Tag Archives: education

What does it take to scale a tablet-based foundational learning program to all the primary  schools in Malawi? In this 3-part interview, Joe Wolf and Kira Keane describe how Imagine Worldwide has approached that challenge and share some of what they have learned in the process.  The tablet-based program at the center of Imagine Worldwide’s work, developed by software partner onebillion, serves as a supplement for regular instruction, with each child in a school spending a targeted 150 minutes per week working independently on problems related to reading and mathematics. Imagine Worldwide partnered with the Government of Malawi to rollout the program in 500 primary schools in 2023-24, with the ultimate goal of expanding to all 6000 primary schools in Malawi, serving 3.8 million learners in standards [grades] 1-4 annually. Joe Wolf is the Co-CEO and Co-Founder of Imagine Worldwide and Kira Keane is the Director of Communications. (Photos/graphics are from Imagine Worldwide unless otherwise noted.)

TH: Can you describe for us some of the key steps or phases you went through as you developed your work to test and then to scale-up this tablet-based program in Malawi? 

Joe Wolf: The first phase of our work was all research oriented. We wanted to see if these learner-centric tablet models could work – were they really effective for children? – before asking under-resourced systems to spend time, energy, and capital on them. That meant we had a prolonged research phase that included nine randomized controlled trials. That was across different contexts, different languages, different implementation models, different countries – really exhaustively trying to prove that these solutions can, in fact, add significant value. 

The second phase was what we call “learning to scale:” What are the processes that need to be done repeatedly well to scale within these contexts? We purposely spread our work out across seven countries, with different implementation models, different implementation partners, different types of structures to really test what needs to be done repeatedly well so that these systems can adopt the work at scale. Then, only in the last three years, we’ve put the pedal down and said, “Okay, I think we’re ready to really think about scaling.” And we were only able to act on scaling thanks to the leadership of the government of Malawi, who saw the learning gains of our pilot programs and saw how this edtech intervention could support their national goals of improving foundational skills.  At that point in 2022, we served around 6000 children, but we increased it to about 700,000 children by the beginning of 2025. That’s a 100x increase in the last two years, which I think is a testament to the scalability of the model, the execution of the team, and the leadership of our government partners. 

TH: What’s the third phase? Implementation? 

JW: I would say it’s scale plus continuous improvement. Now, our research is less efficacy oriented and more implementation oriented. How do we make it better and better and better? To address that, we have four levers we focus on: 

• Access: How do we serve more and more children and make the solutions easier and easier to implement? 
• Cost-effectiveness: How do we bring down the recurring costs to be as low as possible? We’ve brought costs down around 75% in the last five years, and we think there’s still room to go. Our key inputs are all highly deflationary, so we’re getting better economies of scale as we grow. Right now, we’re at about seven dollars (USD) per child per year. We think we can get that under five dollars (USD) as we get better economies of scale. 
• Advocacy: How do we use data to improve the implementation model in the software so that the efficacy of the program continues to go up and up and up? It’s one of the beauties of technology that it can iterate and improve. You’re not building a building and putting in books and then five years later it’s deteriorated. We actually have the ability to use data to continuously improve through this flywheel of innovation.

“It’s one of the beauties of technology that it can iterate and improve. You’re not building a building and putting in books and then five years later it’s deteriorated. We actually have the ability to use data to continuously improve through this flywheel of innovation.“

• Sustainability. How do we work with our government partners to build operational and financial sustainability?  And how do we do it starting day one, where we’re building the “muscles” within the existing education system, as opposed to the classic approach of starting off outside the system and then trying to hand it off to the system. Too often, if you haven’t done a good job of building that internal muscle, and then things fall apart. So we’ve really taken the system strengthening approach, acknowledging that there are capacity and infrastructure gaps within the countries where we work and that there are key functions that need to be built that don’t currently exist within some of these systems. We’ve tried to give it time so that, by the end of the implementation phase, the system has already been doing the work for an extended period of time. That way, you don’t have this fall off as you try to hand-off everything to the system itself.

Kira Keane: I just want to underscore a couple of points that Joe made. For Imagine, this notion of the continuous improvement loop, it’s not like we did things, something went wrong and we’re like, “Oh, we have to fix this.” This was an intentional design element from the very beginning: How do we get continuous feedback to improve both the software itself and the implementation model? And the other point is that our key question is “How do we serve as many children as possible?” The need is so immense and the population growth will be so intense over the next 10-15 years so we really need to be focused on scale. That means working with our government partners to aim for generational impact, really looking at country-wide scale, and focusing on how we design for that.

“The need is so immense and the population growth will be so intense over the next 10-15, years so we really need to be focused on scale.  That means working with our government partners to aim for generational impact, really looking at country-wide scale, and focusing on how we design for that.”

JW: I’ll add two more things to what Kira said. The ecosystem is exhausted by pilots – by small things that don’t scale, that don’t have evidence, that take a lot of time and resources. Scale from day one very much aligns with where the governments are. They have a big problem with the lack of foundational learning among their students, and they need big solutions. Little, tiny things are just distracting and take too much time and energy. The second thing is that we have positioned our organization to be temporary in nature, so our job is to put ourselves out of business as quickly as we possibly can. We don’t see these as “Imagine Worldwide” programs in Malawi or “Imagine Worldwide” in Sierra Leone. These are programs of the government in Malawi and of the government in Sierra Leone that we are helping to support. We’re helping to build capacity and infrastructure to build muscle within the systems. But as soon as the government is ready to maintain this on its own, we are more than pleased to step out of the way and to move on to the next challenge. I think that positioning is really important for the governments. It’s really important for the funders. It’s really important for us and our team. Too many times, an NGO establishes itself and 50 years later, the NGO is still there, doing the work. We need this work to be sustainable within existing systems. Part of that is a commitment for us to get out of the way. We have to believe in sovereignty and the power of governments to run themselves, while also acknowledging that the use of technology in a place like Malawi is new, and so there is going to be a period of time where we have to build some functions that do not currently exist.

TH: That certainly resonates with my experiences in the US where we’ve seen multiple improvement efforts collide in schools in ways that can actually undermine their capacity for improvement. What made Malawi a good context for you to work on scale-up?  

JW: The work in Malawi actually predates the partnership with Imagine. There was a program called “Unlocking Talent,” with the software developer onebillion that became our partner. The onebillion CEO went to Malawi, I think, 15 years ago, fell in love with the country, and developed the product. The first product they developed was in Chichewa, in Malawi. In other words, this was not developed in the West and then adapted to the context. This actually was developed within the Malawian context. We became a research partner to look at impact and to help do the RCT work. That has now evolved into a much more scalable model that we call the BeFIT Program. It’s serving standards [grades] one through four, whereas the first program was only standard two. 

There have been a whole bunch of iterations along the way to develop our general approach, but it basically evolved by thinking about what it would take to actually scale the program much more cost effectively to many more students in more systems. If you look at the other places that we worked, you’ll see that we started with finding local partners, mostly local NGOs, some local social-oriented businesses, and then turned over a lot of the functions to those local partners to see what worked in different contexts. From that, we have built a series of centralized functions that we’re now drawing on in our country partnerships, as opposed to having it be completely decentralized. We learned a lot from the initial more decentralized exploration, but we’re now in the process of creating more standardization. Part of scaling depends on acknowledging that you can’t have fifty different bespoke operations. You need to have systems and standards and data systems. When you have 6000 children in Malawi, using a total of 1000 devices, you can do some things by hand; but now we’re trying to serve millions of children in Malawi, with hundreds of thousands of tablets. We now need data driven systems in order to be able to manage that equipment in the field. 

TH: Let’s follow the arc of that evolution in Malawi. What are some of the steps that were crucial to your learning and to the development of the model?

JW: In Malawi, we took seven or eight years to do the research and to get the right level of government buy-in to understand what was working. That included learning things like what’s the infrastructure for the typical school in Malawi? Just to give you the context, that means more than 100 children per teacher and inadequate levels of teacher training. There’s very rarely basic infrastructure in place, so no electricity and certainly no internet connectivity. That’s the reality of the average class in Malawi. So as you think about the components of our model that have emerged the first was what you would call the infrastructure component. We put solar power into all of our schools, addressing questions like: 

• Where do solar panels go? 
• How does the solar electricity feed a bank of lithium batteries? 
• How do the tablets get stored and secured overnight so that they’re charged and they’re safe? 
• How does all that equipment get distributed to children in a really efficient manner, so that you’re getting as much asset utilization as possible and as much learning time as possible? 

In the end, our research consistently shows that the number of minutes each student uses the content is directly correlated to the level of learning. So we’re addressing these 101 things that need to be done in terms of the infrastructure and operations to maximize that time on task. And that has to take into account that the school day and the school periods are very short in Malawi and you have a lot of children in the classroom. So even just getting kids in and out of a classroom is a lot harder than in many other contexts.

TH: You just described those complexities really effectively, but for those of us who aren’t familiar with the context, can you go into it even more deeply? What does it really take to get a program like this up and running at scale? 

JW: I think that in addition to a foundational learning organization, we are, in a lot of ways, also a supply chain logistics company. Learning gains are still our north star, but the reality is you’re talking about a phase one of BeFIT that involved launching the program in five hundred schools in five months across half of the country of Malawi, including very rural districts. So we have to deal with the logistics of getting five hundred secure storage cabinets into those schools. We have to deal with the logistics of getting 100,000 tablets distributed across those 500 schools and of getting the solar equipment put into 500 schools. That’s a significant operational lift, and you have to approach that with a level of rigor in terms of those key functions, if you’re going to be able to scale, and you’re going to be able to do that on time. And we had to do that on budget in the middle of a huge macro-economic meltdown in terms of currency and raw materials. In the grand scheme of things, once the equipment is in place, kids can get learning very, very quickly. There’s not a huge lift in terms of adult training. There’s not a huge lift in terms of the role of the adult in the model itself; the content has been built to be autonomous, meaning the child can be self-directed. The tablets themselves have been built to be very robust. A lot of enhancements have been made to make the tablet durable. There’s a long battery life so it can be used throughout the day. Every part of the tablet has been built with screws so that a component can be swapped out if something breaks. So every part of the context has been taken into account in order to get that equipment into the field and utilized. This is one of the big learnings: you have to start with the context in mind, and you have to start with the learning objectives in mind. You then make a series of software decisions, and then you make a series of hardware decisions. Too often in education, it goes the other direction, where people buy stuff, but then they haven’t really thought about what’s going to go on the stuff? What’s the training required? What are the charging and security components of it? What is our learning objective at the end of the day? You have to start with learning, move into the context, and think about all the infrastructure decisions that need to be made in order to make that learning possible in that context. 

“You have to start with learning, move into the context, and think about all the infrastructure decisions that need to be made in order to make that learning possible in that context.”

KK: I think it’s also important to flag that in working on the logistics we included the government from day one. That means things like using the delivery trucks the government already had. Trying to manage that coordination may have been a little slower or less efficient in some ways, but too often people design an implementation model, put a bow on it, and then hand it to the government without including them from inception. 

Next Week: Building the Capacity to Improve and Sustain Foundational Learning Through Government and Local Partnerships in Malawi: Joe Wolf and Kira Keane on the Evolution of Imagine Worldwide (Part 2)