Engineering metrics
DORA Metrics vs SPACE Framework: Measuring Speed and Sustainability in 2026
DORA tells you how fast your pipeline moves. SPACE tells you whether your engineers can hold that pace. Running one without the other is how teams hit elite benchmarks the week before three senior engineers resign.
TL;DR
DORA measures delivery outcomes, speed and stability. SPACE measures the human conditions that make those outcomes repeatable, satisfaction, flow and collaboration. Neither is complete alone, and AI assisted delivery has widened the gap between them. The fix is one combined scorecard that flags mechanical risk and human risk on the same page.
Key takeaways
- The velocity trap is optimizing DORA in isolation. The dashboard turns green while the team quietly burns down.
- DORA measures the system. SPACE measures the conditions the system runs on. They answer different questions and neither substitutes for the other.
- AI changed the reading. DORA’s research now describes AI as an amplifier that magnifies whatever your organization already is, strong or dysfunctional.
- Overlay the two frameworks and specific pathologies become diagnosable instead of anecdotal.
- Track vital signs, not everything. Five metrics on one page beats forty across six tools.
Engineering leaders who optimize DORA in isolation routinely reach elite benchmarks right before a wave of attrition. The pipeline looks healthy on Monday. Three key engineers hand in notice by Friday. Nothing in the delivery data predicted it, because delivery data was never designed to.
That is the velocity trap, and 2026 made it easier to fall into. When a meaningful share of your diff is machine generated, throughput charts get more flattering at exactly the moment they become less informative.
Part 1. The machine: DORA metrics
DORA, the DevOps Research and Assessment program now run by Google Cloud, measures the outcomes of your software delivery process. It answers one question. Is our delivery pipeline efficient and reliable?
Four keys, with the benchmarks popularized by Accelerate:
Deployment frequency. How often you ship to production.
Elite: on demand, multiple times per day.
Lead time for changes. Time from commit to running in production.
Elite: under one hour.
Change failure rate. Share of deployments that need a hotfix or rollback.
Elite: 0 to 15 percent.
Mean time to recovery. How quickly you restore service after a failure in production.
Elite: under one hour.
The blind spot. DORA is gameable. A team can reach elite numbers by working nights, skipping documentation and deferring every piece of technical debt. It works for two quarters. Then the engine, which is made of people, throws a rod.
Worth knowing: DORA itself has moved on from the tidy four tier ranking. The 2025 research replaced the elite to low performer ladder with seven team archetypes, precisely because the old ladder flattened teams that were fast and fragile into the same bucket as teams that were fast and durable.
Part 2. The human: the SPACE framework
SPACE, developed by researchers from GitHub and Microsoft, measures the conditions that make performance sustainable. It answers a different question. Is this environment healthy enough to keep producing those outcomes next quarter?
Productivity is multidimensional. SPACE covers five areas:
Satisfaction and well-being. Do engineers have the tools and conditions to do good work, and are they running hot?
Performance. Are we delivering reliable, quality outcomes? This is where SPACE overlaps with DORA.
Activity. Counts of actions such as commits and pull requests. Useful as context, dangerous in isolation, and more distorted now than it has ever been.
Communication and collaboration. How easily knowledge moves and how quickly a blocked engineer gets unblocked.
Efficiency and flow. Can people work in uninterrupted blocks, or does the calendar shred the day into fragments?
SPACE is deliberately not prescriptive. That is its strength as a lens and its weakness as an implementation plan, which is exactly why so many teams read the paper, agree with it, and change nothing.
Part 3. What AI changed about both
Two frameworks written before AI assisted development became normal now have to survive it. Here is what the research shows.
DORA’s 2024 report found that rising AI adoption came with a measurable decline in delivery stability. The 2025 report, focused specifically on AI assisted software development, found the throughput picture had improved while the stability penalty persisted, and framed the core conclusion this way: AI acts as an amplifier. It magnifies the strengths of high performing organizations and the dysfunctions of struggling ones. It is not a shortcut past either.
AI does not fix a weak system. It runs that system faster.
Practically, three things break if you keep reading the old dashboard the old way:
Deployment frequency and lead time inflate. Generated code arrives faster, so the numbers that measure arrival speed look better whether or not the change was worth shipping.
Activity metrics lose their meaning entirely. Commit and pull request counts were always weak proxies. When a share of them originates from an agent, they stop describing human effort at all.
Stability and rework become the honest signals. Change failure rate, recovery time and how much code gets rewritten within a few weeks are the metrics that still tell you whether speed was real.
The other useful development is consolidation. The authors behind DORA, SPACE and DevEx collaborated on the DX Core 4, which folds all three into four counterbalanced dimensions: speed, effectiveness, quality and impact. If you are starting from scratch, that is a reasonable place to start. If you already run DORA, the overlay below gets you most of the value without a migration project.
Part 4. The synergy: four diagnoses
The value shows up when you overlay the two frameworks instead of reviewing a delivery dashboard in one meeting and an engagement survey in another. Read together, they name specific organizational pathologies.
1. The burnout rocket
2. The frustrated artist
3. The silent slog
4. The amplified illusion
Part 5. The combined system and sentiment scorecard
Do not track everything. Track vital signs. Five metrics that fit on one page, reviewed in the same meeting, with a named owner for each.
| Dimension | Source | What to measure |
|---|---|---|
| Speed | System | Deployment frequency, weekly average per team |
| Stability | System | Change failure rate, plus rework rate on code merged in the last 30 days |
| Well-being | Survey | Monthly pulse on one statement: I can sustain my current pace of work |
| Flow | Hybrid | Share of working days containing at least two uninterrupted hours |
| Friction | System | Build and CI wait time, plus median time a pull request sits waiting for review |
Blue rows come from your systems and answer whether the machine is working. Amber rows describe the conditions your engineers actually work in. Review them together or the trade-off between them stays invisible.
One rule that keeps this honest: never review a speed metric without its counterweight in the same frame. Deployment frequency travels with change failure rate. Throughput travels with sustainable pace. A number without its counterweight is an invitation to game it.
Part 6. Rolling it out in one quarter
Weeks 1 to 4: instrument the system side
Pull the three system metrics from tools you already have. Do not build a custom warehouse project. Publish the baseline even if it looks bad, and say out loud that no one is being ranked on it.
Weeks 5 to 8: add the human side
Launch a monthly pulse with a small number of questions. Report results at team level and above, never at individual level. Response rates collapse the first time someone feels identified.
Weeks 9 to 12: run the overlay
Put both halves in one review. Diagnose each team against the four patterns above, pick one intervention per team, and commit to measuring whether it moved anything before adding a second.
Conclusion: monitor the driver, not just the car
Track only DORA and you are pressing the accelerator without checking the fuel gauge. Track only SPACE and you are checking the fuel while forgetting to drive.
DORA is your speedometer. SPACE is your engine health check. In an AI assisted organization both readings drift, so the pairing matters more than either instrument did on its own.
Engineering leadership in 2026 comes down to holding two numbers in view at once: how fast the system moves, and how long the people running it can keep it there.
Alex Circei is CEO and co-founder of Waydev. Waydev V2 measures DORA, developer experience and AI impact on one scorecard, so engineering leaders can see delivery speed and its human cost in the same view.
