Attempts to scale up information and communications technologies (ICTs) in developing countries are often unsuccessful. Many innovations that hold huge promise fail to grow or deliver a significant development impact.
The architects of ICT innovations have blamed a catalogue of technical problems and management errors for this failure. But few have looked for clues in the processes of scaling.
Conventional wisdom suggests that scaling an ICT begins with a pilot study, led by a firm or an NGO. Then the focus turns to building the structures needed to diffuse the innovation, such as marketing and distribution processes.
But have successfully scaled ICT innovations actually followed such a process?
The history of two successful cases in Kenya – mobile phones and the M-Pesa mobile money service – tells a very different story.
Recent research I conducted with my colleague Richard Heeks examined these successful innovations. 
Surprisingly, ICTs did not remain as static as the conventional scaling model suggests: such technologies were not finished after the trial period. Reinvention continued after the pilot.
For instance, some firms were introducing mobile phones into Kenya as long as ten years ago, but it was only in recent years, as they reached more remote rural users, that these firms altered their designs to improve their function in areas with poor mobile reception.
They adapted after realising that their phones could not meet these users’ needs. The availability of different languages, the phones’ durability in dusty settings and a decent battery life where electricity is sporadic are examples of features that significantly affect how useful phone models are to low-income users.
And reinvention revolved around much more than the technical elements. It also included adaptations to such things as pricing, distribution and marketing.
M-Pesa is a vivid example of this. In its early days, the service was mainly marketed as a solution for urban-to-rural money transfer under the slogan “Send Money Home”.
But over time, marketers have also focused on new functions that benefit lower income users – to secure liquid cash by turning it into e-cash when travelling to a dangerous area, for example. As a result, the innovation has been perceived more positively and used more widely.
Perhaps more surprisingly, we found that this constant reinvention was often influenced – even driven, at times – by actors other than the lead firm: local entrepreneurs, for example, or active users.
This is something that’s often overlooked about M-Pesa. Much of its dramatic growth can be linked to business model innovations undertaken autonomously by networks of independent dealers and agents involved in operating the service.
This occurred when agents began to engage in less formal subcontracting relationships, reducing barriers to entry for smaller entrepreneur-run kiosks and kick-starting the expansion of agent services into rural and slum areas.
These innovations were later absorbed into M-Pesa’s core structure as external processes and business models were adapted and improved.
Constant reinvention is central to successful scaling. It allows the innovation to adapt to problems that weren’t anticipated during pilots. And partners outside the lead firm are key to reaching low-income users.
Because they are closer to the action, such partners can have a better idea than large, often remote firms about how innovations can be most useful on the ground.
The global mobile handset industry is another example of this. While firms have often been rather distant from low-income users, those that have more successfully spurred take-up in Africa, such as Nokia and Tecno, have actively built links with their distribution partners to get a better perspective of real-world problems and needs.
This suggests that the growth of ICT innovations will be limited if firms or organisations blindly follow the conventional model of piloting innovations and then scaling for diffusion.
Local partners are vital
Although our research mainly focused on market-based innovation processes, these findings are likely to be relevant to a wider set of ICT-for-development interventions – for example, ICT systems piloted in one part of the world and then scaled across a large number of sites without any further consideration given to adaptation.
Of course, pilot studies are an important first step. They help in building basic understanding. But firms and development organisations should be open to the constant reinvention that will enable them to successfully scale innovation.
To put this in perspective: if M-Pesa had fulfilled its original goal, it would have remained a system for microfinance distribution. It was its openness to reinvention that made it a more widely useful ICT service in the long run.
Partners close to low-income users play a central role – so engaging them is one way to encourage reinvention.
In essence, successful scaling to lower income groups needs to actively integrate bottom-up reinvention. Original innovators may need to take a step back – perhaps controversial for firms and NGOs long educated in the mantras of decisive management and long-term planning, rather than an ‘adapt and follow’ philosophy.
This is about building and nurturing partnerships, particularly with those involved in operating the innovation, and about ongoing ‘systemic’ learning.
Firms or development organisations need to offer partners the flexibility, freedom and skills to reinvent – and then avoid closing down reinventions that sit outside the ICT’s original aim, so they can absorb those that can ensure the technology reaches as many people as possible.
Christopher Foster is a researcher on ICT use in emerging and developing markets at the University of Oxford, United Kingdom.
His recent research includes work on local innovation in the Kenyan mobile sector and an examination of the impact of fibre optic internet on firms in East Africa. He can be contacted at firstname.lastname@example.org and on twitter @cgfoz
 Chris Foster and Richard Heeks Innovation and scaling of ICT for the bottom-of-the-pyramid (Journal of Information Technology, December 2013)