Research software for teams that care about rigor.
GlassFisch is being built by experienced research and field professionals to make sophisticated research operations more connected, accessible and dependable.
Built from field experience.
Designed for better research.
Experience behind the platform
GlassFisch is being built by people who have designed survey instruments, run field operations and prepared data for analysis, including in international development research.
Team experienceGlassFisch is shaped by more than 25 years of combined professional experience in survey research, impact evaluation, questionnaire development, large-scale field operations and data-quality assurance.
Across their professional careers, members of the team have contributed to research assignments involving institutions such as the World Bank, KfW, GIZ, United Nations agencies, the University of Mannheim in Germany and the University of Washington in the United States.
Their hands-on experience spans established research technologies, including SurveyCTO, ODK, KoboToolbox, Survey Solutions, CSPro and Stata. This practical background informs GlassFisch's approach to questionnaire authoring, statistical quality checks and research operations.
These references describe the previous professional experience and technical background of individuals contributing to GlassFisch. They do not imply institutional affiliation, sponsorship, endorsement, partnership or software integration. Product and organization names belong to their respective owners.
Why we started building GlassFisch
Research projects often depend on several disconnected tools. Questionnaire authoring, translation, field preparation, data collection, quality checks, corrections and analysis may each be handled in a separate system, with files, scripts and spreadsheets carrying the work from one stage to the next.
Every one of those transitions has a cost.
- Duplicated effortThe same setup rebuilt in several tools.
- Technical bottlenecksRoutine changes waiting for one specialist.
- Coordination challengesTeams working from different versions.
- Delayed quality checksProblems found after teams have left the field.
- Difficult translation workflowsLanguage versions drifting from the source.
- Additional operational riskMore hand-offs, and more places for errors to enter.
GlassFisch grew out of years of solving these problems in real research projects, one study at a time, with whatever tools were at hand. The platform is our attempt to bring that experience into one connected place. Not every connection described on this page is available yet; the current development stage is set out below.
We believe researchers should spend more time producing reliable evidence and less time connecting fragmented tools.
From research design to analysis-ready evidence
Eight connected stages of research operations. Each one shapes the quality of what reaches the analysis.
Research Design
Research questions, sampling and the analysis plan define what the data must support.
Questionnaire Authoring
Questions, routing, constraints and calculations turn the design into a working instrument.
Translation & Review
Each language version is translated, reviewed and checked against the source questionnaire.
Field Preparation
Teams, assignments, devices and interviewer training are ready before the first interview.
Data Collection
Interviews are conducted with routing and checks applied as answers are entered.
Quality Review
High-frequency checks and back-checks raise signals for human review while teams are still in the field.
Corrections & Documentation
Confirmed issues are corrected with a recorded reason, and the original values stay traceable.
Analysis Readiness
A consistent, documented dataset is ready for the analysis plan.
Reliable evidence depends on the quality of each step, not only the final dataset.
An illustrative chain of research operations, in more detail than the research lifecycle on the homepage. It describes stages of research work, not the release status of GlassFisch modules.
From fragmented workflows to connected research operations
Research teams often move their work between tools built for different stages. GlassFisch is being designed so that each stage builds on the one before.
Traditional research workflow
- Multiple disconnected systems
- Repeated configuration
- Translation coordination challenges
- Heavy dependence on specialist coding
- Delayed quality review
- Fragmented operational oversight
GlassFisch design direction
- More connected research workflows
- Structured questionnaire authoring
- More manageable multilingual preparation
- Accessible quality-check configuration
- Clearer organizational oversight
- Better continuity across research stages
The right-hand column describes the direction GlassFisch is being designed toward, not capabilities that are all available today. It compares ways of working, not specific products.
Designed around the realities of rigorous field research
Research methodsThe methods below are established research practice, described in plain terms. They shape what GlassFisch is being built to support; naming them is not a claim that each is available in GlassFisch.
Instrument design
- Computer-assisted personal interviewing (CAPI)
- Interviewers administer the questionnaire on a tablet or phone, so routing, checks and timestamps travel with every interview.
- Skip patterns and conditional logic
- Questions appear only when they apply, so respondents are asked what is relevant and nothing is skipped by mistake.
- Range and consistency constraints
- Answers are checked against plausible ranges and earlier responses while the interview is still under way.
- Questionnaire calculations
- Derived values, such as household totals or ages from dates, are computed during the interview and can steer later questions.
- Randomization
- Treatment modules, question order or response options are assigned at random, for experimental designs or to balance order effects.
Quality during fieldwork
- Missing-data classifications
- Don't know, refused and not applicable are recorded as distinct codes, so missingness can be examined rather than guessed at.
- High-frequency checks (HFCs)
- Routine checks run as data arrive, so problems surface within days rather than after fieldwork closes.
- Paradata-aware review
- Interview metadata, such as duration and timing, helps reviewers decide where to look more closely.
- Back-checks and supervision
- A sample of interviews is revisited or re-asked to confirm that data were collected as intended.
From data to evidence
- Data validation
- Rules confirm that the dataset is complete, consistent and structured as the questionnaire intended.
- Reproducible research workflows
- Every step from raw submissions to the analysis file is documented, so others can trace and repeat it.
- Analysis readiness
- A clean, documented dataset with its codebook and recorded decisions, ready for the analysis plan.
How this page distinguishes its statements
Advanced questionnaire design, made more approachable
Planned product directionFrageDock — Questionnaire Builder is the questionnaire-authoring component of GlassFisch Collect, where a research design becomes a working instrument.
Researchers and project staff should be able to build sophisticated questionnaires without every routine change requiring specialist programming support. Specialist programmers remain essential for the most demanding instruments; the aim is to let them focus there, while routine changes stay with the research team.
- Conditional logicShow, hide and route questions according to earlier answers.
- Complex calculationsDerive values during the interview and reuse them in later questions.
- Validation constraintsPlausible ranges and consistency rules, with messages interviewers understand.
- Structured questionnaire designSections, rosters and repeated groups in a clear structure.
- RandomizationSupport for experimental and order-balanced designs.
- Multilingual preparationLanguage versions prepared and reviewed alongside the source questionnaire.
- Clearer authoring workflowsChanges reviewed, tested and versioned before they reach the field.
FrageDock is part of GlassFisch Collect, not a separate product. The list describes the intended authoring experience; FrageDock is not yet publicly available.
Making rigorous data-quality review more accessible
Planned product directionMembers of the team have spent years configuring and running Stata-based high-frequency checks during large-scale field data collection. Run throughout fieldwork, such checks can help identify:
- Inconsistencies between related answers
- Unexpected or implausible values
- Patterns of missing data
- Interview-quality concerns
- Operational issues that need investigation
A flag is a signal, not a verdict. An HFC flag is a signal for review, not conclusive proof of an error or interviewer misconduct. People decide what it means.
Writing and maintaining those checks has usually required Stata expertise. Field Intelligence is being designed to make quality-check configuration more accessible to authorized research professionals, including those who are not Stata experts.
Intended outcomes
- Earlier attentionProblems noticed while teams are still in the field and can still resolve them.
- Transparent reasonsEach flag shows the rule behind it, so reviewers can judge it.
- Recorded decisionsA traceable path from signal to review decision and any correction.
- Shared ownershipQuality review that does not depend on one specialist's scripts.
Research trust is built into the process.
Rigorous research depends on more than good data. These four principles guide how GlassFisch is being designed and built.
- Pillar 1
Quality data
Evidence is only as reliable as the data behind it. GlassFisch is designed to support sound instruments, timely review and a documented account of each data decision.
- Sound questionnaire logic
- Consistency checks
- Better field review
- Documented data decisions
- Pillar 2
Quality software
Research teams need tools that behave predictably and stay understandable. We aim for clear workflows, careful engineering practice and fewer technical bottlenecks.
- Clear research workflows
- Reliable design practices
- Reduced technical bottlenecks
- Practical usability
- Pillar 3
Data privacy
Respondents entrust research teams with personal information. GlassFisch is designed to keep it within appropriate organizational boundaries and to show no more of it than a task requires.
- Responsible data handling
- Controlled access
- Sensitivity to respondent information
- Appropriate organizational boundaries
- Pillar 4
Data security
Access should follow role and purpose. GlassFisch is designed around permission-aware access, restrained data exposure and safeguards that refuse access when a condition cannot be confirmed.
- Permission-aware design
- Restrained data exposure
- Fail-closed safeguards
- Audit-conscious operations
The pillars describe design principles. They are not certifications, audits or guarantees, and they do not imply that any system is free of risk.
Research data should not become AI training material.
Our data-handling direction is to keep customer research data outside AI model training. We are designing GlassFisch around responsible access, organizational control and research confidentiality.
- Quality checks in GlassFisch are deterministic, configurable rules, not AI judgments, so the reason for each flag can be shown.
- Automated signals support human review; they do not replace it.
- Where AI assistance is considered for future features, our direction is to work from aggregated, minimized results rather than respondent-level records, and only with the explicit approval of the organization concerned.
This is a statement of direction. It is not a description of verified technical controls or a contractual commitment. Binding data-handling terms will be set out in the Legal & Trust documents, which are in preparation.
Built for the people behind the evidence
GlassFisch is being designed for the people who plan, run and review research in the field.
Universities and research centers
Research groups running surveys alongside teaching and publication.
Principal investigators
Oversight of instruments, fieldwork and data quality across a study.
Impact evaluation teams
Experimental and quasi-experimental studies where data integrity carries the design.
Research and consulting organizations
Teams running several studies, often for different clients, at the same time.
Data collection firms
Field operations at scale, across many teams, forms and languages.
Monitoring and evaluation teams
Recurring data collection that follows results over time.
Field managers and supervisors
Day-to-day coordination of teams, assignments and follow-up.
Research professionals without specialist programming backgrounds
People who know what an instrument must do and want to build it themselves.
Building carefully. Verifying thoroughly.
Current platform capabilities- Building and verifyingCurrent stage
- Planned controlled pilotInvitation-based, when enrollment opens
- Public plansTo be announced
GlassFisch is being prepared for a planned controlled pilot, with invitation-based participation when enrollment opens. Enrollment has not opened.
GlassFisch Collect for Android is in development. Offline mobile fieldwork will become available following completion of integration and release verification.
Other feature availability varies by module; the homepage workflow shows the current status of each. We would rather take longer than offer researchers something they cannot rely on, so our approach is to verify each capability before it is offered.
Built from field experience. Designed for rigorous research.
Our ambition is to help research teams bring together sound methods, practical operations and responsible technology, so that more time can be spent producing reliable evidence.
The GlassFisch Team
Explore our Knowledge Hub