In December 2024, the Innovate team closed the Global Parenting Initiative’s (GPI) 2024 webinar series by looking back on three years of building digital parenting tools. IDEMS leads the team with our Kenyan partner INNODEMS, and we work alongside the University of Oxford and partners across GPI. The panel was chaired by IDEMS’ Director, who is also GPI’s Innovate Lead. It included INNODEMS’ Director; IDEMS mathematical scientists and data specialists who lead app development, chatbot development and data analysis, several of whom first joined us as Impact Activation Fellows; and two University of Oxford doctoral researchers.
Our main takeaway is that the most valuable results came less from any single product than from how the work was organised: across studies, across countries, and with the people building the technology involved in the decisions.
Two starting points, shared learning
When GPI set out to digitise the evidence-based PLH programmes, we started from the view that there is more than one valid way to do it, because different approaches serve different purposes and audiences.
ParentApp started from research. It went through a pilot, an optimisation study and then an RCT in Tanzania, and the results went well beyond our expectations. We wanted the trial to test the parenting content, not simply the effect of using an app, so it needed a strong active control. INNODEMS built one, WASHApp. It proved so popular that the researchers briefly worried it would weaken the comparison, but ParentApp was still effective against it.
ParentText started from implementation. During COVID-19 it was rolled out quickly in the Philippines, Malaysia, South Africa and Jamaica. What we learned from those deployments fed into ParentText 2.0, with versions in Malaysia, South Africa, China and Mexico.
Both tools share the same content but took different routes, and each learned from the other. When ParentText reached its own RCT in South Africa, it borrowed ParentApp’s active-control approach and built WASHText as its control.
Practical lessons beyond the formal findings
In the ParentApp for Teens optimisation study, which formed the core of one panellist’s doctoral research, many design decisions were made before results from earlier stages were in. One was leaving out phone-call support because it was too complex and costly to implement. Drop-out showed gaps that there were not yet systems to address. Those observations, together with INNODEMS staff in the field talking to parents and fixing technical problems, changed how the RCT was run. The trial reached completion rates above 90%, which is almost unheard of for a digital intervention.
A second doctoral researcher focused on the content. Families in early phases said they wanted more sessions and more to learn. A digital programme is not limited the way weekly in-person sessions are, so, with IDEMS, Clowns Without Borders and UNICEF, this work explored how learning through play could strengthen relationships between parents and adolescents and keep families engaged for longer.
ParentText faced a different test. Our chatbot development lead described running up to five country deployments at once. That pressure pushed the team to build a global version that partners could customise on their own and feed back into without the versions drifting apart. When the South African trial opened, recruiting 20,000 users was expected to take about a month, and 10,000 joined in the first four hours. The system held up because it had been built for exactly this.
None of these lessons would show up in a results table. They come from people who can see across studies and are trusted to act on what they see.
Foundations, not one-off apps
Each deployment pushed us to build something reusable. Our app development lead describes the result as an app builder with three layers: developers write the underlying code, a middle layer holds the logic, and authors edit content without coding. That structure matters for projects such as ParentApp for Kids, where a research-led version in Tanzania and a version with PLH in Kuwait started from one structure and have since diverged in places.
Capacity that stays local
INNODEMS was part of GPI from the start, both to build its team’s skills and so that it can take on many of IDEMS’ roles over time. Its Director described how graduates with little technical background learned to build apps by authoring ParentApp and ParentText in spreadsheets. The team has since built WASHApp; Parenting for Respectability app; and a conflict-mapping app that uses socio-economic, climate and news data. Because the team belongs to the communities it builds for, its solutions start from local needs.
Our view: give innovators a seat at the table
For us, this is what “impact activation” means. Much of the impact we see can be traced back to learning from elsewhere in GPI, and that cross-learning is what makes a research programme more than a set of separate projects.
This was possible because of how GPI was designed. In most research projects, the technology team is brought in late and told what to build. In GPI, Innovate was a core theme from the proposal stage, with IDEMS on the executive team, and funders including the LEGO Foundation and Oak Foundation supported that integrated approach. There was reflection during the webinar that GPI has learned that involving innovators earlier in designing research studies makes the work more responsive and creates more chances to learn, a lesson it wants to share with the field.
We agree. In digital work the details matter, and the people building the tools can often be the first to notice connections between studies. A research programme that gives them a voice in decisions undoubtedly will learn faster and achieve more.