Optimize Your Time to Green (TTG)

Time to Green (TTG) is the time from when a pull request (PR) opens and triggers CI to the moment all checks are green and the PR is review-ready.

TTG is a practical sub-metric of Time to Merge (TTM): by compressing TTG (lower is better), you remove the biggest day‑to‑day bottlenecks that developers feel, which in turn improves overall TTM.

Why is this important?

The biggest day‑to‑day waste in engineering teams: constant context switching and PR babysitting. The common loop is:

• 🧑‍💻 Write code
• 🧑‍💻 Push PR
• ⏳ CI runs
• ❌ CI fails 2 minutes later
• ⏳ Discover it much later
• ‍🧑‍💻 Switch context to debug and trigger CI run
• ⏳ Re-running CI
• ❌ Flaky test fails CI
• ‍🧑‍💻 Switch context to debug and trigger CI run
• ⏳ Re-running CI
• ✅ CI is finally green
• ‍🧑‍💻 Reach out to someone to review

This delay compounds across teams and drastically slows delivery. Nx Cloud fixes this.

How to improve TTG

High TTG usually comes from three sources: slow failure discovery, disruptive PR babysitting, and raw execution time. Tackle them in this order.

Prerequisite: Connect your workspace to Nx Cloud

If you haven't already, run the following command to connect your workspace to Nx Cloud:

1) Get failure feedback immediately (avoid late discovery)

When you don't notice CI failed, you lose time before you can act. Tighten the loop so failures surface where you’re working.

What to do: See failures immediately where you work by getting a notification in your editor: install Nx Console.

2) Eliminate PR babysitting (minimize context switching)

The expensive loop is switching branches to fix, re‑pushing, waiting, and repeating; especially with flakes.

What to do:

• Approve fixes instead of branch‑hopping: enable Self‑Healing CI to analyze failed tasks, propose and verify fixes, and commit to your PR after approval.
• Also enable flaky task detection and retries to automatically re-run flaky tasks in the background while you keep working undisturbed.

3) Shorten actual CI time (make the pipeline fast)

Once feedback and context switching are handled, compress the compute side.

What to do:

• Reuse work with Remote caching (Nx Replay).
• Run more in parallel with Distributed task execution (Nx Agents).
• Scale long suites with E2E test splitting so they finish quickly.

Measure TTG and diagnose bottlenecks

Doing too much work on PRs

• Use Nx Affected to only run what changed.

Cache hit rate is low

• Ensure tasks are cacheable and deterministic. Define outputs and configure inputs/namedInputs correctly. See: Configure Inputs, Configure Outputs, Inputs Reference
• Standardize Node/PNPM versions across dev and CI; avoid environment variables that unintentionally affect inputs.
• Use Remote caching (Nx Replay) to share results between CI and dev machines.

Agents are idle, or queue time is high

• Increase or right-size distributed capacity and parallelism with Nx Agents.
• Use Dynamic Agents to scale based on PR size.
• Remove unnecessary serialization (global locks, overly strict dependsOn).

E2E suites take too long

• Enable E2E test splitting so large suites run across agents.
• Ensure tests are shardable (no hidden global state, independent specs).

Flaky task rate is high

• Enable Flaky task detection and automatic retries.
• Isolate and quarantine persistently flaky suites to keep pipelines green.

Late failure discovery / PR babysitting

• Install Nx Console for instant failure and fix notifications in your editor.
• Enable Self‑Healing CI to propose and validate fixes automatically (ensure the npx nx-cloud fix-ci step runs with if: always()).

Talk to us

If you still need help feel free to reach out to us.