sales-skills/sales-kardsort

KardSort Platform Help

KardSort (kardsort.com) is a simple, low-cost, indie card-sorting + tree-testing tool for UX teams — the budget end of the information-architecture (IA) research cluster, built and maintained by a single developer. It runs card sorting (open/closed/hybrid, with rich-media cards — images/audio/video/YouTube — and participant-created cards, and letting a card go into more than one category) and task-based tree testing (import a sitemap or text tree, or build one from a card sort), with built-in analytics for both: a similarity matrix, hierarchical-clustering dendrogram, card ambiguity, participant agreement, and split-half reliability on the card-sort side; success/directness, first-click correctness, path analysis, node heatmaps, and Sankey navigation flows on the tree-test side. You can link a card sort to a tree test for correlation analytics and IA-validation scoring, add pre/post questionnaires with power filters (segment by answer), password-protect and schedule studies, and run in 7 languages.

Two things to say almost every time:

• A clean card sort or tree test proves the structure works, not that anyone will pay. Read the similarity matrix / dendrogram / findability score as evidence the IA is right — if the user is still deciding whether to build, keep the findings but take the go/no-go to a real behavior test via /sales-idea-validation.
• KardSort has NO public API, MCP server, webhooks, or Zapier. The only programmatic surface is export (CSV, Casolysis, SynCaps V3); a "pipe results into my CRM/warehouse" ask is a scripted export job, and the documented analysis path is the third-party cardsort Python package (github.com/katoss/cardsort) reading the Casolysis CSV. If a documented REST/webhook pipeline is required, route to /sales-userintuition.

Step 1 — Gather context

If references/learnings.md exists, read it first for accumulated platform knowledge.

Ask only what you can't infer:

1. What do you want from KardSort?
   • A) Pick/run a study — card sort (open/closed/hybrid) vs tree test, rich-media or participant-created cards
   • B) Read the analytics — similarity matrix, dendrogram, agreement; tree-test success/directness/path/Sankey
   • C) Link studies — connect a card sort to a tree test for correlation / IA-validation scoring
   • D) Get data out / automate — CSV / Casolysis / SynCaps V3 export and the cardsort Python package (no API)
   • E) Choose a plan — Free vs Starter vs Premium (all paid plans are one-time payments, not subscriptions)
2. Are you testing an IA you've already got, or still deciding whether to build? The second is an idea-validation question — flag it in Step 2.

Skip-ahead: if the prompt already names the study type or the question is specific, go to Step 3.

Step 2 — Route or answer directly

| If the user's question is about… | Route to | |---|---| | Comparing card-sort/IA tools across the market, or the tool-agnostic validate-before-building method | /sales-idea-validation {question} | | The MCP-native budget twin (same battery + an idea-validation engine + an MCP server) | /sales-validatethat {question} | | The deep-IA reference-standard peer (benchmark card-sort/tree-test analysis, own panel) | /sales-optimalworkshop {question} | | The broad self-serve all-in-one usability peer (own panel, session recording, prototype tests) | /sales-uxtweak {question} | | A documented API / webhook-native research pipeline (KardSort has none) | /sales-userintuition {question} | | Running a real behavior demand test (smoke-test page, waitlist, pre-sale) instead of an IA test | /sales-idea-validation or /sales-funnel {question} |

When routing, give the exact command: "This is a {domain} question — run: /sales-idea-validation {original question}"

Otherwise, answer KardSort-specific questions using Step 3.

Step 3 — KardSort platform reference

Read references/platform-guide.md for the full reference — the module/automation-surface table (what's export-accessible vs UI-only), best-effort pricing and the one-time-payment plan gates (study/participant/card caps, data-retention windows), the Study → Card → Category → Participant data model with JSON shapes, the export formats (CSV / Casolysis / SynCaps V3) and the cardsort Python package recipe, and the no-API data-out playbook. For the export column schema the cardsort package expects, read references/kardsort-api-reference.md — KardSort has no REST API, so that file documents the export/analysis surface instead.

Answer using only the relevant section — don't dump the full reference.

Step 4 — Actionable guidance

• In every response, say the caveat: an IA/usability result is not demand. State plainly — even when the user only asked how to fix navigation — that a card sort or tree test proves a structure works (category agreement, findability) but does not prove strangers will pay. Keep the findings; for any build-or-not question, take the go/no-go to a real behavior test (smoke test, pre-sale) via /sales-idea-validation.
• Flag it as export-only when any automation/API/pipeline comes up. KardSort has no public REST API, MCP server, webhooks, or Zapier (verify — it may change). The only data-out is export — CSV (cards & categories, with remarks), Casolysis (recommended), SynCaps V3 (.txt), questionnaire responses, participant info. The documented analysis path is the third-party cardsort Python package reading the Casolysis CSV — one row per participant × card placement, columns card_id, card_label, category_id, category_label, user_id (pip install cardsort → create_dendrogram / get_cluster_labels). "Sync results to HubSpot/Snowflake on a schedule" is a scripted export job, not REST — if a documented pipeline is required, route to /sales-userintuition.
• State that paid plans are one-time payments, not subscriptions. Starter and Premium are one-time purchases (Starter = one month of access, no auto-renewal; Premium = a year, no auto-renewal) — say this explicitly so the user doesn't expect a recurring plan, and present every price as best-effort pointing to kardsort.com/pricing.
• Match the tool to the question, and name the outputs to read. Recommend by job: card sort = how users group/label content (open = discover, closed = validate; hybrid = both) — whenever you recommend a card sort, tell the user to read its results via the similarity matrix and the dendrogram (plus participant agreement); tree test = can users find things in your nav (findability, no visuals → success/directness/ first-click/Sankey path). Design the tree with a card sort, then validate it with a tree test — and link the two for correlation / IA-validation scoring.
• Warn about the plan caps and data-retention window before they hit them. Free is 1 published study / 10 participants / 10 cards / 20 tree nodes / 7-day retention; Starter and Premium raise these (up to ~1000 participants / 1000 cards / 500 nodes) and extend retention to 30 / 90 days. Present all limits as best-effort and flag that results are purged after the retention window — export before it lapses.
• Set participant counts realistically. A stable similarity matrix / split-half-reliable card sort wants ~30+ participants; a findability tree test similar. Say it's a rule-of-thumb, not a guarantee, and note the per-plan participant cap may force a paid tier for a properly-powered study.

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) — pricing, the one-time-payment model, the per-plan caps, the data-retention windows, and the export-format list; verify at kardsort.com.

• IA/usability success ≠ demand. A clean card sort or tree test proves the structure works, not that anyone will pay. Route the build-or-not go/no-go to /sales-idea-validation.
• No REST API, no MCP, no webhooks, no Zapier. The only automation surface is export (CSV / Casolysis / SynCaps V3); the documented analysis path is the third-party cardsort Python package on the Casolysis CSV. Don't design a live REST/webhook pipeline around it — route pipeline needs to /sales-userintuition.
• Paid plans are one-time payments, not subscriptions. Starter (~one month of access) and Premium (~one year) are one-time purchases with no auto-renewal — set expectations accordingly; access lapses rather than renews.
• Free tier is a real evaluation cap, not just a trial. 1 published study, 10 participants, 10 cards, 20 tree nodes, 2 questions, 7-day retention — too small for a powered study. A 30-participant card sort needs a paid tier.
• Results are purged after the retention window (7 / 30 / 90 days by plan). Export (Casolysis/CSV/SynCaps) before the window lapses or the data is gone — there's no API to pull it back later.
• A stale listicle may say KardSort has "no similarity matrix/dendrogram." That's outdated — the current product ships 15+ card-sort KPIs and 20+ tree-test KPIs; trust kardsort.com over old comparison articles.
• Payment is Stripe/PayPal only. If those aren't supported in the user's country, the maker offers manual alternatives on request (contact via kardsort.com) — don't assume a blocked checkout means no access.

Related skills

• /sales-validatethat — The closest twin: the same budget card-sort/tree-test battery plus an Idea-Validation Engine, a Figma plugin, on-site embeds, and — unlike KardSort — an MCP server (run studies from Claude/Cursor). Route here when the user wants an automation surface or idea validation baked in; KardSort wins on one-time pricing and rich-media/participant-created cards. Install: npx skills add sales-skills/sales --skill sales-validatethat -a claude-code
• /sales-optimalworkshop — The IA "reference standard" peer: deep card sorting (OptimalSort) + tree testing (Treejack) + first-click (Chalkmark) with benchmark analysis and an own panel — route here when the job is pure IA rigor and price matters less; also UI-first (no API). Install: npx skills add sales-skills/sales --skill sales-optimalworkshop -a claude-code
• /sales-provenbyusers — Another budget twin: a cheaper full-battery suite (card sort/tree test + first-click/five-second/preference/surveys) with one-time pricing, but no rich-media/participant cards, no MCP, and no cardsort-package export path — just CSV. Route here when method breadth matters more than rich cards or a scripted export. Install: npx skills add sales-skills/sales --skill sales-provenbyusers -a claude-code
• /sales-uxtweak — The broad self-serve all-in-one usability peer (card sort + tree test + first-click / five-second / preference + prototype/website tests + session recording + surveys) with its own 130-country panel — route here for breadth and a built-in panel; also UI-only/no API. Install: npx skills add sales-skills/sales --skill sales-uxtweak -a claude-code
• /sales-lyssna — The ex-UsabilityHub usability + IA suite with a larger panel and a usable free plan — route here to compare, or when panel size / free-tier generosity matters. Install: npx skills add sales-skills/sales --skill sales-lyssna -a claude-code
• /sales-idea-validation — The tool-agnostic validate-before-building method + the full research/usability/validator landscape (use this to decide build-or-not; a passing IA test is not demand). Install: npx skills add sales-skills/sales --skill sales-idea-validation -a claude-code
• /sales-userintuition — The API/webhook/MCP-native real-human research peer — the pick when you need a documented programmatic research pipeline KardSort can't offer. Install: npx skills add sales-skills/sales --skill sales-userintuition -a claude-code
• /sales-funnel — Build the smoke-test / fake-door landing page that turns an IA hypothesis into a real demand test. Install: npx skills add sales-skills/sales --skill sales-funnel -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: "I need to fix my app's navigation but I'm on a tiny budget — what do I run?"

User says: "Users can't find settings in my app. I want to test the menu structure without paying Optimal Workshop's prices." Skill does: Recommends an open card sort first if the labels/categories are in doubt (read via the similarity matrix / dendrogram), then a tree test to validate findability — and to link the two for IA-validation scoring. Notes the Free tier's 1-study / 10-participant / 7-day-retention cap means a real 30-participant study needs Starter or Premium — and that those are one-time payments, not subscriptions. Adds that a clean tree test proves the nav works, not that the product sells. Result: The user runs a properly-powered, linked card-sort → tree-test study on a paid one-time tier for a fraction of the peer price.

Example 2: "How do I get KardSort results into my analysis pipeline / automate this?" (developer/automation)

User says: "Can I pull card-sort results out programmatically and cluster them myself instead of clicking around?" Skill does: States plainly that KardSort has no public REST API, MCP server, or webhooks — the only data-out is export (CSV, Casolysis [recommended], SynCaps V3). Points to the documented path: export the Casolysis CSV (columns card_id, card_label, category_id, category_label, user_id) and analyze it with the third-party cardsort Python package (pip install cardsort → analysis.create_dendrogram(df) / get_cluster_labels(df, cards)). For a scheduled warehouse sync, wrap that export+script in your own ETL; if a documented REST + webhooks pipeline is a hard requirement, routes to /sales-userintuition. Result: The user builds a scripted Casolysis-export → cardsort clustering job instead of hunting for an API that doesn't exist.

Example 3: "Is KardSort's free plan enough, and is it really a subscription?"

User says: "I want to try card sorting cheaply — what does the free plan cover and what am I paying monthly?" Skill does: Explains Free = 1 published study, 10 participants, 10 cards, 20 tree nodes, 7-day retention — fine for a trial, too small for a powered study. Clarifies that Starter and Premium are one-time payments, not subscriptions (Starter ≈ one month of access, Premium ≈ one year, no auto-renewal), so nothing recurs. Flags that results are purged after the retention window — export (Casolysis/CSV) before it lapses. Presents every figure as best-effort and points to kardsort.com/pricing. Result: The user picks the right one-time tier for a real study and exports results before retention lapses.

Troubleshooting

"Which do I run — a card sort or a tree test?"

Symptom: The user isn't sure whether to test categories/labels or navigation findability. Solution: Use a card sort when the question is how users group and name content (open = create/label to discover a structure; closed = sort into your categories to validate them; hybrid = both → read via the similarity matrix / dendrogram + participant agreement). Use a tree test when the structure exists and the question is whether users can find an item (success/directness/first-click, Sankey path, no visuals). Run the card sort to design the tree, then the tree test to validate it — and link them for correlation/IA-validation scoring. Remember: a passing test proves the structure works, not that anyone will pay.

"How do I get KardSort results out / automate exports?"

Symptom: The user wants results flowing into a CRM, warehouse, notebook, or Slack. Solution: KardSort has no public REST API, MCP server, or webhooks. Data-out is export — CSV (cards & categories, with remarks), Casolysis (recommended), SynCaps V3 (.txt), questionnaire responses, participant info. The documented analysis path is the third-party cardsort Python package on the Casolysis CSV (pip install cardsort). Glue a scheduled export into your own ETL, or drive the cardsort package from a script. If an automated documented pipeline is required, route to /sales-userintuition (REST API + HMAC webhooks + MCP).

"My results disappeared / the study says the free limit is reached."

Symptom: Data is gone, or the user can't publish/collect more. Solution: Two separate causes. (1) Data retention — results are purged after the plan window (Free 7 days, Starter 30, Premium 90); export (Casolysis/CSV/SynCaps) before it lapses, there's no API to recover it. (2) Plan caps — Free allows 1 published study / 10 participants / 10 cards / 20 nodes; a bigger or second study needs Starter or Premium (both one-time payments). Check which limit was hit before assuming a bug, and export existing data first.