Developer Impact Index (DII): Measure What AI Actually Ships

The Question Every Engineering Leader Is Being Asked

Your company has rolled out AI coding tools. Cursor. Claude Code. GitHub Copilot. Maybe all three. The CFO wants to know if it's working. The board wants a number. And you're looking at a dashboard full of suggestions accepted, prompts run, and lines of code generated — wondering if any of it actually means anything.

It doesn't. Not in the way that matters.

Today, we're launching Deveval — and with it, the Developer Impact Index (DII). A single, defensible score that tells you whether your engineers are using AI to ship value, or just using AI.

Activity Is Easy to See. Output Is Harder.

The developer productivity measurement problem isn't new. But AI made it worse.

Before AI tools, the effort-to-output ratio was roughly stable. An engineer who opened 20 pull requests probably shipped more than one who opened 5. The signal was noisy, but it was correlated. You could squint at GitHub and get a rough read.

AI broke that correlation entirely.

An engineer running Cursor all day can generate 10× the code volume of a year ago. PRs flood in. Commits stack up. The activity dashboard looks incredible. And yet — the features don't ship. The merge rate is flat. The defect rate is climbing. The business logic never lands.

Activity is easy to see. Suggestions accepted. Prompts run. Pull requests opened. Lines of code generated. Every AI tool ships a dashboard that makes these numbers look impressive.

Output is harder. Merged business logic. Features shipped to users. Defects avoided. First-pass review rates that tell you whether code is actually ready when it's submitted.

The tools your team already uses — GitHub native insights, Jira, Linear — were designed to track effort. None of them were built for the question you're actually being asked: Is AI making my team more productive, or just more busy?

Why Existing Tools Can't Answer This

We've spent time evaluating 38+ engineering teams and analyzing 122+ engineers in private beta. The pattern is consistent.

Teams that look productive on activity dashboards often aren't. Teams that look quiet sometimes ship the most. The gap between apparent motion and actual delivery is real — and it's widening as AI tools become standard.

General-purpose tools aren't designed to surface this gap. They measure what's easy to measure: commits, tickets closed, velocity points. They can't tell you whether the code that got committed solved a problem, introduced a bug, or sat in a draft PR that never merged.

Boards and CEOs are being asked to validate AI ROI. CTOs are being asked to defend their team's performance. And the only data available measures the wrong thing.

That's the problem Deveval was built to solve.

Introducing the Developer Impact Index

The DII is a 0–100 composite score derived from GitHub data. No agent rollout. No tool installations on developer machines. No surveys. Connect GitHub, run the analysis, get results — in under 5 minutes.

The score is built from four weighted dimensions, each measuring a distinct signal of real engineering output:

Code Quality (CQ) — 30 points

Architecture decisions. Maintainability signals. Defect risk indicators. Review signal quality — whether the code submitted for review is ready when it arrives, or consistently returns for rework. This is the dimension that separates engineers who write code that ships from engineers who write code that stalls.

Delivery Capability (DC) — 30 points

Merged business logic. Feature completeness. Actual shipped output. This is the dimension AI tools disrupt most — because they inflate the inputs (code volume) without guaranteeing the outputs (merged, working features). A high DC score means the work is landing.

Engineering Efficiency (EE) — 25 points

PR cycle time. First-pass review rates. Productive AI adoption — the ratio of AI-assisted code that survives review versus AI-assisted code that gets reverted or heavily modified. This dimension tells you whether AI is compressing time-to-merge or just adding noise to the review queue.

Contribution Breadth (CB) — 15 points

Cross-repo ownership. Review participation. Team leverage — how much an engineer multiplies the output of others, not just their own. Senior engineers with low CB scores are often bottlenecks in disguise. High CB scores signal the kind of force-multiplier behavior that scales teams.

Each engineer receives an overall DII score, a red/yellow/green status, and a one-sentence summary. The output is readable by a non-technical CEO without any translation layer. It's designed to be the answer when the board asks the question.

What This Means for Engineering Leaders

Three things became clear across the 38+ teams we've evaluated.

First: AI adoption isn't uniform. In nearly every team, 20–30% of engineers are using AI tools to genuinely compress cycle time and ship more. The rest are generating volume that doesn't convert to delivery. Without DII, you can't tell them apart from the outside.

Second: The board question doesn't go away. "Is AI helping?" is now a standing agenda item for engineering at the board and C-suite level. The DII gives CTOs and VPs of Engineering a defensible, data-backed answer — not a gut check, not a narrative, not a dashboard screenshot.

Third: Contractor and outsourced team evaluation is broken. Most companies have no objective way to assess external engineering output. They rely on progress reports and demo calls. DII applies the same analysis to any GitHub contributor — internal or external — giving procurement and engineering leadership an objective negotiation anchor.

Five Minutes. No Dashboards. Board-Ready by Default.

Setup takes under 5 minutes. Connect GitHub. Deveval analyzes automatically. You get a report.

The report is one click to PDF export. It's formatted for a board room, not a Slack channel. Non-technical founders don't need an engineering translator to read it. Red means attention required. Yellow means monitoring. Green means performing.

There are no ongoing meetings required. No weekly dashboards to maintain. No check-ins with team leads to synthesize before you understand what's happening. The data comes from GitHub, which your team is already using. Deveval reads the signal that's already there.

How to Get Started

Deveval is available today with three tiers:

• Free: Up to 5 engineers, 1 repository, one-time 30-day evaluation. Enough to run a real assessment on a core team before committing.
• Starter: $3.50/seat/month (billed annually). Up to 100 seats, monthly automated evaluations, PDF export. Built for teams that want ongoing accountability.
• Team: $5.60/seat/month (billed annually). Multi-team comparison, larger organizations, cross-team benchmarking.

The free tier is real. Connect your GitHub and run your first report before you decide whether this is worth a budget conversation. Five minutes is short enough to do it before your next board meeting.

The Era of "We're Using AI" Is Over

Every engineering team is using AI tools now. The question is no longer whether you've adopted them — it's whether they're translating into shipped output.

That's what Deveval measures. That's what the DII answers.

Connect your GitHub at developereval.com. The first report is on us.