CAWI Programming
From a Word questionnaire to an online survey ready to field.
Logitude is a research firm running CATI, CAWI and CAPI surveys. We programmed a CAWI questionnaire from its Word specification: screens, rotations, routing, input controls and a standalone HTML export, with a testable prototype before going live.
Turning a Word questionnaire
into a reliable survey.
A Word specification describes questions, filters, rotations and sometimes ambiguous instructions. Programming it by hand is slow, and every routing error biases the survey.
The aim was to speed up go-live without sacrificing fidelity to the original questionnaire.
A programming tool
built for CAWI.
The questionnaire is imported then refined in a screen editor, previewed as a respondent would see it, shared with the team and exported as a standalone HTML file.
The .docx text is converted into a structured survey by a single structured-output LLM call (Claude or GPT, user’s choice). It produces options, rotations, exclusive options, display conditions and terminations. It is a first draft: ambiguous instructions are refined in the editor.
Screen types: chapter, single choice, multiple choice, matrix, ranking, 0-10 scale and numeric. Rotation engine (pinned or exclusive options), routing conditions with operators and interview termination, live preview as a respondent.
Respondent flow served from tracked links with session resume, quota groups and segments, response collection, analysis tables (sex, age, socio-professional category…) and CSV export.
Roles (admin, client, panelist), member invitations, a discussion thread per survey with email notification, and export of the survey as one HTML file that runs with a double-click.
From specification
to survey, faster.
Web stack
and structured AI
Next.js and Supabase for the tool, Claude or GPT for the import, and an in-house rotation and routing engine to guarantee identical behaviour in the tool and the HTML export.