
How does DX measure developer productivity?
DX Core 4, AI measurement and 2024–2026 data
Slide contents
1. How does DX measure developer productivity?
DX Core 4, AI measurement and 2024–2026 data
2. Median PR throughput growth was 7.76%
November 2024 – February 2026 · more than 400 companies
3. Five questions about productivity
What changed in the team’s work after adopting AI?
4. DX grew out of DevEx research
Pull Panda, DX launch, Core 4, Atlassian and AI reports
5. Core 4 connects three perspectives
DORA, SPACE and DevEx asked different questions
6. Every key metric has counterweights
Speed, effectiveness, quality and business impact
7. One DXI point: 13 minutes weekly
DX converts a survey index into developer hours
8. Collection changes what a number means
Telemetry and surveys answer different questions
9. Weighting PRs changes the unit
PR throughput and TrueThroughput both describe the flow of changes
10. AI measurement has three dimensions
Utilization, impact and cost
11. A 300% ROI rests on three conditions
DX’s calculation for a 200-engineer organization
12. Review context against five properties
CAFE(S) offers questions for reviewing an agent’s inputs
13. Agents need verifiable guidance
Apply CAFE(S) to a test instruction
14. Surveys capture different sides of work
Time lost in 2024 and perceived AI time savings in the Q4 2025 report
15. The median hides company-level variation
Company-level change in PR throughput, November 2024 – February 2026
16. TrueThroughput rose from 1.43 to 2.51
Q4 2025 – Q3 2026 · DX report sample spans 500+ companies
17. Spend grows faster than throughput
Total AI spend in the DX sample, Q4 2025 – Q3 2026
18. AI relieves local friction, not systemic
Change in DXI factors from April to September 2026
19. AI is mostly busy with upkeep
AI sessions by work type, July–September 2026
20. Quality holds, but CFR has shifted
Change failure rate and recovery time in Q3 2026
21. Higher AI share, longer review wait
July–September 2026 · organizations grouped by AI code share
22. Comparability depends on the metric definition
Check method, unit, denominator, cohort, window and aggregation
23. Producing got cheaper, trusting did not
Productivity shows only across metrics