AI-native survey
programming in minutes.
Upload your survey spec or write it with prompts. Review and refine with built-in AI assist. Ship in hours, high accuracy, without the grind.
- qid: Intro type: info title: Welcome text: | Thank you for participating in this research about streaming video services. The survey takes about 12 minutes.- qid: S1 type: single text: In which country do you currently live? instruction: <i>(Single response)</i> required: True order: fixed skin: radio options: 1: United States 2: United Kingdom 3: Canada 4: Australia# Punch hMarket immediately after S1 — carries the# country label for piping and quota.- script: | if S1.value is not None: hMarket.set(S1.value)- qid: hMarket type: single visibility: hide title: Market (from S1) — used for pipingUpload a spec.
Get a working survey.
Drop the Word doc or paste the questionnaire. Fieldbase reads the question text, options, programming notes, and routing instructions — and writes the survey code for you to review.
- S1single
- {}script
- hMarketsingle
- S2single
- {}script
- S3single
- Q1multi
- Q2multi
- Q3gridSingle
- Q4cbc
- qid: S1 type: single text: In which country do you currently live? required: True options: 1: United States 2: United Kingdom 3: Canada 4: Australia
Describe the change. The AI writes it.
You accept, refine, or skip.
Scope with @, describe what you want. The AI proposes a diff — lint-checked and in your DSL — with Accept, Refine, or Reject buttons. Every change is logged in a thread the whole team can trace.
How AI assist works →- qid: S1 type: single text: In which country do you currently live? required: True options: 1: United States 2: United Kingdom 3: Canada 4: Australia# Market quota check — fires on S1 selection- script: | if hMarket.value is not None: cell = hMarket.value if not quota("qtMarket", cell).isOpen: endSurvey("quotafull")- qid: S2 type: number text: What is your age? min: 13 max: 99 required: True# Age gate — out-of-scope respondents terminated- script: | if S2.value >= 65: if S2.value >= 65: if S2.value < 18 or S2.value >= 65: endSurvey("terminate")- qid: Q1 type: multi text: Which of these brands have you used in the past 3 months? min: 2 options: brandList- qid: Q2 type: gridSingle text: How satisfied are you with each of these features? rows: brandListSee your results instantly.
Export clean data.
Auto-generated topline reports with cross-tabs and filters. Deterministic column naming across every export — no cleanup, no surprises between waves.
| Total | Dell | Apple | HP | Lenovo | |
|---|---|---|---|---|---|
| Quality | 3.8 | 3.4 | 4.5 | 3.2 | 3.6 |
| Value | 3.2 | 3.6 | 2.1 | 3.8 | 3.7 |
| Ease of use | 3.9 | 3.5 | 4.6 | 3.3 | 3.4 |
| Availability | 3.7 | 4.2 | 3.8 | 3.9 | 2.8 |
Built for how modern
survey work actually happens.
Each capability is designed around the way survey programmers actually work today — fast, code-first, AI-augmented, with the research methods and output quality serious projects need.
Code-first authoring
Clean, readable syntax for every question type and every routing rule. Autocomplete and live validation from a single contract — every rule defined once, enforced everywhere.
single · multi · grid · cbc · maxdiff · cardSort Scripting & routing →Conjoint and MaxDiff included
CBC with three response modes — choice, rating, ranking. Upload a design file, reference it in a single line. Utility-ready exports drop straight into your analysis tool.
design: project://laptop_cbc_v2 Choice modeling & CBC →Quotas that hold
Cell locking on every completion. No over-quota, no soft counts, no manual reconciliation — even under heavy concurrent traffic. The numbers you set are the numbers you get.
quota("AgeGender", 1).isFull() Quota engine →Live respondent preview
Every change updates the preview as you type. Branching, loops, piped values, hidden logic — see exactly what a respondent sees, without deploying. Test routing in seconds.
Survey simulator →Analysis-ready exports
Deterministic column naming — every export column is mechanically derived from the question structure. No surprises, no manual mapping, no cleanup before analysis.
QBudgetr1c1 · QBudgetr1c2 · QBudgetr2c1 Export formats & datamap →Theme framework
22 professional themes with seven style variants. Every visual property is a token — change how a survey looks without touching what it does. Respondent-facing UI that holds up to client review.
Theme framework →Every capability, in depth.
The platform's real surface — question types, scripting, design assets, simulation — each with its own page you can read, share, or send to a stakeholder.
15+ types — single, multi, grids, cardSort, CBC, MaxDiff, lookup, rank. 60 skin variants for different respondent contexts.
Explore question types →
Python-like - script blocks. Conditional routing,
quota gates, computed hidden values, screenouts — real logic.
Upload CBC / MaxDiff design files, reference in one line, and verify precision before fielding with Design Health + Precision scorecards.
Explore choice modeling →Fake respondents against the real quota engine. Catch routing gaps, load-test quotas, verify CBC and MaxDiff coverage.
Explore the simulator →Prompt. Author.
In your language.
Prompt the AI in the language your team thinks in — French, Japanese, Spanish, German, Portuguese. The AI responds in-language; the generated content stays in that language too. Twelve languages supported today, more as teams ask.
qid, type, options) and
identifiers stay English, so any reviewer on your team can read
the YAML regardless of the content language.
Built for the way research data has to be handled.
Your data isn't used to train AI
Survey content, respondent answers, and the code you author are never added to AI training data. Your work — and your clients' data — stays your own.
ISO 27001 certified, GDPR aligned
Certified under ISO 27001 and aligned with GDPR through our parent company, Inginit. Data is encrypted in transit and at rest, with tenant isolation by project. Full details on the Inginit trust center.
You own and control your data
Export the full project — code, design files, responses — at any time. Delete on request, including from backups within standard retention windows. Any operational metrics we keep for product improvement are anonymised aggregates that can't be traced back to you.
Fieldbase is built by Inginit. For certifications, sub-processors, retention windows, and to request a DPA, see the Inginit trust center.
Get in early.
Fieldbase is in private beta. Email us with a few words about your team and the kind of research you run — we'll set up a call and get you onto the platform.
Email [email protected]Replies within 48 hours · No mailing list, just a real conversation.