Scores of lawsuits have pushed retirement plan sponsors to shorter, easier-to-navigate menus, but - as Ian Ayres and Quinn Curtis argue in this work - we've only scratched the surface of retirement plan design. Using participant-level plan data and straightforward tests, Ayres and Curtis show how plan sponsors can monitor plans for likely allocation mistakes and adapt menus to encourage success. Beginning with an overview of the problem of high costs and the first empirical evidence on retirement plan fee lawsuits, they offer an overview of the current plan landscape. They then show, based on reforms to a real plan, how streamlining menus, eliminating pitfalls, and adopting static and dynamic limits on participant allocations to certain risky assets or 'guardrails' can reduce mistakes and lead to better retirement outcomes. Focusing on plausible, easy-to-implement interventions, Retirement Guardrails shows that fiduciaries need not be limited to screening out funds but can design menus to actively promote good choices.