Measuring impact on the SDGs through AI
How do you know whether you're truly contributing to the SDGs? It's one thing to set objectives and priorities, it's quite another thing to get there.
All of the best plans have feedback loops: you need insight into what’s working and what’s not so that you can be more effective going forward. But impact measurement is tricky and resource-intensive. Getting insights out of data is your first hurdle. This is where Artificial Intelligence can help - either by generating it yourself or by using models and data built by others.
The Sustainable Development Goals (SDGs) have unleashed a wave of enthusiasm: organisations across sectors and across the globe are looking at how they can contribute. They’re choosing priorities and drawing up strategies, but our research shows that they’re still not measuring much. This is a real pity because data can do two things: first- it can give insight into whether you’re on track with implementation and second- it can tell you whether your plans are having the desired effect, allowing you to adjust them so that they are even more effective.
This second aspect is commonly referred to as impact measurement and it is something that organisations have been struggling with for decades. Impact data is often difficult or costly to gather because it involves looking at what is happening in long and sometimes complicated value chains. In the past, only a few organisations have been willing to devote the time and resources necessary to gain these insights. But luckily, times are changing.
It’s not enough to simply say you’re contributing to the SDGs - you have to do it. And data is getting cheaper and easier to access.
We're in an Artificial Intelligence boom. The past few years have known major advancements in Machine Learning. Models pop up left and right, automating tasks like recognising trees on maps, sorting species of fish, and even mapping bird calls. These tasks might seem mundane, but they are everything but. Indeed, they automate work normally done by trained staff, more often than not scarce white-collar specialists. This opens up a world of possibilities. Indicators that used to be prohibitively expensive can now be approximated frequently and cheaply. Provided you know how to use AI.
The availability of data is growing at a staggering rate: a projected ten-fold increase between 2017 and 2025. Chances are that what you'd like to measure is being measured: if not by you, then by someone else. Suppose you're interested in increasing the bee-friendliness of the farms you source from. One indicator is the availability of flowering crops at the farm: one in four cultivation strips. You could hire your own private army of geospatial analysts to tally these on a satellite image. Or: you could make a machine learning model.
The expertise to create and apply Machine Learning models is a scarce resource. Some organisations are privileged enough to be able to grow it in-house, but this is not something available to everyone. The good news is that these models and data, once available, can be shared and used by different parties across the value chain to understand and increase their SDG impact.
If we want to achieve the SDGs we need to make sure that the largest number of organisations have access to the best available data and models. Of course, this will not always be for free. In some cases, organisations have built the data and models at great expense and so we understand the hesitation. However, there is a win-win to be found by tapping into the market for valuable indicators. The solution is a well-equipped social enterprise.
For organisations that own the data and know how to apply the models: think about setting up a social enterprise to make this expertise available to others.
1) Make an inventory of other organisations that you expect to need the same indicators you need. Good candidates are other parties along your value chain, competitors, or complementary products.
2) Make your business case based on how much this indicator is worth to you and to others. Assess whether the collective need warrants the creation of a model.
3) Form a social enterprise to fulfil this need. Build a model, and offer it as a service it to all interested parties.
As organisations increasingly monitor and manage their contributions to the SDGs, the demand for good data is bound to grow. We expect to see a rise in data-driven social enterprises. Anyone with Machine Learning expertise can capitalise on this market, from start-up to corporate. Not only the creators, but the users profit as well.
Jeroen Goudsmit, mathematician and manager Forensic Services for PwC The Netherlands, Linda Midgley, SDG Lead for PwC EuropePwC