Traditionally, there’s been a three-step journey to social impact: innovate, learn and scale. Have ideas, identify what works, then amplify them. But more and more, innovations are not simply changing what we scale, but how we scale.
These recent innovations include behavioural insights, human-centred design and data science. They may well be familiar — often the private sector is ahead of the curve when it comes to squeezing value out of innovations. But now the public and non-profit sectors are starting to use them to scale too.
Most often, they turn to technology to cut costs: they want to do the same thing, but cheaper. Technology can achieve that — but that’s only a fraction of its potential. Emerging techniques and technologies can enhance impact, or open the horizon for a radically new kind of intervention. They are changing the scale up process at every step, from dissecting the problem to designing the intervention and delivering it.
Designing the intervention
Whether an intervention will scale successfully is often determined at the very beginning by its design. Getting that blueprint right is crucial, and two innovations are changing how it’s done: data science and human-centred design.
Start with data. Using computers and data science techniques to sift through vast amounts of data can reveal otherwise imperceptible patterns. In the design stage, this can help with understanding the problem at hand — and providing a potential entry point to tackle that problem.
The Behavioural Insights Team (BIT), a spinout from the UK government, shows how it can be used. They recently chewed through over a decade’s worth of road safety data in Essex, a region in the UK. This let them pop some preconceptions. For example, occupational drivers were not more likely to be involved in accidents, nor were visitors just passing through. In fact, drivers in accidents were disproportionately local to the area — as were drivers with prior speeding convictions.
Clearly that previous brush with the law had not changed their behaviour. So BIT decided to modify the letter that drivers receive after they are caught speeding. Using behavioural insights, they simplified and rejigged it. A randomised controlled trial found that the new letter reduced reoffending by 20%.
Data science identified the entry point; behavioural insights sculpted the intervention. Together, they let BIT avoid acting on prejudice and instead design a highly targeted and scalable intervention.
Behavioural insights are central to a broader innovation making its way into the public and non-profit sectors: human-centred design. This requires placing the human perspective at the heart of the design process: asking not what people need, but what people want, and how they will interact with the product.
Whitney Pyles Adams, who runs the Scale X Design accelerator of the NGO CARE, believes it is vital for any scale up. “One thing that is very prevalent is that if you, for example, don’t have clean water, then any solution for clean water is a better one — but we know that’s not the case,” said Adams. “People would sometimes rather go without a latrine than use a terrible one. So we have to shift the mindset for our staff to understand desirability: what the recipient cares about, which may not be the same as what we care about. If you don’t have a strong value proposition for the people you’re scaling for, it’s absolutely not going to scale.”
Expand access and segment users
One of the clearest examples of how technology has led to a step change in scale can be seen in the expansion of access.
Broad internet coverage and the increasing pervasiveness of smartphones and tablets mean that more people than ever are connected. They’re breaking two bottlenecks that once stopped scale: geography and material resources.
Take Brazil’s new approach to schooling in the Amazon. Not long ago, many children were faced with a choice when they reached high school age: move to Manaus, the state capital, or stop going to school. But recently the state government created a so-called Media Centre in which lectures are broadcast via a bi-directional camera, allowing teachers in Manaus to lecture and interact with students in hundreds of classrooms at the same time.
Clearly this is a compromise: the standard of education won’t be as good as having teachers in the classroom. But in an environment with very limited resources, it has made it possible for 300,000 students from 2,300 remote villages scattered across the Amazonas to continue their schooling.
The flip side of expansion of access is the growing need to segment users according to how best to serve them. The private sector excels at doing this kind of audience segmentation with data tools. It’s an approach that has potential, both good and bad: it can reinforce discrimination, but it can also provide personalised service and help distribute resources more efficiently.
The Graduation Approach, a program to tackle extreme poverty that we covered in a case study, is starting to experiment with segmentation. Their program is extensive and expensive, involving a mixture of asset transfers, coaching and food support.
At the moment, the program treats the extreme poor as a homogeneous category — which it clearly isn’t. They are currently figuring out how best to slice and dice that group, by gender, demographic, geographic or on psychosocial lines. Some may receive cash rather than asset transfers like tools or livestock; others may be asked to pay something back once their business gets going.
Segmentation can be problematic, but it is essential to maximise cost-effectiveness — something any scale up should aspire towards.
Deeper and sustained engagement
In addition to expanding access, digital devices offer the opportunity for deeper engagement. Increasing numbers of programs are designing apps for this purpose.
Take the Nurse-Family Partnership, a 40-year old US program that sends nurses to visit young poor first-time mothers for up to two years. It recently hired Hopelab, a social innovation lab, to augment its impact through technology. After extensive interviews with mothers and nurses in the program, they discerned that the bond between them was the key to the program’s success. Then, the question was: “How do we effectively combine the human touch with technology?” said Margaret Laws, CEO of Hopelab.
They came up with Goal Mama. This app is multipurpose: it helps mothers and nurses communicate between visits, it digitises much of the administration and it introduces goal-setting and nudge elements for the mothers. The idea was to make the program “more engaging for the clients, more efficient for the nurses, and more easily replicable across a distributed network,” said Laws. “This is a program that has 287 sites across the US and is in five other countries.”
Apps aren’t limited to industrialised countries. In fact, some governments are incorporating them into their Graduation Approach. In these instances, human-centred design is especially important. “Apps are typically designed for people who have some level of education and literacy, including digital literacy,” said Ana Pantelic, Chief Strategy Officer at Fundación Capital, a development organisation. “But we were saying: can we design apps for people who are illiterate and have maybe never used a device?”
In Colombia, for example, they are using apps that teach financial literacy and business skills to replace some of the in-person coaching in the program — one of its costliest elements.
All the data produced by these apps is a complementary benefit. There’s GPS data to track tablets as they circulate through the community; data from how people use the app that helps the designers refine it; and integrating data sources from, for example, the app and mobile money can paint a picture of how the program is shaping local commerce.
Through data, an app can have knock-on effects on a program’s design, segmentation and evaluation. All of which can sharpen its effectiveness and efficiency.
Evaluation and forecasting
Over time, whether or not it involves an app, any scale up should accumulate huge amounts of data. Unfortunately many public and non-profit bodies lack the basic data infrastructure to ensure that this data is high quality, well organised and accessible. Fixing that should be a priority, because parsing big data with AI and machine learning offers big opportunities.
AI involves computers going through vast amounts of data to find patterns and make predictions — all without explicitly being told what to do. It can help make sense of increasingly large and impenetrable data sets, meaning programs could move away from relying on intuition and rough estimates, and onto using AI-powered predictions.
Take Crisis Text Line, a texting service that anyone in a crisis can reach out to in order to get support from a trained crisis counsellor. The organisation recently analysed the messages they have received over the years and the outcomes they led to. “Now they know, within the first two texts, whether this is going to need to go to 911,” said Laws.
This isn’t a case of AI replacing humans; it’s a case of AI complementing humans. AI can be a helping hand for people with overwhelming case loads — a common feature of programs that scale up.
But AI is not simply a piece of software to be installed. AI systems require careful preparation of data, as well as lots of monitoring and customisation. The best route to AI for many organisations may be through partnerships that give them a step change in capacity.
Program design, coverage, engagement and evaluation — these are just some of the steps where emerging technologies and techniques can change the process of scaling up.
Ultimately, though, they cannot replace the basics: understanding the organisation that’s delivering a program, the population receiving it, and the system it needs to survive in.
Technology is a bonus — but it can be a substantial one. And once it is integrated, the data starts rolling in. The public and non-profit sectors just need to develop the data infrastructure to manage it, and capacity to analyse it.
(Picture credit: Pexels)