PickFu (pickfu.com) platform help — instant consumer split-testing and market research: run a poll (head-to-head, ranked, star/emoji rating, open-ended, click or five-second test) past an on-demand panel of 15M+ verified respondents targeted by 90+ demographic traits, and get quantified preference plus written rationale in minutes. Unlike UI-only message-testing peers it exposes a real REST API, an MCP server, and a CLI. Use when setting up a PickFu poll or writing unbiased non-leading questions, split-testing a logo/ad/Amazon listing/book cover/app icon/business name/copy, targeting a consumer audience by demographics, reading vote percentages plus verbatim comments, budgeting credits or a poll cost, wiring the PickFu API, MCP, or CLI into an automation, or fixing shallow or off-target respondent answers. Do NOT use for the validate-before-building method or comparing research tools (use /sales-idea-validation), or B2B-panel message/positioning testing (use /sales-wynter).
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
PickFu (pickfu.com) platform help — instant consumer split-testing and market research: run a poll (head-to-head, ranked, star/emoji rating, open-ended, click or five-second test) past an on-demand panel of 15M+ verified respondents targeted by 90+ demographic traits, and get quantified preference plus written rationale in minutes. Unlike UI-only message-testing peers it exposes a real REST API, an MCP server, and a CLI. Use when setting up a PickFu poll or writing unbiased non-leading questions, split-testing a logo/ad/Amazon listing/book cover/app icon/business name/copy, targeting a consumer audience by demographics, reading vote percentages plus verbatim comments, budgeting credits or a poll cost, wiring the PickFu API, MCP, or CLI into an automation, or fixing shallow or off-target respondent answers. Do NOT use for the validate-before-building method or comparing research tools (use /sales-idea-validation), or B2B-panel message/positioning testing (use /sales-wynter).
argument-hint
[describe what you need help with in PickFu]
license
MIT
version
1.0.1
tags
["sales","market-research","platform"]
github
https://github.com/pickfu
PickFu Platform Help
PickFu (pickfu.com) is an instant consumer split-testing and market-research platform. You
create a poll — head-to-head A/B/C, ranked, star/emoji rating, open-ended, click test,
five-second test, single/multi-select, or Likert — pick a consumer audience from a panel of
15M+ verified respondents (90+ demographic targeting traits, 15+ countries), and get back
quantified preference (vote %) plus a written explanation from every respondent — usually
in minutes. Common jobs: testing logos, ads, Amazon/app-store listings, book covers, app
icons, business/domain names, packaging, pricing, and copy.
Its niche vs the rest of the research/validation cluster:
Consumer panel, fast + self-serve. The mirror image of Wynter (/sales-wynter): PickFu's
panel is broad consumer, Wynter's is B2B professionals. For hard-to-reach B2B targeting,
Wynter is the better instrument; for consumer creative/message testing at speed, PickFu.
Creative/message split-testing, not usability or idea validation. PickFu tells you which
option wins and why with real people — not whether a built product is usable
(/sales-lyssna, /sales-uxtweak) and not build-or-not (/sales-idea-validation).
A real developer surface. Unlike UI-only peers (Wynter, Articos), PickFu ships a REST
API, an MCP server, and a CLI (@pickfu/cli) — so polls can be created, watched, and
read programmatically.
Two things to say almost every time:
A poll measures preference (stated), not demand. "Option B won / people said they'd buy" is
what a reader prefers when asked, not a purchase. Keep the winning option + verbatim comments;
take the willingness-to-pay / conversion go/no-go to a real behavior test via /sales-idea-validation.
Results live or die by question wording. A leading or vague question ("Isn't this logo
great?") produces misleading results. Write neutral, goal-based questions — PickFu collects
the "why" automatically, so ask the real question, not "do you like this?".
Step 1 — Gather context
If references/learnings.md exists, read it first for accumulated platform knowledge.
Ask only what you can't infer:
What do you want from PickFu?
A) Set up a poll — pick a question type (head-to-head / ranked / rating / open-ended /
click / five-second) and write neutral questions
B) Target an audience — build a consumer panel by demographics/traits (or general
population)
C) Read results — vote %, average scores, verbatim comments, AI Insights/Highlights
D) Budget — credits, PAYG vs PickFu+, what a poll costs
E) Automate — REST API / MCP server / CLI / Zapier
Consumer audience, or B2B/professional? If B2B, PickFu's consumer panel may be thin —
consider /sales-wynter.
Skip-ahead: if the user wants the validate-before-building method or a cross-tool
comparison, that's a /sales-idea-validation question — route in Step 2.
Step 2 — Route or answer directly
If the user's question is about…
Route to
The validate-before-building method, or comparing research/validation tools across the market
/sales-idea-validation {question}
B2B message/positioning testing on a professional panel (seniority/industry/company size)
/sales-wynter {question}
Running a real behavior demand test (smoke-test page, waitlist, pre-sale) after a poll
/sales-idea-validation or /sales-funnel{question}
Usability / prototype / IA testing on a built product or design (not preference testing)
/sales-lyssna or /sales-uxtweak{question}
Synthetic-audience message simulation (AI personas instead of a real panel)
/sales-societies{question}
Analyzing existing survey/VoC data into themes (post-collection)
/sales-customer-feedback or /sales-trill{question}
When routing, give the exact command: "This is a {domain} question — run: /sales-idea-validation {original question}"
Otherwise, answer PickFu-specific questions using Step 3.
Step 3 — PickFu reference
Read references/platform-guide.md for the full reference — the question-type / automation
table, best-effort credit pricing and plan gates, the Survey → Question → Response data model,
audience targeting, and how to read vote % vs comments. Read
references/pickfu-api-reference.md for the developer surface — REST endpoints (base
https://api.pickfu.com/v1), the MCP server (https://mcp.pickfu.com/mcp), and the CLI.
Answer using only the relevant section — don't dump the full reference.
Step 4 — Actionable guidance
Say the caveat: a poll is preference, not demand. Whatever the user asks, make explicit that
a poll winner (or people saying "I'd buy this") means the option wins with a real reader when
asked — a genuine, useful signal — but it is stated preference, not observed demand or a
purchase. Point the user to the verbatim comments (the "why", objections, confusion) as the
real value rather than trusting the top-line %, then take the willingness-to-pay / conversion
go/no-go to a real behavior test (smoke test, pre-sale) via /sales-idea-validation.
Insist on neutral, goal-based questions. Tell the user to write non-leading questions and
let PickFu collect the "why" automatically — never "do you like this?" or "isn't X better?".
Framing bias is the #1 cause of misleading PickFu results. Use targeting traits to match the
real buyer, and if answers are shallow/off-target, that's a question/audience-design fix —
rate responses (thumbs up/down) so the panel improves and low-quality respondents are flagged.
Size a poll in credits and call pricing best-effort. Billing is pay-as-you-go from ~$1 per
response with a minimum poll around $15; cost scales with sample size (15–500) ×
targeting specificity × question count. In API/credit terms note 1 credit = $2 USD (differs
from Wynter's $1). PickFu+ (subscription) and prepaid credit bundles add discounts. Give a
directional figure but present every number as best-effort and point to pickfu.com/pricing
and the in-app cost calculator to confirm.
For automation, use the API/MCP/CLI — and poll for completion (there is no webhook). PickFu
has a real REST API (base https://api.pickfu.com/v1, Authorization: Bearer sk_..., 100
req/min), an MCP server (https://mcp.pickfu.com/mcp, OAuth or Bearer), and a CLI
(@pickfu/cli). Create a poll with POST /surveys (then POST /surveys/{id}/publish) and
read results from GET /surveys/{id}/responses. There is no completion webhook — detect a
finished poll by polling GET /surveys/{id} (watch numResponses/status) or pickfu survey watch <id>. Never auto-retry POST (it can duplicate a poll). API/MCP/CLI access is plan-gated (PickFu+/Team/Enterprise) —
confirm on the account's Settings > API Keys page; don't assume it's on the free tier.
Match consumer vs B2B to the panel. PickFu's panel is broad consumer — for B2B /
professional targeting (seniority, industry, company size) route to /sales-wynter instead;
verify the target audience builds out to your sample size before spending credits.
If you discover a gotcha or tip not in references/learnings.md, append it there with today's date.
Gotchas
Best-effort from research (2026-07) — PickFu's pricing (per-response cost, min poll, PickFu+ tiers,
credit = $2), panel size/composition, question-type lineup, plan gates, and API surface change;
verify at pickfu.com and pickfu.com/pricing.
Preference ≠ demand. A poll winner or "I'd buy this" is what a reader prefers when asked,
not a purchase. The verbatim comments (the "why") are the value — take the go/no-go to a real
behavior test.
Leading questions poison results. Vague or leading wording ("do you like this?", "isn't X
better?") is the top cause of misleading PickFu data. Write neutral, goal-based questions; PickFu
collects the "why" automatically.
Priced per response, not per seat; cost escalates with targeting. PAYG from ~$1/response, min
poll ~$15; cost scales with sample size × targeting × question count. 1 credit = $2 USD (not
$1 like Wynter). Size it in the in-app calculator; treat all figures as best-effort.
Consumer-skewed panel. For hard-to-reach B2B/professional audiences the panel may be thin
— use /sales-wynter. Verify the audience builds out to your sample size before committing.
No completion webhook — poll for it. The API has no webhook; detect a finished poll via GET /surveys/{id} or pickfu survey watch. Never auto-retry POST /surveys (duplicates the poll).
API/MCP/CLI is plan-gated. Programmatic access requires a paid tier (PickFu+/Team/Enterprise)
— confirm on Settings > API Keys; the free/PAYG UI account may not have it.
Respondent quality varies — rate to improve it. Some panelists give shallow answers; rate
responses (helpful / not helpful) so poor responders are flagged and future polls improve. This
is a question/audience-design lever, not a dead panel.
Related skills
/sales-wynter — The B2B-panel message-testing counterpart (LinkedIn-verified professionals, scores + quotes, UI-only/no API). Pick PickFu for consumer creative/message testing with an API; pick Wynter for hard-to-reach B2B positioning. Install: npx skills add sales-skills/sales --skill sales-wynter -a claude-code
/sales-pollthepeople — The direct real-panel consumer competitor (same job: vote % + written rationale in minutes) with ChatGPT-native analysis and a cheaper subscription entry (free Lite tier), but UI-only/no API and a smaller US-default 500k panel. Pick PickFu for automation (API/MCP/CLI) or a 15M+ international panel; Poll the People for a cheap US-consumer entry with built-in AI summaries. Install: npx skills add sales-skills/sales --skill sales-pollthepeople -a claude-code
/sales-idea-validation — The tool-agnostic validate-before-building method + the full research/validator tool landscape (use to decide build-or-not; a poll "win" is preference, not demand). Install: npx skills add sales-skills/sales --skill sales-idea-validation -a claude-code
/sales-societies — Artificial Societies, the synthetic-audience counterpart (simulate how a message lands with AI personas before spending on a real panel); its message/creative-tester cousin is /sales-evelance (AI personas score copy/creative/product on 12 psychology metrics) — both are fast, cheap synthetic pre-flights to triage options before deciding with PickFu's real consumer panel. Install: npx skills add sales-skills/sales --skill sales-societies -a claude-code
/sales-lyssna — Self-serve usability/IA testing on a built artifact (five-second/first-click/preference, card sort, tree test) — the right tool when the question is "is my product usable", not "which option wins". Install: npx skills add sales-skills/sales --skill sales-lyssna -a claude-code
/sales-customer-feedback — Real-customer NPS/CSAT/VoC strategy across tools (use to analyze existing feedback; PickFu proactively tests new creative). Install: npx skills add sales-skills/sales --skill sales-customer-feedback -a claude-code
/sales-do — Not sure which skill to use? The router matches any sales objective to the right skill. Install: npx skills add sales-skills/sales --skill sales-do -a claude-code
Examples
Example 1: "Which of my three Amazon main images should I run — this one scored best on PickFu, ship it?"
User says: "I ran a head-to-head of three product main images on PickFu and image B won 58% with great comments. Good enough to make it my live listing and expect more sales?"
Skill does: Credits the result as a real signal — actual consumers preferred B and told you
why, far better than guessing. Then draws the line: that's preference, not demand — a
preferred image is not a proven conversion lift. Directs the user to the verbatim comments (what
they noticed, what confused them, the words they used) as the real payload, suggests using those to
sharpen the image, and to confirm the lift with a real behavior test (A/B the live listing / a
smoke test) via /sales-idea-validation. Notes per-response pricing is best-effort.
Example 2: "How do I create a poll and pull the results into my own dashboard via the PickFu API?" (developer/automation)
User says: "I want to script PickFu — create a head-to-head poll, wait for it to finish, and POST the vote % and comments into our internal dashboard. What's the endpoint and how do I know when it's done?"
Skill does: Gives the real surface — POST https://api.pickfu.com/v1/surveys with
Authorization: Bearer sk_... to create (shows the questions array with type: head_to_head),
then explains there is no completion webhook, so pollGET /surveys/{id} (watch
numResponses/status) or use pickfu survey watch <id> / the MCP get_survey_responses tool,
and read results from GET /surveys/{id}/responses. Flags the 100 req/min limit, never
auto-retry POST, and that API access is plan-gated (confirm at Settings > API Keys). Points to
references/pickfu-api-reference.md for the full JSON shapes.
Example 3: "My PickFu results feel random / the comments are shallow — is the panel bad?"
User says: "Half my respondents just say 'looks good' and the votes swing every time I run it. Is PickFu's panel junk?"
Skill does: Explains it's usually a question/audience-design problem, not a dead panel:
vague or leading questions invite low-effort answers, and a too-broad audience adds noise.
Prescribes neutral, goal-based questions (let PickFu collect the "why"), tighter targeting
traits to match the real buyer, a larger sample size for stability, and rating responses
(thumbs up/down) so poor responders are flagged and the panel improves. Notes that for B2B targets
the consumer panel may be thin — consider /sales-wynter.
Troubleshooting
"Where's the PickFu webhook to fire when a poll finishes?"
Symptom: You're looking for a webhook to trigger a pipeline when a poll completes.
Cause: PickFu's REST API has no completion webhook — completion is detected by polling.
Solution: Poll GET /surveys/{id} and watch numResponses/status (or status: completed), or use pickfu survey watch <survey-id> in the CLI / the MCP get_survey_responses tool. Then read GET /surveys/{id}/responses. Respect the 100 req/min limit and back off on 429. Note API access is plan-gated — confirm at Settings > API Keys.
"My PickFu poll results swing or feel shallow — is the panel bad?"
Symptom: Votes change between runs, or respondents give one-line generic answers.
Cause: Usually a question/audience-design problem — vague/leading questions and too-broad targeting, not a dead panel.
Solution: Write neutral, goal-based questions (PickFu collects the "why" automatically), tighten targeting traits to the real buyer, increase sample size for stability, and rate responses so poor responders are flagged. For B2B targets, the consumer panel may be thin — consider /sales-wynter.
"Why is my PickFu poll so expensive / why did the cost jump?"
Symptom: The poll total is higher than expected, especially with tight targeting or many questions.
Cause: PickFu is pay-as-you-go from ~$1/response (min poll ~$15); cost scales with sample size × targeting specificity × question count, and in credit terms 1 credit = $2 USD.
Solution: Use the in-app cost calculator before launching, trim to the questions that matter, right-size the sample, and if you run many polls compare PAYG against PickFu+ / prepaid credit bundles. Treat all figures as best-effort — confirm at pickfu.com/pricing.
