Time commitment:~3–4 days per week, variable by phase. Duration: period tbc. Start-asap
OCH's global research team conducts large-scale, multi-country survey research and has developed a growing library of quantitative datasets and segmentation outputs across geographies. We are looking for an experienced quantitative analyst to join the team and contribute across a range of analytical work — from foundational data preparation and exploration through to advanced statistical modelling.
The core analytical focus of the role centres on two interconnected workstreams: the rigorous development of survey-based clustering and segmentation models, and the design of a classification framework that allows new respondents to be assigned to existing segments efficiently and reliably. Beyond this, the analyst will also handle day-to-day data management tasks including dataset cleaning, variable harmonisation, and exploratory cross-tabulation work. The role sits within the research methods function and involves close collaboration with OCH's Head of Data & Research Methods.
KEY RESPONSIBILITIES
Data cleaning & preparation
Clean, recode, and structure incoming survey datasets - including applying advanced data quality checks and filters, raking & weighing, missing data, etc.
Conduct foundational data exploration including frequency distributions, cross-tabulations, and basic descriptive analyses, primarily in SPSS
Work fluently across survey data formats, principally SPSS (.sav) and R-native formats
Cluster analysis & segmentation
Conduct advanced cluster analysis on complex, multi-country survey datasets, working hand in hand with the Head of Data & Research Methods regarding analytical decisions and final segmentation outputs
Evaluate and compare clustering approaches (e.g. k-means, hierarchical, latent class analysis, and others as appropriate) with a view to producing segments that are statistically robust, meaningful, and cross-nationally comparable
Manage the specific methodological challenges of complex survey data: dealing with varying variable types (nominal, ordinal, continuous), handling of translated or culturally non-equivalent items
Iteratively test and refine cluster solutions, systematically varying parameters and documenting the impact of each decision on outputs
Classification model development
Using existing, labelled segmentation outputs as a training base, design and fit (machine learning / train-test) an appropriate classification model to enable assignment of new respondents to established segments
Evaluate candidate classification approaches (e.g. random forest, logistic regression, LDA, gradient boosting, or others) and select the most appropriate given the data structure, segment separability, and intended use
Assess model performance rigorously using appropriate validation strategies (e.g. cross-validation, held-out test sets, confusion matrices, precision/recall)
Iterate on model specifications, documenting all variations and intermediary outputs
'Golden questions' identification
Identify the minimum set of survey questions ('golden questions') that are most predictive of segment membership — i.e. those that would need to be included in future quantitative research instruments to allow reliable classification of new respondents
Apply appropriate variable importance and feature selection techniques to identify and rank candidate questions, and validate their predictive power
Produce clear recommendations on the golden question set, including supporting evidence and sensitivity analyses
Classification / calculator tool
Design and implement a practical classification tool or calculator that can be applied to future survey datasets to assign respondents to segments based on the golden question set
Ensure the tool is well-documented, reproducible, and usable by the Head of Data & Research Methods without requiring re-running of the full modelling pipeline
Methodological documentation
Maintain detailed records of all analytical iterations, including variations in parameters, model specifications, and the rationale behind decisions taken
Document all intermediary outputs in a structured and retrievable format
Produce final methodological documents for each workstream — written to a standard that would allow a qualified analyst to understand, reproduce, and build upon the work
Flag methodological uncertainties or trade-offs explicitly, rather than presenting a single opaque output
REQUIRED EXPERTISE & EXPERIENCE
Solid, demonstrable experience (typically 4–7 years) working with quantitative survey or polling data (or equivalent) in an analytical capacity
Fluency with SPSS for data cleaning, cross-tabulation, and exploratory data analysis, including confident management of variable and value labels, codebooks, and data transformations
Advanced proficiency in cluster analysis methods, with hands-on experience selecting and comparing approaches on real survey datasets
Proven experience fitting and validating classification models using labelled training data
Advanced R proficiency — all modelling and classification work is expected to be conducted in R, with clean, documented, reproducible scripts
A rigorous, structured approach to analytical work with a strong documentation habit
KEY SKILLS & ATTRIBUTES
Statistically rigorous and methodologically confident, with the seniority to take end-to-end ownership of complex analytical problems
Detail-oriented and systematic, with a natural inclination to document decisions and iterations thoroughly
Comfortable working autonomously and at depth on a focused analytical brief
Able to communicate methodological choices clearly in writing, for a technically informed audience
Self-directed, structured, and reliable in managing their own workflow
Originally posted on