Skip to main content
A healthy Lightdash deployment is adopted by the people who need it, governed so access and content stay trustworthy, and efficient so warehouse cost tracks the value it returns. This guide sets out what good looks like across those three areas, why each one matters, and how often to review it.
You can request a Lightdash Instance Health review as part of the Enterprise plan. View pricing

Adoption

Are people signing up to Lightdash, and staying active once they’re there?

Access governance

Is access granted safely and kept current?

Content lifecycle

Is content trustworthy, maintained, and free of clutter?

Cost and performance

Is warehouse spend tracking the value returned?

​
Adoption

Adoption measures whether the people provisioned in Lightdash are actually using it, and whether anyone who was active has quietly dropped off. What good looks like:
  • Most provisioned users run at least one query every 90 days
  • Weekly-to-monthly stickiness sits above 40%
  • No cohort of users who were active and then went silent
Lapsed users are an early signal that something isn’t working for your instance. When someone stops querying, decisions in their team start getting made without the data. Tracking adoption by team (using groups) shows you where to focus enablement, rather than treating the instance as one undifferentiated number.
Review usage monthly. When a team’s activation drops, run a short enablement session focused on the spaces they already use, and appoint a champion to follow up. Usage analytics gives you the activation and engagement numbers to track.

​
Access governance

Governance keeps access and credentials safe as your team and content grow. What good looks like:
  • Access is granted through groups rather than per user
  • Admins are kept to a small active set
  • No personal access tokens are left without an expiry
  • Content created by people who have left is reviewed and archived, not left to drift
Over-broad access and tokens that never expire are a standing security risk, and content left behind when people leave erodes trust in what is current. Managing access through groups also makes onboarding and offboarding atomic: add or remove a person once, and their access follows.
Review access quarterly. Grant project and space access via groups and roles, and rotate stale tokens.
Setting a maximum token lifetime (PAT_MAX_EXPIRATION_TIME_IN_DAYS) blocks new tokens from being created without an expiry. On self-hosted instances this is an environment variable; on Cloud, request it from Lightdash support. It does not retroactively revoke existing no-expiry tokens, so any already created must be rotated or revoked manually.

​
Content lifecycle

Content lifecycle is about keeping your charts and dashboards current and easy to navigate, so people trust what they find. What good looks like:
  • Little or no content with zero views
  • Each active space is aligned to a group, so a team is clearly responsible for it
  • The canonical charts and dashboards are marked as verified so people know what to trust
Stale and duplicated content drives shadow reporting and undermines confidence in the numbers; dead content that is still on a scheduled delivery also wastes warehouse cost. A tidy, maintained, verified content set is what lets new users (and AI agents) start from the right place.
Review content quarterly. Archive unused dashboards and charts into a dedicated archive space, align each active space to the group responsible for it, and verify the content you want people to start from. When moving content between projects, use content promotion.

​
Cost and performance

This area keeps query performance healthy and warehouse spend proportional to the value Lightdash returns. What good looks like:
  • p95 query latency is within your target
  • The error rate is low
  • No single explore dominates warehouse time
A handful of heavy explores or failing scheduled deliveries can account for a disproportionate share of spend and slow the experience for everyone. Watching cost by explore, user, and context tells you where to optimise instead of guessing.
Review cost monthly. Investigate the heaviest explores with your data team for missing aggregations or overly broad filters, and prune or fix failing scheduled deliveries.
These practices map to the sections of a Lightdash instance health review. If you have run one, work the report’s action plan from top to bottom: it is ordered by priority and links each recommendation back to the relevant section above.