Skip to main content
Community Contribution

Three Ways Data Science is Changing Monitoring and Evaluation

May 20, 2019
John Mataya

Three Ways Data Science is Changing Monitoring and Evaluation

As a Data Scientist working in the context of monitoring and evaluation in the international development industry, I get exposed to interesting data science problems every day. It’s fascinating to see firsthand how an industry is borrowing innovations from Silicon Valley and applying them to make a lasting difference in people’s lives across the globe. Data Science methods borrowed from computer science are generally a bit atypical in the field of international development which is strongly rooted in the domains of social science. However, looking forward we can expect some changes on the horizon. Here are a few of the changes that I’m most excited about within the intersection of data science and monitoring and evaluation for international development.

Smarter Surveys Lean on Data Science

Surveys are common and expensive for many international development projects. But in the future, we are more likely to start combining big data resources – think satellite images and mobile phone data – with smaller scale surveys and machine learning to replicate more traditional surveys that are often used on our monitoring and evaluation projects Using big data and machine learning, we’re at a point where we can begin to predict attributes, demographics, and behavior of individuals that can be compared to the relationships observed in smaller scale surveys. This approach really began with Blumenstock’s work in Rwanda, but it’s now being institutionalized in social science thanks to pioneers like Matthew Salganik and we’re likely to see more computational social science techniques being applied to our monitoring and evaluation surveys Utilizing such a methodology could be transformative when used on projects related to migration, refugees, conflict, and famine – giving decision makers near real-time information to act and intervene. This approach isn’t without drawbacks and obstacles. Getting access to big data resources is not always easy and still can be expensive. However, access to data is getting easier and the costs of doing machine learning are dropping and becoming more commonplace. We are close to witnessing some exciting changes in how surveys are done.

Text Analytics Will Unlock New Insights in Our M&E Work

Text is not traditionally thought of as a data source that can bring new insights to international development projects. However, Data Scientists armed with programing language such as Python and R can now scrape and transform PDFs, websites, interviews, and transcripts into meaningful insights. We are just beginning to apply these types of data science techniques to our work in the context of international development projects. At DAI, we are analyzing and quantifying text in new ways. For example, we are examining word frequencies, word correlations, sentiment, and topic models in some of our monitoring and evaluation work and even on USAID reports handed over to our projects as PDF documents. This is not perfect for every situation, but over time the resources available to analyze text and utilize it as a data source to monitor and evaluate our projects will be easier and become more widely adopted.

Computer Vision Opens New Unforeseen Opportunities

Computer vision better known as facial detection or object detection is becoming more commonplace and I believe it is only a matter of time before such technology can be used in new and unforeseen ways on our monitoring and evaluation projects. Computer vision is a subset of machine learning and artificial intelligence which has exploded due to the dropping costs of higher resolution imagery and advances in machine learning models that use convolutional neural networks to more accurately identify patterns such as people and objects in images. This hit home recently when I was in a Power BI meet up group hosted at Microsoft where Patrick Leblanc was able to classify and quantify audience reactions based on their facial expressions during a presentation in real time. The demonstration used Azure’s out of the box facial recognition platform with Microsoft’s reporting platform, Power BI. While this type of technology is still in its infancy and currently requires customization, it won’t be long before it becomes easier to set up and implement such as system. Imagine how such a platform could change how international development projects monitor participation at events, training, and conferences if we want to lean on new metrics beyond the more traditional approaches that might just count event participation. In conclusion, this is an exciting time to be working on monitoring and evaluation projects as a Data Scientist. As data science techniques begin to change social science and international development, we can expect some positive changes – better decisions, improvements to activities, and superior results for our international development projects.

John Mataya, Data Scientist, DAI Global