How we design and conduct research
The design choices, quality controls and ethical process behind every study we run. Written for the people who will review our methods — and for anyone who wants to know what they are commissioning.
Designing the study
Every engagement begins with a scoping conversation, not a proposal. We establish what decision the research will inform, what is already known, what specifically needs to be found out, what would count as a credible answer, and what constraints — time, access, budget, ethics — shape what is possible.
Only then do we recommend a design. Sometimes that design is smaller than what was requested, because a smaller well-designed study answers the question and a larger poorly-designed one does not. We would rather say that at the start.
“A question we can only half-answer produces a study nobody should rely on.”
One of the four tests we apply to every research question we take on.
Study designs we use
QUANTITATIVE
Cross-sectional surveys · cohort studies · case-control studies · quasi-experimental designs (difference-in-differences, propensity score matching, interrupted time series) · randomised designs where appropriate and ethical
QUALITATIVE
In-depth interviews · key informant interviews · focus group discussions · observation · case studies · document and policy analysis
MIXED-METHOD
Sequential explanatory, sequential exploratory and convergent designs, with an explicit statement of how the strands integrate — specified in the protocol, not improvised at analysis
EVALUATION
Baseline · midline · endline · process · outcome · impact · formative · summative · developmental, with theory of change and results framework development where these do not already exist
REVIEW
Systematic reviews · scoping reviews · rapid evidence assessments · meta-analysis, following PRISMA
HEALTH ECONOMICS
Cost analysis · cost-effectiveness and cost-utility analysis · burden-of-disease estimation · out-of-pocket expenditure and financial-protection studies
Sampling and sample size
Sampling frames are constructed and documented, not assumed. We specify the sampling method, justify the sample size against the primary outcome and the precision or power required, state the expected non-response and how it is handled, and document any deviation between the planned and achieved sample in the final report.
Where a sample cannot support a claim, we say so rather than making the claim.
Instrument development
Instruments are developed against the research questions, drawing on validated tools where they exist. Where a tool is adapted or translated, we document the process: forward and back translation, expert review, cognitive testing and piloting.
Piloting is not optional. A questionnaire that has not been tested in the field is a hypothesis about a questionnaire.
Data collection and quality
We do not maintain a permanent in-house field team. Data collection runs through vetted field partners with nationwide coverage in Bangladesh, and we supervise fieldwork directly rather than subcontracting responsibility for it.
Enumerators are trained for each study on the instrument, on consent procedures and on ethical conduct. Fieldwork is supervised, with back-checks and spot re-interviews agreed in the protocol. We use KoboToolbox and ODK for computer-assisted data collection, with built-in validation, skip logic and GPS and timestamp capture.
Quality checks run during collection, not after: completion rates, outlier detection, interviewer-level consistency, duplicate detection, GPS and timing checks. Cleaning decisions are logged in a documented script so every change is reproducible and reversible. Personal identifiers are separated from analysis datasets.
Analysis plans are written before the data is seen
Quantitative analysis in Stata, R and SPSS, following a pre-specified analysis plan. Qualitative analysis using a documented coding framework, with inter-coder agreement where the design requires it. Where we deviate from the plan, we report the deviation and why.
QUANTITATIVE
Descriptive and inferential statistics · regression and multivariable modelling · survival analysis · multilevel models · sensitivity analysis
QUALITATIVE
Thematic and framework analysis · documented codebook · inter-coder agreement · negative-case analysis · audit trail
HEALTH ECONOMICS
Cost and cost-effectiveness analysis · financial protection and catastrophic expenditure · equity analysis
Ethical review
Every study involving human participants is submitted for ethical review before data collection begins. We do not begin fieldwork on an unapproved protocol, whatever the timeline pressure.
The Ideology does not yet operate its own institutional review board. Protocols are submitted to the relevant government, institutional or organizational ethics committee — in practice, the ethics committee of the partner institution, the commissioning body, or the Bangladesh Medical Research Council where the study falls within its remit. Establishing our own ethics review committee is a stated objective.
We say this plainly rather than implying a capability we do not yet have. Read the full ethics policy →
Consent and participant protection
Informed consent is obtained in the participant's language, in writing or by documented verbal consent where literacy or context requires. Participants are told what the study is for, who is funding it, what will happen to their data, that participation is voluntary, and that they may stop at any point without consequence. Consent forms and information sheets are submitted as part of ethical review.
Reporting
Every study produces a technical report stating methods, results, limitations and interpretation, with the instrument and analysis approach annexed. Most also produce a decision-maker brief, a plain-language summary, a presentation, or a visual product — agreed at design stage, not at the end.
We report what we found, including findings that are inconvenient. A report shaped to a preferred conclusion is not research.
| Stage | Checkpoint |
|---|---|
| Design | Internal methodological review before the protocol is finalised |
| Ethics | Approval before any data collection |
| Instrument | Pilot and revision before full fieldwork |
| Fieldwork | Supervision, back-checks, daily data review |
| Data | Documented cleaning; reproducible scripts |
| Analysis | Pre-specified plan; internal review of outputs |
| Reporting | Internal review before client delivery |
Discuss your study design
Send us the question you need answered. We will tell you what design would answer it, what it would require, and whether the timeline you have in mind is realistic.