Tips for becoming a data-driven organization

CANBERRA — As organizations strive to better develop their programs, investment in good data practices is critical. Still, there remain barriers in the development sector that may prevent an organization from using data properly or achieving as much as they could. At Devex World, Harpinder Collacott, executive director at Development Initiatives, Maria Ruth Jones, lead of Development Impact Evaluation Analytics at the World Bank, and Laura McGorman, global policy lead at the Facebook Data for Good portfolio, discussed how development organizations can become data-driven. Expanding upon this discussion, Collacott and McGorman spoke with Devex about the skill sets organizations should pursue and the partnerships that can help build long-term success. “At a certain point, data sets large and small should influence everything you do.” --— Laura McGorman, global policy lead, Facebook Data for Good portfolio Here is the interview, edited for length and clarity. What is the best way to approach data strategies within organizations to achieve the best value? Collacott: Firstly, collaboration is essential to strengthen the data ecosystems — within organizations and between organizations. The people who produce the data need to be speaking to the people who will use the information emerging from the data. The users need to understand the limitations of the data, what it can and cannot do, as well as the producers clearly understanding what questions those using the information are seeking to answer. Remember, most people want the information which comes from the data — they don’t want the data itself. They want to know what it says and what it means. Translate it from data to information and linking it to their problem is critical to ensure use. But data cannot tell you what to do. What it can do is give you insights, shine a light on where the problem may be and ensure you can monitor progress to address the problem. So a data-driven organization needs to be very clear what problem they are trying to address, and then what data does it need to inform its thinking. What does leadership and culture look like in an organization that is using data well to drive its business needs? McGorman: I think being data driven has to be habitual and baked into everything you do. Research on habits says that they should be second nature — like brushing your teeth. In a data-driven organization, I think that means that you have a daily set of practices that leverage insights over gut feel — whether it's by surveying your employees or your partners to see what they think before proceeding with a plan, or by making sure you review progress on last year's goals before you set goals for next year. At a certain point, data sets large and small should influence everything you do. What are the tools, expertise and knowledge that should exist? What investment is needed to support this? McGorman: I think people are the most important investment you can make towards building a data-driven organization. Employees will be inspired by other people that they work with, but you need a small cadre of data leaders within a business paving the way. I'm inspired every day by the talented people who work at Facebook on data science, geospatial data, and differential privacy. Working with people like that helps pull me forward. Collacott: Consider your data management approach and invest in data management. Consider what data governance processes you have, put these in place so that collaboration and coordination across organizational silos is supported. Make sure you invest in understanding data ethics, privacy, and rights issues. Create communities to champion the data-informed approach and bring together the stakeholders from across the silos. And invest in long-term capacity and skills development rather than one-off training, workshops, or events Data science is an important function within our organization. We have also invested in data management and storage solutions which allow data to be transformed and all the transformations clearly linked in the data storage system to create efficiencies. Training up more people to be confident in downloading and integrating the data from there has been important but also creating dashboards for the data has been really helpful to make sure those who want to make decisions from the data can read it in an accessible way. They need it translated for them. When considering the approach to be data-driven, how important is trust in data, and how do you build this? Collacott: Trust is critical. If the data is not trusted it will be dismissed. Building trust in data requires there to be strong regulations around the data internally — frameworks for collection, ethics, and data privacy legislation. But there also needs to be clear information about the limitations of the data — what it can and cannot be used for so there are not unreasonable expectations and manipulation of the data to tell a particular story. McGorman: I think people's trust in data is critical, but I also think that people's trust in each other is more important. As Facebook Data for Good, we try to be as open as possible about potential biases in our data, which are fairly straightforward and based on smartphone and internet penetration in a country. That being said, I think it's more important that all of us working together on data partnerships trust one another's expertise and desire to make responsible decisions. At Facebook we strive to make data sets that are privacy protecting and work to be transparent about when our data might not be representative. In turn, our partners know how to appropriately triangulate our data with other sources and discuss it with certain caveats in mind. The trust that goes both ways in these kinds of partnerships is paramount and even more important than trust in the data itself.