The framework that DORA itself outgrew
For ten years, DORA's four metrics were the gold standard for measuring software delivery performance. Deployment frequency, lead time for changes, change failure rate, mean time to recovery. The four numbers told you whether your engineering org was elite, high, medium, or low performing.
In 2025, DORA itself moved away from the elite/high/medium/low buckets and introduced seven archetypes based on eight measures, with labels like "The Legacy Bottleneck" and "The Harmonious High Achiever". The 2025 DORA Report also provided the first in-depth look at how AI is changing core metrics, finding that AI adoption improves throughput but increases delivery instability.
The CTO implication: the four-metric DORA dashboard you built in 2022 doesn't capture what your team is actually doing in 2026. The framework was honest about its evolution. Most engineering measurement programs haven't caught up.
Real Estate Investment AI
Your models aren’t wrong. Your data is. Here’s how real estate teams fix AI failures before they cost millions.
What DORA still does well
Worth keeping. The four classic metrics still capture genuine signal:
- Deployment frequency: how often you ship
- Lead time for changes: commit-to-production duration
- Change failure rate: percentage of deploys causing problems
- Mean time to recovery: how fast you recover from incidents
If you're not at those numbers, work on getting there. If you are, the DORA-plus-five framework below is what's next.
The five metrics elite engineering teams add to DORA
Metric 1: AI-attribution productivity gains
The 2025 DORA report found teams using AI report a 7.5% increase in documentation quality, 3.4% better code quality, 3.1% faster code reviews, and 1.8% reduction in code complexity. Those numbers are useful precisely because they're modest. They tell you what AI is actually contributing in your specific org.
Track this by surveying your engineers monthly: what fraction of code shipped this month was AI-assisted, where did AI most accelerate the work, where did AI add review burden. The aggregate over a quarter tells you whether your AI tooling is producing the value the vendor's pitch promised.
Metric 2: PR review time at team and org level
The AI-era hidden tax. PR review time grew 91% in high-AI-adoption teams while throughput grew 21% and merge volume grew 98%. The review bottleneck became the new constraint.
Track PR review time as a first-class metric. Watch the trend. If it's growing faster than throughput, the AI gains are being eaten by review burden, and senior engineer attrition is the next signal you'll see.
Metric 3: Senior engineer attrition rate
The unspoken health metric. Engineering orgs that are working well retain senior engineers. Engineering orgs that aren't, don't.
Watch the rolling 12-month rate. A healthy org sits below 15% senior attrition. A struggling org climbs above 20%. The trend matters more than the level. If your rate has gone up two years running, the working conditions are deteriorating regardless of what the throughput metrics say.
Metric 4: Cost per shipped feature
The business-aligned metric. Engineering throughput in feature count means nothing if cost per feature is climbing. Track the trend over rolling quarters.
Programs that capture both the engineering cost (loaded engineer time per feature) and the infrastructure cost (compute, storage, AI inference) of shipping a feature can defend their budget to finance with specifics. Programs that don't are running on the CFO's patience.
Metric 5: Time from incident to architectural change
The learning metric. Programs that ship one or two architectural improvements per significant incident are learning from operations. Programs that ship none are accumulating debt that will compound.
Track this as a count over rolling 90 days. Healthy orgs ship 2-5 incident-driven architectural improvements per quarter. Stagnant orgs ship zero.
What the new measurement looks like in practice
The DORA-plus-five dashboard has nine metrics, not four. Each tells you something the original four didn't.
A team can be elite on the original four DORA metrics and unhealthy on the additional five. The throughput is high. The senior attrition is high. The PR review time is climbing. The cost per feature is rising. The incident-driven learning is flat. The system is shipping but isn't healthy.
Conversely, a team can be medium on DORA and healthy on the rest. They ship moderately fast, recover well, and the metrics underneath show retention, sustainable review pace, controlled cost, active learning. They're the team you want to invest in, not the team with the prettier headline numbers.
The 2026 measurement question isn't "what's our DORA tier." It's "what does our DORA tier look like underneath."
How Logiciel fits this conversation
Most engineering leaders who reach out to us about metrics have a working DORA dashboard and a growing sense that the headline metrics aren't capturing the org's actual health. They're seeing the underneath signals (review time, attrition, cost per feature) move and don't have a framework for what to do.
The work we do is the metrics layer extension. We help instrument the additional five, build the dashboard, and pair the metrics with the operating cadence that converts the signals into decisions.
PropTech AI Infrastructure Roadmap
They’re stuck because the data layer they need doesn’t exist yet
Call to Action
The 30-minute move
Book a working session with a senior Logiciel engineer. Bring your current DORA dashboard. We'll walk through what's missing and tell you which underneath metric is highest leverage to instrument first.
Book the 30-minute metrics session →
Frequently Asked Questions
Are DORA metrics still useful?
Yes. They capture a real layer of performance. They just don't capture all the layers. Add to them; don't replace them.
How often should we review these metrics?
DORA metrics weekly, the additional five monthly. The trends matter more than any single reading.
What's the most underrated metric?
Senior engineer attrition. Most engineering scorecards don't include it. The orgs that watch it carefully outperform on every other dimension over time.
We're a small team. Do all nine apply?
The trends apply at any size; the thresholds shift. A small team can run informal measurement on most of these and still capture the signal.
What if our DORA numbers look good but underneath looks bad?
That's worth flagging to leadership. Headline numbers and underneath numbers diverging is the leading indicator of a team that's going to flip from "elite" to "in trouble" within 2-3 quarters. --- Sources cited: - 2025 DORA Report: seven archetypes, AI impact, percentile distribution - DORA elite benchmarks: <26hr lead time, <5% CFR, <6hr MTTR - AI impact: 7.5% docs / 3.4% code quality / 91% PR review time
