Source: Dalberg Data Insights, Technical Data Fellowship Case Study Assessment, 2024/25
Tags: data science case study, SDG 4, education data analysis, Nairobi Kenya, primary school infrastructure, Dalberg Data Insights, development data, geospatial analysis, census data
Difficulty: Intermediate
Prerequisites: Familiarity with Python, R or JavaScript for data analysis. Basic understanding of descriptive statistics, geospatial data concepts and data visualisation. Some exposure to international development frameworks (SDGs) is helpful but not required.
This case study sits at the intersection of data science and international development. Dalberg Data Insights (DDI) is a data and technology consultancy that works with governments, NGOs and development organisations to turn raw data into actionable policy recommendations. The exercise tests whether a candidate can take messy, real-world education and demographic datasets, identify where gaps in access to quality schooling are worst, and translate findings into a six-slide presentation that a non-technical government official can act on. It maps directly to Sustainable Development Goal 4 (quality education) and mirrors the kind of analytical work development data scientists do day-to-day: combining facility-level data with census demographics, building composite indicators, and making resource-allocation recommendations under constraints.
You are a data scientist at DDI, tasked with analysing Nairobi's primary school infrastructure and demographic data to advise a county education officer on where to direct a new grant. The deliverable is a maximum six-slide presentation (exported as PDF) supported by code, with clear visualisations and stated assumptions. The core skill being tested is turning two datasets into prioritised, evidence-based investment recommendations aligned with SDG 4.
Dalberg Data Insights (DDI)
A data and technology consultancy within the Dalberg group that partners with governments and development organisations. Think of it as the analytics arm that turns raw data into policy-ready recommendations.
SDG 4 (Sustainable Development Goal 4)
The United Nations goal to "ensure inclusive and equitable quality education and promote lifelong learning opportunities for all." In simple terms, this is the global benchmark for whether children everywhere can access decent schooling.
Education Facilities Data
A dataset containing information on primary schools across Nairobi, including geo-locations, infrastructure capacity and other indicators. Originates from 2013 but is treated as representative for this exercise. Think of it as a census of school buildings and what they have.
Kenya Population and Housing Census
Demographic data for Nairobi covering population counts, age distributions and household characteristics. In simple terms, this tells you how many people (and children) live where.
Geospatial analysis
Analysis that uses geographic coordinates (latitude/longitude) to map facilities, measure distances and identify spatial patterns such as clustering or gaps in coverage. Think of it as "data science with a map."
Infrastructure capacity
The physical resources a school has: classrooms, desks, latrines, water points. This is what determines how many pupils a school can realistically serve at an acceptable standard.
Pupil-to-classroom ratio
The number of enrolled students divided by the number of usable classrooms. A key proxy for overcrowding. High ratios signal that existing infrastructure is under strain.
Composite indicator / index
A single score built by combining several variables (e.g. infrastructure quality, access distance, overcrowding) into one number per area or school. In simple terms, it collapses multiple dimensions of "need" into a ranking you can act on.
Accessibility analysis
Measuring how easily a population can reach a service, typically by distance or travel time. For schools, this means asking: how far does a child have to walk to the nearest primary school?
Actionable recommendation
A finding that specifies what to do, where to do it and why, backed by evidence. The opposite of a vague conclusion like "more investment is needed."
The Nairobi County Education Officer has secured an external grant to improve school infrastructure and resource allocation across the county
The officer needs data-driven insights to decide where the money should go, because spreading it evenly would waste impact
The overarching goal is equitable access to primary education for all pupils in Nairobi, aligned with SDG 4
DDI's role is to analyse existing datasets, find the biggest gaps in education infrastructure, and propose where investment will have the most effect
Access: Can children physically reach a school within a reasonable distance?
Capacity: Once there, is the school overcrowded? Are there enough classrooms, desks, latrines?
Equity: Are some sub-counties or wards systematically worse off than others? Do demographic patterns (population density, age distribution) align with school provision?
The case study is specifically about primary schools, not secondary or tertiary education
You are playing the role of a data scientist at DDI, not an academic researcher
The audience is a government official who is not technical but needs to make investment decisions
Your output must be actionable: "invest in Ward X because it has the highest pupil-to-classroom ratio and the lowest school density relative to its under-14 population"
Vague conclusions like "infrastructure needs improvement" will score poorly
Dataset 1: Education Facilities Data
Comprehensive information on primary schools across Nairobi
Includes geo-locations (latitude/longitude), infrastructure capacity indicators and school characteristics
Originates from 2013 but the exercise treats it as still representative
Typical fields you would expect: school name, location coordinates, number of classrooms, pupil enrolment, type of school (public/private), available facilities (water, latrines, electricity)
Dataset 2: Kenya Population and Housing Census
Detailed demographic information for Nairobi at the sub-county and ward level
Population counts, age breakdowns, household characteristics
Allows you to estimate the school-age population (typically ages 6 to 13 for primary) per geographic area
How the two datasets connect
The education facilities data tells you what exists (supply). The census data tells you who needs it (demand). The analytical core of the exercise is joining these two: comparing where schools are and what capacity they have against where children live and how many there are. Geographic identifiers (sub-county, ward or coordinates) are the join key.
Suggested analytical approaches
Compute school density per ward (number of schools per 1,000 school-age children)
Calculate pupil-to-classroom and pupil-to-teacher ratios where the data permits
Map school locations against population density to identify underserved areas
Build a composite "need index" combining overcrowding, access distance and demographic pressure
The exercise explicitly allows additional assumptions and external datasets to enrich the analysis
Presentation (the primary output)
Maximum six slides, excluding cover page
Exported and uploaded as a PDF
Must stand alone without a verbal presentation: the slides should tell the full story
No dashboards (Tableau, PowerBI); this is a static presentation
Should include clear, compelling visualisations derived from the data
All assumptions must be explicitly stated
A strong six-slide structure typically looks like this:
Slide 1: Problem framing and hypothesis (where should investment be directed and why?)
Slide 2: Data overview and methodology (what data was used, how it was processed, key assumptions)
Slide 3: Supply-side analysis (school distribution, infrastructure quality, capacity gaps)
Slide 4: Demand-side analysis (population demographics, school-age children per area)
Slide 5: Gap analysis (where supply falls short of demand, composite need index, map)
Slide 6: Prioritised recommendations (specific areas, specific interventions, expected impact)
Code (the secondary output)
All analysis code submitted as a single file (.py, .ipynb, .R, .js) or a zipped folder
No datasets included in the upload
Code is assessed on readability, modularity, consistency and documentation
Comments and docstrings matter; a reviewer will read the code to understand your analytical choices
What gets marked
The assessment criteria map directly to what a client engagement at DDI would require:
Data understanding and processing accuracy
Code quality (readable, modular, documented)
Depth of analysis in mapping facilities, evaluating accessibility and assessing demographic disparities
Thoughtfulness in identifying vulnerabilities and targeting recommendations
SDG alignment and practical implementability
Creativity in metric development
Clearly documented assumptions
Professional, non-technical-audience-friendly presentation
"More schools = better access." Not necessarily. If new schools are built in areas already well-served while high-density informal settlements remain underserved, access does not improve where it matters most. Distribution matters more than count.
"The 2013 dataset is too old to use." The exercise explicitly states to treat it as representative. Questioning the data's age is fine as a stated assumption, but refusing to analyse it or caveating every finding with "but this is old data" misses the point.
"I should build a dashboard." The brief specifically says not to. The deliverable is a static PDF presentation. Building a Tableau or PowerBI dashboard will not score well, regardless of how polished it looks.
"Technical sophistication in the code is what matters most." The assessment criteria weight communication, actionability and thoughtfulness just as heavily as technical execution. A beautifully engineered pipeline that produces vague recommendations will underperform a simpler analysis with sharp, location-specific findings.
The exercise tests consulting-style data science, not academic research. Your slides are the product, not a paper.
Visualisations must be clear to a non-technical audience. Label axes, use plain titles, avoid jargon in chart annotations.
Stating assumptions explicitly is a scored criterion. If you assume the under-14 population proxies school-age children, say so.
SDG 4 alignment is specifically called out. Frame at least some recommendations in terms of SDG 4 targets (e.g. 4.a: build and upgrade education facilities that are child, disability and gender sensitive).
Creativity in metric development is valued. Going beyond obvious ratios (e.g. building a vulnerability index, estimating walking distance to nearest school, identifying gender-specific gaps in latrine provision) will differentiate your work.
The presentation must be self-contained. A reviewer will read it cold, with no voiceover. Every slide needs enough context to stand on its own.
True or False: The deliverable for this case study is a Tableau dashboard. (False: it is a maximum six-slide PDF presentation.)
Fill in the blank: The two datasets provided are the Education Facilities Data and the Kenya ______ and Housing Census. (Population)
True or False: The Education Facilities Data originates from 2023. (False: it originates from 2013.)
Fill in the blank: SDG 4 aims to ensure inclusive and equitable ______ education. (quality)
True or False: You are expected to present the slides in a recorded video. (False: the presentation must stand alone without you presenting it.)
Q: What is the primary objective of the Nairobi County Education Officer in this case study?
A: To ensure equitable access to quality primary education for all pupils in Nairobi, using a recently secured grant to address the most critical gaps in school infrastructure, guided by data-driven recommendations aligned with SDG 4.
Q: How would you identify the most underserved areas in Nairobi for primary education?
A: Join the education facilities data with the census data at the ward or sub-county level. Compute the ratio of school-age children to available school capacity (classrooms, enrolled places). Areas with the highest ratios, the longest average distances to the nearest school, or the poorest infrastructure quality scores are the most underserved.
Q: Why is it important to state assumptions explicitly in this exercise?
A: Because the datasets are imperfect (the facilities data is from 2013, the census may not perfectly match school catchment areas), and analytical choices (e.g. defining "school-age" as 6 to 13, using straight-line distance rather than walking distance) affect the results. Stating assumptions lets the reviewer assess whether your logic is sound and lets the client decide whether to accept them.
Q: What makes a recommendation "actionable" versus "vague" in a development data context?
A: An actionable recommendation names a specific location, a specific intervention and a reason grounded in data. For example: "Prioritise classroom construction in Kibera ward, where the pupil-to-classroom ratio is 120:1, three times the county average." A vague recommendation says: "More classrooms are needed across Nairobi."
Q: How does SDG 4 relate to this case study?
A: SDG 4 is the global framework for quality education. Target 4.a specifically calls for building and upgrading education facilities. This case study asks you to identify where those investments are most needed in Nairobi, making your recommendations a direct operationalisation of SDG 4 at the county level.
This case study connects to geospatial data science (working with coordinates, computing distances, creating choropleth maps), a skill set used across public health, urban planning and logistics. It also connects to survey and census data analysis, which is foundational in development economics and demography. The framing as a consulting deliverable for a government client connects to data communication and storytelling, the discipline of turning analytical findings into slides that drive decisions. Finally, the SDG alignment connects to the broader field of monitoring and evaluation (M&E) in international development, where indicators are tracked against global frameworks.
Dalberg Data Insights, DDI Technical Data Fellowship, case study assessment, data science for development, SDG 4, quality education, Nairobi Kenya education, primary school infrastructure, education facilities dataset, Kenya Population and Housing Census, geospatial analysis, school mapping, pupil-to-classroom ratio, pupil-teacher ratio, accessibility analysis, composite index, vulnerability index, education equity, sub-county analysis, ward-level data, choropleth map, development consulting, data-driven policy, resource allocation, grant prioritisation, M&E, monitoring and evaluation, school-age population, infrastructure capacity, data visualisation for policy