission count.
Deployment needs the Tally MCP connected once (an API key from Tally's settings plus one `claude mcp add` command — the skill gives you the exact steps and verifies before proceeding). Without it, questions-only mode is fully functional.
Results *analysis* — mapping responses back to hypotheses and proposing status changes — isn't built yet. Today the plugin reports submission counts; reading the results against your bets is a planned update.
## What it writes
| File | What's in it |
| --------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `startup/surveys/{slug}.md` | Purpose, target audience with a realistic response goal, distribution plan, and the numbered questions — each with its type and a `Tests: [[hypothesis]]` link. Tally deployments add the form ID and URL |
## Good to know
* Response goals are sized by channel reality: cold audiences convert around 5–10%, warm ones 20–30% — the goal in the file reflects that math, not wishful thinking.
* Yes/no questions are avoided for core hypothesis tests — they collapse the nuance you're trying to measure.
* Strongest signal comes from triangulating: surveys tell you *what*, interviews tell you *why*.
## Related
Do these first — surveys confirm what conversations discover.
How survey questions tie into the same evidence trail.