As Mereka grew to 100+ GitHub repositories across several engineering teams, no one could see engineering activity across all of them in one place.
Key takeaways
- Automate manual reporting and give the time back to the team.
- Make DORA metrics and team health indicators visible to the whole team.
- AI assistance wrote three in four commits of our build; automated tests kept the numbers accurate.
- Explain every number. Open metrics build trust.
The challenge
Before the dashboard, basic questions such as “How productive was last quarter?” or “Is any team carrying too much work?” meant working through GitHub’s interface, copying data into spreadsheets and hoping the numbers were accurate.
- No central view. Each repository existed on its own, with no organisation-wide view of engineering activity.
- Manual reporting. Engineering managers estimated they spent 4 to 6 hours every month compiling activity reports.
- No industry-standard metrics. DORA metrics, the standard measures of software delivery performance, were impossible to calculate without dedicated tooling.
- Blind spots on team health. We had no systematic way to spot knowledge silos, overloaded teams or collaboration patterns.
- No AI impact tracking. As teams adopted AI coding tools, we could not measure whether they were helping.
The solution
We built our own team analytics dashboard and used AI throughout the build. It collects data from GitHub every 6 hours, calculates key metrics and shows them in an interactive dashboard that leadership and team leads can open at any time.
- DORA metrics at a glance. Deployment frequency, lead time for changes and mean time to recovery, calculated automatically.
- Team health indicators. Workload balance across teams and “bus factor” analysis, which shows where knowledge sits with too few people.
- AI attribution tracking. The share of AI-assisted and human-written code across the organisation.
- Collaboration insights. Network graphs showing how teams work together across repositories.
How AI helped us build it faster
During development, three in four commits to the dashboard (231 of 306) were written with AI assistance. We used Claude for architecture decisions, complex problem-solving and day-to-day coding.
The same shift has spread across Mereka. The share of AI-assisted commits across all our repositories rose from about a third in early 2026 to 78% in October 2026. Team members outside engineering, such as Miranda and Faiz, now push fixes directly without waiting for a developer.1
The results
| Metric | Before | After |
|---|---|---|
| Time to compile the monthly activity report | 4 to 6 hours | 30 seconds |
| Repositories tracked | Manual spot-checks | 100+ with full history |
| Metrics available | 5 to 10 basic counts | 50+ computed metrics |
| Data freshness | Weeks old | Updated every 6 hours |
Source: Mereka’s tech team, October 2026
Managers now spend that time supporting their teams. Because we built the dashboard for production use, over 2,000 automated tests check that the system is reliable and the numbers are accurate.1
How our teams work differently
- Self-service analytics. Managers check the numbers themselves instead of asking for a report.
- Conversations that start from data. Sprint planning and process discussions start from the same information for everyone.
- Every number explained. Each metric shows how it was calculated, its data sources and its caveats, which is why people trust the data.
The hardest decisions were about governance. We had to agree what the metrics would be used for. We removed evaluative labels such as “top performer” and “highest score”, because a dashboard the team does not trust gets quietly undermined.
Source: Rashvin Pal Singh, “The AI ROI stack”, LinkedIn, 25 May 2026.
What’s next
The dashboard is the first step. We plan to grow it into a “team operating system” by connecting more of the tools our teams use every day.
| Integration | What it enables |
|---|---|
| ClickUp | Connect project tasks to code changes, from task assignment to deployment. |
| Slack | Automatic alerts for overloaded teams, sprint completions and unusual activity. |
| Google Drive | Track document collaboration alongside code, so contributions beyond programming are recognised. |
Sources and methods
1 Mereka’s tech team, October 2026
Figures come from Mereka’s tech team: the dashboard’s own GitHub data for commits, repositories and metrics, and engineering managers’ estimates for reporting time before the dashboard. Measured in October 2026.
