Frequently asked questions

Vindex catches the five things that make requirements fail in sprint: vague scope, missing acceptance criteria, stories too big to estimate, hidden dependencies, and untestable outcomes. Scoring is based on INVEST, the industry-standard checklist for user story quality, but you don't need to know the framework to get value from the results.

No. Vindex evaluates stories in the tools you already use (Jira, Azure DevOps, Asana) and surfaces story health scores back in those tools. It's not a replacement for refinement either; it's the signal that tells you which stories need refining.

No. Vindex gives you structured, plain-English recommendations for how a story could be improved. A human still writes the story. That's deliberate; refinement is a conversation, and Vindex is there to make the conversation sharper.

The unique user story. Every subscription plan includes a monthly quota of new stories, and once a story has been counted, re-scoring it is free forever, in any later month. Score, fix, re-score, confirm: the improvement loop never costs extra. Backlog grooming of previously scored stories never touches your quota.

Identity, not text. Each integration tracks the platform's own permanent record identifier (the Jira issue id, the Asana task gid, the Azure DevOps work item id, the ServiceNow sys_id), so editing, renaming, or moving a story never makes it a new one. Deleting a story and recreating it does create a new story, as does splitting one story into several.

Send a stable external_id with each story and re-scores of that id are free forever. If you don't send one, Vindex falls back to a fingerprint of the story's content: unchanged content stays free, edited content bills as a new story. The response tells you which rule applied (identity_resolution), so there are never surprises on a bill.

New stories pause until the quota resets on the first of the month (00:00 UTC) or you upgrade. Everything already scored keeps working: re-scoring counted stories stays free and available even at the limit. Unused quota doesn't roll over.

1 credit = 1 unique story per project, with every re-scoring round included. Score a client's backlog, they fix their stories, score it again: round two of the same stories consumes 0 credits. The estimate always shows the split before you confirm: how many stories were previously scored in the project (0 credits) and how many are new. If your upload has no ID column, stories can't be matched to earlier rounds and the full selection bills as new. The estimate says so up front.

A percentage that summarises how ready a story is for sprint planning. It's the arithmetic mean of the six INVEST dimension scores, stored on a 0–10 scale and always displayed as a whole percentage (for example 64%). The default quality gate is 70%; below that, a story shouldn't enter a sprint without refinement.

70% and above is healthy. 50% to 69% needs review - surface it in refinement. Below 50% is at risk and should not go into a sprint. See Interpreting scores for the full breakdown.

Dimension scores are stored as integers from 1 to 10 and displayed as percentages. Below 50% is a clear signal the story doesn't satisfy that dimension well enough to move forward. On the results page, and in the API's dimension_fail_counts, that threshold is what powers headlines like “31 of your 48 stories are missing testable acceptance criteria.”

The default gate is 70%, but company admins can tune it to match their own definition of ready. If you need per-project gates or team-specific thresholds, contact us; that's on the roadmap.

Jira Cloud, Azure DevOps, and Asana are supported today, along with CSV upload, the direct API, and AI assistants over MCP (Claude, ChatGPT, Codex, Cursor, GitHub Copilot). Trello and Monday are on the roadmap.

Yes, for the initial install: Jira Forge apps require a site admin, Azure DevOps extensions an organization admin, and Asana apps a workspace admin. Once installed, scoring works for every user without further admin involvement.

The public /evaluate flow scores the first 5 stories in your CSV so you can see what Vindex finds before signing up. Full-backlog scoring is available once you connect an integration or register an account.

Scores are stored in a Vindex Data Platform. CSV uploads from the public flow are retained for 24 hours and then deleted automatically by a TTL index; no CSV contents are written to the filesystem.

No. Scores are scoped to your customer_id, which is resolved to your company at query time. Portal endpoints return 404 Not Found for any score that belongs to a different company; we never reveal records across company boundaries.

No. Vindex records the model and provider used for every scoring run, but model_metadata.temperature is always excluded from API responses and callback payloads.