Security, privacy, and troubleshooting

How Tailwin separates customer data

Most customer records carry an organization ID. Signed-in requests resolve an active organization, verify membership or authorized partner access, and scope reads and writes to that organization. Partner act-into-client actions and privileged impersonation are audited.

Public pages use bounded tokens or public IDs for a specific purpose, such as report sharing, proposal acceptance, content approval, a receipt, a job application, or a concierge widget. A public token is access: revoke it when exposed.

Secret handling

• Store platform credentials in the deployment secret manager.
• Store customer connector credentials through Tailwin’s encrypted connection or source fields.
• Never place real values in Markdown, screenshots, source code, or chat transcripts.
• MCP, webhook, widget, ingest, and share secrets should be scoped and revocable.
• After a visible credential is shared, rotate it even if it was shared with an authorized operator; screenshots and chat histories persist.

Privacy boundaries

Tailwin intentionally avoids visitor fingerprinting, session replay, form-field capture, and default person identification. Crawler tracking records bot request metadata, not bodies. Talent has retention and deletion behavior. Cross-organization benchmarks require opt-in and a minimum cohort.

The customer remains responsible for notices, consent, data-processing terms, content rights, employment review, and lawful third-party account access.

General troubleshooting sequence

1. Confirm the active organization, site, and date range.
2. Read the explicit status or reason; do not translate missing into zero.
3. Confirm the relevant source, credential, and account ID.
4. Confirm the web and worker versions match.
5. Check whether the action is synchronous or queued.
6. Review the source’s last sync/error or the relevant operational log.
7. Retry once only when the action is idempotent or free. Paid Press, SERP, and other vendor operations require their specific recovery procedure.
8. Capture the time, organization ID, site/source ID, route, and error without including secrets.

Common problems

Problem	Likely cause	Next step
UI works but scheduled work does not	Worker stopped, stale image, or Redis unavailable	Check worker health and ensure web and worker run the same release
Source remains empty	First sync did not finish, account lacks scope, or wrong org	Review source status and last error in the owning client org
AI scan is plausible but not real	Development mock mode enabled	Confirm runtime engine mode and health before using the result
DataForSEO research returns 503	Platform credentials absent	Configure and restart, or use labeled Search Console fallback
Rank check appears stuck	Standard Queue result still pending	Allow the bounded polling window; use the task ledger before retrying
Crawler hits absent	Client-side install or CDN bypass	Move detection to edge/origin and run live verification
Public share reveals too much	Dashboard/source/audience mistake	Revoke the token immediately, correct the dashboard, issue a new token
Checkout succeeds but plan does not change	Stripe webhook path, signing secret, or event setup is wrong	Inspect webhook delivery and entitlement processing

Report a problem

Provide:

• what you expected and what occurred;
• exact time and timezone;
• organization, site, source, or dashboard identifier;
• the page or operation;
• safe error text and a redacted screenshot;
• whether a retry occurred and whether the action could incur cost.

Do not provide passwords, API keys, authorization headers, raw cookies, webhook secrets, or unredacted personal data.