STIR staff run a meeting with local government officials in Karnataka to prepare them as network leaders with the organization’s system-led delivery model.

Nonprofits involved in areas that address basic needs like education or health often expect that if a project or intervention is successful, government will ultimately adopt and manage it for the longer haul. For a project to reach its potential scale and sustain impact over time, many view this as the most realistic end game. Conventional wisdom holds that governments want fully formed, tested versions of programs or interventions so that it can easily scale them up.

But is this really accurate? Recent evidence suggests that even after these “adolescent” innovations successfully pass through randomized controlled trials (RCT)—the highest bar for evidence—only a handful ever reach meaningful scale or lead to policy change.

Our own experience during the past five and a half years at STIR Education, which aims to reignite and scale teacher motivation, has led us to question this wisdom. Peer organizations who also have experience in this area—including Landesa, Clinton Health Access Initiative, Mothers2Mothers, and the Department for International Development’s Research on Improving Systems of Education program—have similar concerns.

At STIR, we believe programs are most enduring and effective when they are delivered in partnership with governments from the outset and fully integrated into social systems. In partnership with the Ugandan and Indian governments, we have connected with 75,000 teachers and 2.6 million children in five years. Recently, we’ve been asked by both governments to fully integrate our teacher development model into national and state systems. Through this effort, we can support our government partners to improve learning outcomes for 60 million children during the next five years.

A local government official leads a network meeting in Northern Uganda.

To reach this point, we’ve made many mistakes, and our painful lessons may help others undertake a similar journey.

Lesson one: Begin at the end. Start by thinking about your goal, and work backward from there.

We soon learned we needed to test government demand right from the start, almost before we started piloting the intervention. Early input from governments often fundamentally changes the nature of the innovation itself. For example, our initial conception of STIR focused primarily on the micro-practices of teachers such as the physical layout of classrooms. But in government corridors, we heard concerns about teacher motivation—an issue the findings from our pilot reinforced. We had to make a big pivot toward building our model around networks of teachers, which allowed us to scale our model more quickly and more deeply embed it in the education system.

Lesson two: Cost really matters.

Early on, a senior Indian official pointed out that the government annually spends 10,000 rupees (about $200) on one child’s education, and that amount wasn’t going to change any time soon. That became our yardstick; we wanted our intervention to cost less than 1 percent of that per-pupil amount. That level—just $2—is where we are now. In our next chapter of growth, we hope to further lower costs to just 0.2 percent (40 cents).

Honestly, at times we’ve been sorely tempted to relax these cost constraints. If we climbed to a higher price, we were sure we could achieve short-term impacts much faster. But we’ve generally managed to resist. In our experience, beginning with a higher-cost innovation makes it too difficult later on to shift gears to a lower cost structure. In addition, the higher expense brings with it the temptation to add more inputs from outside the system, such as groups of parallel support staff. By contrast, maintaining the low cost from the start forces you to work within the existing system and find latent resources you can tap into.

This approach is more challenging and more painful in the short-term. It takes time to figure out which resources and structures are truly latent, and which are unavailable for a reason.

But it pays long-term dividends. In a recent RCT, we calculated the value of the improvements in student learning, as reflected in improved student earnings over their lifetime. We found that each dollar spent on the innovation translated into a one hundred-fold increase in earnings. Each dollar in was working incredibly hard to achieve that impact.

Lesson three: Working well with governments is itself a core competency.

As social entrepreneurs, we tend to focus obsessively on our innovations, and neglect whether we’ve built the core competencies that will give us a realistic chance to support a government scale up what we do.

Government systems can be challenging for social entrepreneurs and NGOs. We focus on a “co-creation” model with our government partners, which includes clear and transparent communication, rigorous evidence-based learning, redesigning the teacher development model based on the evidence, and working to make innovations come alive by inviting ministry officials to visit our partner districts and schools.

Using this approach, we’ve been pleasantly surprised by the willingness of government partners to iterate or change direction. For example, the RCT mentioned previously suggested that we could do more to improve gains in reading, so we’ve been brainstorming with the government in Uttar Pradesh about how to collaborate in the same districts with other respected NGOs, such as Room to Read, that already have technical prowess in this area. We’ve been humbled to see that, once trust is established, leaders in these systems can commit to pushing through necessary changes.

Lesson four: Adopt an “insider-outsider” approach.

If you gain the trust of a government, you can be granted the leeway to do things differently—what we call an “insider-outsider” approach. For example, STIR teacher network meetings, in which teachers learn and gain support for making tangible change in their classrooms, is now part of the official government system. But it still feels very different from a typical government teacher-training program in almost every way; we encourage teacher-led discussions to identify and solve problems, and then reflect on and adapt the solution before evaluating it, all in a supportive environment.

In addition, an NGO that develops these government adoption “muscles” is in a better position to protect the project. Many external validity studies of social innovations show that when an NGO implements the project, the impact is strong—but once the government takes over, it collapses. With a well-constructed partnership with government, that collapse need not be inevitable.

Our program quality data is much the same no matter whether our own staff or government staff are responsible for delivering the program. This is for two reasons. First, when government takes ownership, people in the system consider it more legitimate than when we as an external agency do it. Second, when government runs it, it becomes part of an official system: The innovation becomes the default for all the teachers in the system, removing barriers to access.

Overall, our experiences suggest the need for a more positive and humble approach toward government. Indeed, NGOs and social entrepreneurs add value by piloting and developing promising innovations. But sustainably scaling up and building capacity within the system requires that we partner with government early on as a co-creator. Innovations can be gradually introduced and integrated into the system. All of us—whether implementers, influencers, or funders—should give equal, or even primary, credit to the work of governments and the leaders within it.

Perhaps it’s time for a new approach—a turn away from the “lean startup” lens we normally apply to social entrepreneurship—and to rethink the image of the rugged entrepreneur going it alone. This new mentality would promote “lean collaboration” with governments: An approach based on mutual commitment, co-creation, iteration, and—on both sides—a big dose of humility.