At KIT, we use rigorous methodologies and a combination of monitoring and evaluation (M&E) and innovative data sources to help our clients and partners use evidence to learn, improve, and make better-informed decisions.
Our Data Science & Products service offering is built on three core elements:
1. Advanced data & predictive analytics
At KIT, we are dedicated to providing our clients and partners with maximum insight into their data. From data to evidence. We develop and implement methodological frameworks that use the strengths of advanced econometrics, machine learning, and other predictive modelling techniques to assess for evidence of change using key primary and secondary, structured and unstructured data sources.
2. Interactive reporting & dashboards
From evidence to insights. With reproducible reporting workflows, dashboards, and data storytelling approaches we give our clients digital tools that facilitate access to organised and intuitive insights from their data.
3. Data & information management
Sharing the data with funders and other stakeholders boosts our clients’ transparency and Access to data and evidence. For a successful implementation of M&E and Knowledge Management systems, data and information must be accessible in an easy-to-navigate system. We help clients design, improve, and maintain their data and information management systems.
Our Data Services
Machine learning to estimate income of farming households
What is the income of farming households and does it afford a decent living standard? We developed a supervised machine learning model that can estimate the net annual income of farming households from a few simple variables. In its prototype state, the model has shown great accuracy and potential for application using a sound MEL framework.
Segmentation of beneficiaries using unsupervised machine learning
Different beneficiary groups require a different set of interventions. Using unsupervised machine learning models, we created beneficiary clusters to form the basis of various profiles of individuals and households. Per cluster, we tested which intervention delivers the greatest impact, so our clients can specify their approach.
Predicting success using decision trees
Decision trees are among the best algorithms to generate easy-to-interpret nonlinear effects. Determinants of success are ordered based on their relative importance, and coefficients are assigned to quantify their impact on the outcome variable. We have employed decision trees for an impact evaluation for FMO, where we tested which characteristics determine farmers’ success in loan acquisition.
Intuitive and actionable dashboards
Our data scientists can present your data in a comprehensive, yet intuitive and actionable dashboard. We first help clients define their theory of change and indicators, which we use to develop interactive data dashboards. For AGRA (Alliance for the Green Revolution in Africa), we developed an interactive dashboard that allows the data to be accessed by a country- and crop-based focus.
Geo-spatial assessment of forest change
KIT conducted the end-term evaluation of the “Connecting Production, Protection & Inclusion Partnership Programme” in Indonesia, Brazil, and Liberia. We assessed the outcomes and impacts, including a geo-spatial analysis of deforestation and land cover change.