How much capacity should we allocate to innovation?

Start with 10% and prove you can keep SLOs green. Add planned 70/30 spikes for specific bets. Don’t exceed 15% baseline without executive commitment and evidence that delivery is stable.

What if urgent delivery needs keep stealing the time?

Make exploration a first-class capacity line. If a week gets cannibalized, repay it within the quarter. Track stolen time as debt and report it in the Portfolio Review.

How do we prevent experiments from leaking into prod?

Require flags by default, canaries for rollouts, namespace fences for infra, and CI checks that enforce labels and guardrails. Give each exploration a kill-switch owner.

What metrics convince finance and product this is working?

Lead time to first signal, kill rate, option value (bets promoted), zero SLO regressions, and cost per learning. Show 90-day trends, not anecdotes.

Where does AI exploration fit?

Exactly the same. Flagged endpoints, budgeted inference, red-team evals before exposure, and an explicit data policy in the RFC. The guardrails matter even more with LLMs.

Culture · Oct 6, 2025 · 9 minute read

Stop Saying “20% Time”: A Real Playbook for Innovation Without Blowing Your Roadmap

Innovation time that executives won’t cut, teams won’t resent, and customers won’t feel.

Back to all posts

Stop Saying “20% Time”: A Real Playbook for Innovation Without Blowing Your Roadmap

Key takeaways

Implementation checklist