Skip to content

Using Apache DevLake for DORA + Copilot Correlation

One implementation path for proving engineering impact


This playbook shows one implementation path: deploy Apache DevLake with the gh-devlake CLI, then use Apache DevLake to ingest Copilot, GitHub, and delivery data into a common model for dashboarding and analysis.

One option, not a requirement

You do not need Apache DevLake to correlate Copilot adoption with delivery outcomes. Use this playbook when you want a prebuilt open-source stack for ingestion, normalization, and dashboards. If you already have another engineering analytics platform, reuse the same baseline, segmentation, and ROI steps with your existing data stack.

Decision-maker deck

Download the Apache DevLake decision-maker guide for a concise overview of DevLake architecture, fit vs. alternatives, cost and ownership tradeoffs, phased adoption guidance, and example Copilot impact dashboards.


Prerequisites

  • Analytics-Ready playbook completed — automated data collection and dashboards in place
  • Docker installed locally, or an Azure subscription for cloud deployment
  • GitHub PAT with scopes: repo, read:org, manage_billing:copilot

Time Investment

Plan 1–3 days depending on organization size and GitHub project complexity.


Step 1: Deploy Apache DevLake via gh-devlake CLI

Install the gh-devlake CLI extension and deploy:

gh extension install DevExpGBB/gh-devlake
gh devlake deploy local --dir ./devlake

System Requirements

Requires Docker with at least 4 GB RAM. Stack includes Apache DevLake, its backing services, and Grafana.

gh extension install DevExpGBB/gh-devlake
gh devlake deploy azure \
  --resource-group devlake-rg \
  --location eastus

Azure Costs

ACI typically costs $2–5/day. Stop the instance when not in use.


Step 2: Configure the GitHub Connection

gh devlake configure connection add \
  --plugin github \
  --org my-org

You'll be prompted for your PAT. Apache DevLake ingests repositories, PRs, deployments, and commits.

Multiple Organizations

Repeat for each org. Apache DevLake aggregates data across orgs.


Step 3: Configure the Copilot Connection

gh devlake configure connection add \
  --plugin gh-copilot \
  --org my-org

This configures Apache DevLake's gh-copilot plugin to pull daily active users, acceptance rates, feature-level usage, and seat utilization.


Step 4: Add Scopes and Create a Project

gh devlake configure full

This interactive command lists connections, lets you select repositories (start with 3–5 key repos), creates a project, and triggers the initial Apache DevLake sync.

Start Small

Pick 3–5 representative repositories with active development. Expand later.


Step 5: Wait for Initial Sync

Org Size Repos Sync Time
Small < 20 5–15 min
Medium 20–100 15–30 min
Large 100+ 30–90 min
gh devlake status

Or check the Config UI at localhost:4000 (local) or your Azure endpoint.

Warning

Incomplete syncs produce misleading dashboards. Wait for completion.


Step 6: Open Grafana and Review Dashboards

URL: http://localhost:3002 (local) or Azure Grafana endpoint · Credentials: admin / admin

Apache DevLake ships Grafana dashboards once the data is synced.

Adoption Dashboard

Panel Shows
Active Users Over Time DAU/WAU/MAU trend with adoption tiers
Acceptance Rate Trend Suggestion quality perception over time
Feature Usage Mix Completions vs. chat vs. CLI vs. PR summaries
Seat Utilization Active seats ÷ assigned seats

Impact Dashboard

Panel Shows
PR Cycle Time by Tier Do high-adoption teams merge faster?
Deployment Frequency Are high-adoption teams deploying more?
Change Failure Rate Is quality maintained as velocity rises?
Code Review Duration Are reviews faster with Copilot-assisted code?

Step 7: Establish a Baseline Period

Scenario Baseline Approach
Pre-Copilot data available Use 4–8 weeks before Copilot enablement
Copilot already deployed Use low-adoption teams/repos as control group
No historical data Set current metrics as baseline; measure over next 8 weeks

Metrics to baseline: median PR cycle time, deployment frequency, change failure rate, code review turnaround.


Step 8: Analyze Correlation Patterns

PR Cycle Time by Adoption Tier

Tier Definition Expected Pattern
High > 80% daily usage Shorter PR cycle time
Medium 40–80% daily usage Moderate improvement
Low < 40% daily usage Baseline-like

Look for high-adoption teams closing PRs 15–30% faster than low-adoption teams.

Deployment Frequency

Track deploys/week per tier. Look for a positive trend as adoption increases. Flat frequency + shorter cycle time may indicate more effective batching.

Change Failure Rate

Ideal

CFR stays flat or decreasing while velocity increases — faster without sacrificing quality.

Red Flag

Rising CFR alongside higher velocity — investigate whether reviews or test coverage are being skipped.

Code Review Duration

Compare median time from PR open → first review for high- vs. low-adoption teams. Hypothesis: Copilot-assisted code follows conventions more consistently, speeding reviews.


Step 9: Build the ROI Narrative

ROI = (Time Saved × Blended Developer Rate) - Copilot License Cost
  1. Quantify time saved — e.g., 20% PR cycle time reduction × 500 PRs/month × 2 hrs = 200 hrs saved/month
  2. Apply blended rate — 200 hrs × $85/hr = $17,000/month recaptured
  3. Subtract license cost — 200 seats × $19 = $3,800/month
  4. Net ROI — $17,000 − $3,800 = $13,200/month

ROI One-Pager Template

Use the ROI one-pager template to create a polished executive summary.

Correlation ≠ Causation

This analysis identifies correlations, not causation. Present findings as "teams with higher adoption also show improved DORA metrics." Strengthen claims with controlled rollouts, developer surveys, time-series analysis, and confound analysis.


Result: A working Apache DevLake environment for correlating Copilot usage with DORA metrics, plus baseline measurements, correlation analysis, and an executive-ready ROI narrative.


What to Do Next

  • Deploy Apache DevLake and start your initial data sync today
  • Identify your baseline period and document pre-Copilot metrics
  • Share the Impact Dashboard with your VP of Engineering within 2 weeks
  • Build your ROI one-pager using the template