In human services, we often treat program and system evaluations like the final scores of ball games. A positive result signals a win that paves the way for continued or expanded funding, while anything less sends funders looking for a new direction and programs back to the drawing board. But the best programs and systems improve the way the best teams do: they try something, test it, learn from it, adjust, and repeat. They never stop adapting, and they are never satisfied.

When learning is embedded in their day-to-day operations, using measures that staff can collect and leaders can act on, programs don’t have to wait years for validation. They can test, learn, and course-correct in a matter of weeks, leading to improvements in services and faster achievement of core outcomes.

I spent much of my career designing and implementing large one-and-done evaluations. They each took many years to complete, and each, for the most part, showed that the programs did not meet their goals. These evaluations used rigorous designs and included studies of context and implementation challenges, but they did not provide timely results for funders, program leaders, and staff to use to improve what they were doing. They did not move at the speed of the program and the changing needs of families. They focused on the question of “did it work?”, looking backward, rather than “how can it work better?”, looking forward.

At the same time, long-term impact evaluations remain a critical component of the evidence-building toolkit. Although they may not always align with the timelines or information needs of policymakers and practitioners, they remain indispensable for establishing the foundational evidence that guides future investments, scaling decisions, and policy reform.

More recently, I’ve had the opportunity to co-design and implement continuous improvement efforts—work I’ve found especially invigorating. These efforts focused on small, iterative tests of change, guided by a clear overarching framework and grounded in co-creation among leaders, staff, peer organizations, and, most importantly, clients and customers.

Far from sidelining data, they made it real and actionable for everyone involved, developing measures that reflect what staff and families actually care about and using results to dig into the “why” behind outcomes. They also moved at an energizing pace, with colleagues pushing one another to test new ideas, learn quickly, and continuously improve.

There still is an important role for rigorous, long-term impact evaluation, particularly for mature programs where multiple approaches may be in play. Long-term impact approaches can complement—not compete with—continuous learning.

There is no one best way to engage in continuous learning. Many frameworks for continuous improvement exist, and each has its strengths. One agency or one team can engage the process alone, but there is value in including broader partners both vertically (including leadership, supervisors, working, and families) and horizontally, with teams working on the same challenge in other service or geographic contexts.

The heart of any of these efforts, however, is something like the PDSA cycle (plan, do, study, and act):

• Plan: Identify the “pebbles in your shoe”—the small but persistent challenges affecting staff or outcomes.
• Do: Test a practical, manageable change. Start small and move quickly.
• Study: Pair simple quantitative measures with real-time feedback from staff and families to understand what is happening and why.
• Act: Refine, expand, or abandon the approach based on what you learn.

The power of this approach lies in its cadence. With regular cycles of testing and reflection, programs can detect change early and adjust in real time. 

Being flexible and data driven is the key to success—letting go of what didn’t work and doubling down on things that are moving outcomes in the right direction. There are several important elements of this:

1. Improvement teams need to be given the authority to make changes (perhaps small changes, but at least enough to move the needle on key outcomes).
2. Outcomes need to be clear, defined, and measurable.
3. Some form of routine data collection is required. This may come from existing data systems, or it may require implementing a short survey of workers or another form of data capture.
4. Teams need to be given protected time for reflection away from the press of the day-to-day.

The child welfare field is now experimenting with this approach broadly. The U.S. Department of Health and Human Services’ Children’s Bureau is encouraging this approach for improving child welfare services, moving away from more traditional methods of federal review of state efforts.

More efforts like this are needed. Our field needs fewer delayed answers and more real-time learning. Social services providers, policymakers, and researchers should continue to invest in rigorous, long-term evaluations—but only alongside systems that allow programs to improve as they operate. Continuous learning reduces the gap between evidence and action. To ensure better outcomes for families, governmental and nonprofit agencies need to move from judging programs after the fact to helping them get better every day.