Improve reliability with Heplon

Heplon scores each of your platform goals by the extent to which they are related to reliability issues. For example, if there are many incidents that happen when changes are rolled out, Heplon will flag this for the 'Roll out a change' goal. This guide uses those scores to find weak spots and fix them.

Before you start

Find the weak goals

Open Goals & paths from the sidebar. The goals table shows each goal with a reliability level alongside velocity and standardization.

Look for goals marked red or yellow in the Reliability column. These are the goals where services have related reliability issues.

Open a goal and read its evidence

Click a red or yellow goal to open its detail page. Two parts tell you what is wrong:

  • Paths. The Paths section shows how services meet this goal: the documented paths, the adoption of each, and a bucket for services that follow no documented path. You can view the reliability scores for each path.
  • Incidents. The Scores section reports incidents linked to this goal. These are the failures behind the low score.

Together, these show which way of meeting the goal is failing and which services are affected.

Decide how to fix it

Use what you found to choose one of two fixes.

  • Migrate to a path you already have. If a documented path for this goal is more reliable, move the affected services onto it. Click Migrate services, select the services, and choose the target path. Heplon will open PRs for you.
  • Document a new path. If no existing path is reliable enough, define a better one. Click Add documented path, then migrate services to it once it exists.

Confirm the improvement

After services are migrated, open Key metrics from the Platform group in the sidebar. The Reliability section ranks services by MTTR (mean time to recovery) and CFR (change failure rate).

Find the services you migrated and check their reliability metrics over the following weeks. Heplon indicates the time at which a service was moved to a different path in the plot.