Data Science, Statistics & Mixed Methods
NIRAS offers advanced analytical solutions founded on robust data science, statistics, and mixed-methods research, that complement our robust quantitative design and implementation capabilities. We possess in-house expertise in predictive modelling, causal inference, and multivariate analysis. Our teams have successfully delivered seventeen (17) experimental and quasi-experimental studies across Asia and Africa.
In Figure 1 we provide an overview of how we see data science enhancing evaluation.
Data Science Research & Methods
Unifying methods to turn data into answers
Data Collection Services
Leveraging conventional and untraditional data sources
Data Quality Assess & Improve
Valid data supporting robust and innovative analysis
Data Integration & Visualization
Combining and dynamically presenting data to understand trends and relationships
Data Driven Decision Making & Support Services
Rapid response evaluation and support for decision making
Social Media & Digital Outreach
Harnessing social media for service delivery and to enhance social and behavioural research
Data Science Research & Methods
Unifying methods to turn data into answers
Peter Hargreaves
Expert Profile: Peter is a MEL data scientists and geo-spatial expert with over seven years of professional experience on MEL for sustainable development, poverty reduction, livelihoods, and conservation. His analytical strengths comprise advanced statistical modelling, satellite remote sensing and geospatial analysis, experimental evaluation designs, and the robust management and interpretation of large socio-economic and spatial datasets.
Peter Hargreaves
Expert Profile: Peter is a MEL data scientists and geo-spatial expert with over seven years of professional experience on MEL for sustainable development, poverty reduction, livelihoods, and conservation. His analytical strengths comprise advanced statistical modelling, satellite remote sensing and geospatial analysis, experimental evaluation designs, and the robust management and interpretation of large socio-economic and spatial datasets.
We regularly combine quantitative and qualitative data to contextualise and deepen understanding. For qualitative data analysis, we utilise Computer-Assisted Qualitative Data Analysis Software (CAQDAS), such as MaxQDA and Atlas.ti, to facilitate collaboration, enhance efficiency, and increase transparency.
Over the past six years, NIRAS has been integrating data science into its monitoring and evaluation work, continuously refining its application. Our MEL experts regularly share these innovations with the broader evaluation community, including within the European Evaluation Society ( EES), the UK Evaluation Society (UKES) and the German Evaluation Society (DeGEval), webinars, and collaborations with networks such as the Swedish Evaluation Society and the Monitoring Evaluation, Research and Learning MERL Tech’s Natural Language Processing ( NLP) working group.
The adoption of generative AI and large language models (LLMs) has further expanded our analytical capabilities, particularly in handling complex datasets such as large text corpora, multimedia sources, and social media content. NIRAS adopts an incremental, quality-focused approach to generative AI, emphasising rigorous oversight and ethical practice to ensure transparent and responsible use.
Data Analysis Services Examples
Sentiment analysis for decision-making
As a partner of the UK Government’s Global Monitoring, Evaluation and Learning Partnership (GMEL) consortium tasked with supporting the Conflict, Stability and Security Fund (CSSF), NIRAS explored how big data and data science can support decision-making by highlighting emerging trends.
One of the products developed by NIRAS in 2019-2022 used the former Twitter (now X) to understand how the host population perceives migrants and refugees over time and across different locations. Our team created a web-based interactive dashboard that displays a map of the countries of interest, highlighting negative and positive attitudes on migration over time. They utilised a state-of-the-art AI model, bidirectional encoder representations from transformers (BERT), which was explicitly trained on Twitter data. The dashboard was created using R, a free and open-source statistical programming language, and connected directly to Twitter via an Application Programming Interface (API), allowing data to be automatically updated daily.
Sentiment analysis for decision-making
As a partner of the UK Government’s Global Monitoring, Evaluation and Learning Partnership (GMEL) consortium tasked with supporting the Conflict, Stability and Security Fund (CSSF), NIRAS explored how big data and data science can support decision-making by highlighting emerging trends.
One of the products developed by NIRAS in 2019-2022 used the former Twitter (now X) to understand how the host population perceives migrants and refugees over time and across different locations. Our team created a web-based interactive dashboard that displays a map of the countries of interest, highlighting negative and positive attitudes on migration over time. They utilised a state-of-the-art AI model, bidirectional encoder representations from transformers (BERT), which was explicitly trained on Twitter data. The dashboard was created using R, a free and open-source statistical programming language, and connected directly to Twitter via an Application Programming Interface (API), allowing data to be automatically updated daily.
Natural Language Processing for Strategic Insights
The Ford Foundation’s Building Institutions and Networks (BUILD) initiative was a five-year (2016-2021), $1 billion investment by the Ford Foundation in the long-term capacity of social justice and civil society organisations worldwide. NIRAS leveraged automated natural language processing to analyse grantee partners’ progress reports and other information from the Foundation’s grant management system. NLP results complemented the insights that emerged from focus group interviews and qualitative analysis of grantee narratives, and helped categorise grantee partners’ approach to institutional strengthening and the long-term effects of organisational development on mission impact. Results were used to test the BUILD Theory of Change and inform the design and delivery of the BUILD initiative. Key lessons learned can be found in this report.
Natural Language Processing for Strategic Insights
The Ford Foundation’s Building Institutions and Networks (BUILD) initiative was a five-year (2016-2021), $1 billion investment by the Ford Foundation in the long-term capacity of social justice and civil society organisations worldwide. NIRAS leveraged automated natural language processing to analyse grantee partners’ progress reports and other information from the Foundation’s grant management system. NLP results complemented the insights that emerged from focus group interviews and qualitative analysis of grantee narratives, and helped categorise grantee partners’ approach to institutional strengthening and the long-term effects of organisational development on mission impact. Results were used to test the BUILD Theory of Change and inform the design and delivery of the BUILD initiative. Key lessons learned can be found in this report.
Generative AI for Complex Portfolio and Meta-evaluations
AI’s ability to analyse large volumes of text makes it particularly well-suited for reviewing extensive collections of documents, such as evaluation reports across programmes, countries, and sectors. Multilingual models further broaden the scope by enabling the analysis of reports in multiple languages. Beyond identifying recurring themes and gaps, AI can extract and organise content into structured formats—such as tagging lessons learned, evidence gaps, or recommendations—for further analysis.
NIRAS is applying machine learning and generative AI in the Evaluation of Sida’s Strategic Partnership for Humanitarian Assistance, which assesses the operational performance and effectiveness of 25 humanitarian organisations seeking partnership with Sida. AI will enable the NIRAS team to expand the evidence base and analyse a wider range of sources. By combining Natural Language Processing (NLP) techniques with traditional assessment methods, the team will be able to process and analyse volumes of texts that would be impractical to assess manually within a short timeframe. The NIRAS team will use this tool to analyse documentation from the humanitarian organisations where it is deemed to bring additional insights and/or in cases where the task is considered too daunting or time consuming for manual processes. The quantity of data will effectively provide a degree of triangulation. All findings from these methods will be manually checked by the team of experts and referenced in the assessment reports.
Using AI tools that meet enterprise-grade security standards, the evaluation team—comprising humanitarian experts, data scientists, and MEL specialists—will train the models to generate focused insights aligned to the evaluation criteria and questions. The core team will quality assure AI outputs, refining prompts (and, if needed, underlying models) and validating findings from the AI-generated evidence synthesis. We can also package structured prompts and LLM models into customised, reusable AI agents—enabling clients to run similar analyses independently in the future.
Matt McConnachie
Expert Profile: As a Principal Consultant with the MEL team, Matt is the driving force behind numerous external and internal initiatives that incorporate cutting-edge data science and AI-driven tools into the MEL workflow. His core interest in new technologies is framed by his extensive experience in complex evaluations, adaptive management, learning and uptake.
Matt McConnachie
Expert Profile: As a Principal Consultant with the MEL team, Matt is the driving force behind numerous external and internal initiatives that incorporate cutting-edge data science and AI-driven tools into the MEL workflow. His core interest in new technologies is framed by his extensive experience in complex evaluations, adaptive management, learning and uptake.