Signals & Levers • Elisabeth Hendrickson, Joel Tosi & Charles Humble • GOTO 2026

This conversation reframes software delivery through systems thinking, arguing most orgs are misled by illusions of progress, predictability, and control. Agile/DevOps gains have plateaued; AI and rising complexity expose bottlenecks that rituals and platform teams can’t mask. Microservices often create a false sense of speed while integration delays dominate. Change-control boards masquerade as quality levers but throttle flow.

The authors push leaders to stop treating proxy metrics as truth. Velocity, tokens, and lines of code are opinions in numeric clothing; they optimize for motion, not value. Shift to flow-centric signals—cycle time, lead time variability, WIP—and acknowledge that measuring is itself an intervention: signals become levers as people respond to what’s watched.

Use variability analysis (control charts) to separate signal from noise. Don’t punish normal fluctuations or chase single root causes in complex systems—there usually isn’t one. Build causal models to surface reinforcing loops and find small, high-leverage nudges (e.g., isolate test data, reduce cross-team dependencies) that flip vicious cycles into virtuous ones. Expect time delays; avoid oversteering and change theater when results don’t land immediately.

Culture consistently beats policy. Treat culture as an organizational immune system: it resists misaligned mandates. You can shift it, but only by feeding desired behaviors—language, rituals, origin stories, and rewards (e.g., celebrate cross-team collaboration over heroics). Transformation by decree fails; transformation by incentives, stories, and consistent signals sticks.

Practically: adopt a multidimensional signal framework (capacity, risk, economics, adaptability, etc.) to prevent myopic fixes. Kill velocity targets and token leaderboards. Replace CCBs with progressive delivery and guardrails. Fund enabling constraints—test data bootstrapping, decoupled architectures, focus-limited planning—to reduce variability and restore predictability.