Research Campaign Setup
Creating a New Study
Overview
Yazi supports three core study types — AI Interviews, Surveys, and Diary Studies — each delivered entirely within WhatsApp. You can run any of these as a standalone single-day study, or combine them into a Multi-Day Sequential Study that links multiple activities together over time.
This page covers how to choose the right study type, the universal design principles that apply across all methodologies, and the shared builder features you'll use regardless of which study type you select.
Choosing Your Study Type
AI Interview
A structured survey section followed by an AI-moderated qualitative interview. The AI generates follow-up questions in real time based on what each participant has already said, creating a conversational experience that blends quantitative data collection with qualitative depth.
Best for: Exploring motivations, attitudes, and experiences with qualitative depth at scale
How questions work: You author the survey questions; the AI generates interview questions in real time based on participant responses
Typical sample size: 12–100 participants
Typical question count: 8–12 survey questions + 5–15 AI-generated follow-ups
Participant effort: 8–20 minutes
What you get: Coded data from the survey section + rich verbatims from the AI interview
Flagship capability: The AI adapts its questions based on everything the participant has already said
Survey
A traditional questionnaire with structured questions, delivered as a WhatsApp conversation. Each response is captured individually as the participant answers, so even partial completions yield usable data.
Best for: Quantitative measurement, tracking, benchmarking, and classification
How questions work: All questions authored by you in advance
Typical sample size: 250+ participants
Typical question count: 22–28 questions
Participant effort: 3–8 minutes
What you get: Coded, countable data
Flagship capability: Highest scale and speed
Diary Study
A recurring set of questions participants answer repeatedly over time. Participants return to the same WhatsApp conversation to log entries whenever a relevant event occurs, building a timestamped picture of habits, routines, and experiences over days, weeks, or months.
Best for: Longitudinal behaviour tracking, experience logging, and pattern detection
How questions work: All questions authored by you in advance; repeated each session
Typical sample size: 20–200 participants
Typical question count: 3–6 questions per entry, repeated across multiple sessions
Participant effort: 2–5 minutes per entry, multiple entries over days/weeks/months
What you get: Timestamped session data showing patterns over time
Flagship capability: Captures in-the-moment behaviour as it happens
Quick Decision Guide
"I need numbers and percentages" → Survey
"I need to understand why" → AI Interview
"I need to see patterns over time" → Diary Study
"I need numbers first, then understanding" → AI Interview (the survey section gives you the numbers, the AI section gives you the depth)
"I need evolving research across multiple days" → Multi-Day Sequential Study (combine any of the above)
Single-Day vs. Multi-Day
When creating a new study, you first choose between:
Single Campaign (Single Day) — one standalone activity. Choose a Survey, AI Interview, or Diary Study.
Multi-Day Sequential Study — multiple activities linked together over time. Each activity can use a different methodology. For example: Survey on Day 1 → Diary Study on Days 2–5 → AI Interview on Day 6. See the Multi-Day Studies page for full details.
Two Setup Methods
When you click New Research and select your study structure, you're presented with two setup methods:
Upload Brief with AI
Upload a research brief in Word or Google Doc format, and Yazi's AI will automatically generate your study — questions, answer options, routing logic, and interview configuration. The AI produces a complete draft in approximately 5–6 minutes, which you then edit and refine in the builder.
This is the fastest way to get started. Even a rough brief will produce a strong starting draft. Write out your research objectives, target audience, key topics, and any specific questions you need asked — the more detail you provide, the better the output. Depending on the complexity of your study, this can save hours of manual setup.
Manual Setup
Build your study from scratch in the builder. You choose your study type and add questions one by one.
Both methods produce the same output — a fully configurable study in the builder. You can always start with AI-generated questions and manually adjust from there.
Rendering Modes
All study types can be displayed in one of two rendering modes within WhatsApp:
In-Chat
Questions and answer options appear as native WhatsApp messages directly in the conversation thread — one message per question, with buttons or lists for answer options. This is the most familiar experience for participants and feels like a natural WhatsApp conversation.
Flows
Questions render as a dynamic modal overlay on top of the WhatsApp chat — similar to a browser window within WhatsApp. Participants can move through questions more quickly, making this particularly suited for longer quantitative surveys.
Comparison:
Look and feel: In-Chat uses native WhatsApp messages and buttons. Flows use a pop-up window within WhatsApp.
Speed: In-Chat presents one question at a time at a conversational pace. Flows allow faster progression through questions.
Device compatibility: In-Chat works on all devices and WhatsApp versions. Flows require a newer version of WhatsApp.
Best for: In-Chat is ideal for maximum compatibility, qualitative studies, and emerging markets with older devices. Flows work best for quantitative surveys where speed matters and participants have up-to-date devices.
You can switch between In-Chat and Flows at any time in the builder without rebuilding your study.
Recommendation: Default to In-Chat for maximum reach and compatibility. Use Flows when you're running a long quantitative survey with a participant base you know has up-to-date devices.
Universal Question Types
All three study types support the same set of question types:
Single Select — One answer from a set of options. Renders as buttons (≤3 options) or a scrollable list (4+ options).
Multi-Select — Multiple answers from a set of options. Renders as a list with checkboxes.
Open Text — Free-text response. The participant types their reply.
Rating Scale — Numeric scale (e.g., 1–10, NPS). Renders as buttons or list depending on the range.
Voice Note — Audio response recorded using WhatsApp's native voice note feature.
Image Upload — Photo capture or gallery upload via native WhatsApp image sharing.
Video Upload — Video capture or gallery upload via native WhatsApp video sharing.
Location — GPS pin drop using WhatsApp's native location sharing.
File/Document — PDF or document upload via native WhatsApp file sharing.
What You Can't Ask
WhatsApp's API and app interface do not support certain question formats. These are platform-level constraints that apply regardless of study type:
No grid/matrix questions — The WhatsApp interface cannot render a grid of rows and columns. Break matrix questions into individual rating or select questions instead.
No ranking/sorting questions — Participants cannot drag-and-drop or reorder items on a touchscreen within WhatsApp. Use sequential "most important" / "least important" questions as an alternative.
Universal Builder Features
These features are available across all study types:
Text formatting — Bold, italics, and quotes for emphasis in question text.
Live preview — See exactly how each question renders in WhatsApp before publishing.
Validation warnings — Answer options that exceed WhatsApp's character limits are flagged in red.
Option randomisation — Shuffle answer option order per participant to reduce first-option bias.
"Other" option — Add a free-text "Other" option to any select question. When selected, the participant is prompted to type a response, capturing answers you may not have anticipated.
Routing logic — Skip logic, branching, screen-outs, and looping based on participant responses.
Intro message — Customise the message participants see before the first question.
Closing message — Customise the message shown after the final question.
Drag and reorder — Rearrange questions freely without affecting existing response data.
How Many Questions Should You Ask?
This is the most frequently asked question in research design, and the answer depends on your study type, your participants, and your incentive structure.
General Guidelines
22–28 questions is the sweet spot for a standard quantitative survey with remunerated participants
8–12 survey questions + 5–15 AI follow-ups for AI interviews
3–6 questions per entry for diary studies, where participants log repeatedly
Beyond 32 questions, expect significantly higher drop-off regardless of incentive
WhatsApp vs. Browser-Based Surveys
A common question: "Can I ask as many questions on WhatsApp as I would in a browser-based survey?"
The short answer is yes. The speed at which participants complete questions in WhatsApp is roughly the same as in a browser. The WhatsApp format is not a further limitation on survey length beyond what you'd expect from any mobile survey.
What is different:
Participants get distracted — WhatsApp is a messaging app, and notifications from other chats may interrupt the flow.
Progress is saved — Unlike browser surveys, participants can leave and come back without losing their answers.
Reminders bring them back — You can send automated reminders to participants who started but didn't finish, prompting them to complete within your stipulated timeframe.
This means your effective completion rate can actually be higher than browser-based surveys, because you have tools to re-engage distracted participants rather than losing them permanently.
Every Question Counts Individually
Unlike browser-based surveys where you typically only get data when someone completes the entire questionnaire, Yazi captures each response as it's submitted. If a participant answers 15 out of 28 questions and drops off, you still have 15 usable data points.
Design principle: Place your most important questions first. Treat every subsequent question as a bonus. Even with drop-off, you'll have data on what matters most.
Answer Option Display Rules
How your answer options render in WhatsApp depends on the number and length of options:
3 or fewer short options → Tappable buttons directly in the chat
4 or more options → Scrollable list accessed via a burger menu button
Option text exceeds ~24 characters → Automatically switches from button to list format
The builder flags options that exceed the character limit in red so you can adjust before publishing.
Tip: Keep option text concise. If you need to provide context or definitions, put them in the question text rather than the answer options.
Multi-Language Support
All study types support multi-language surveys with the same configuration:
Write all questions and options in your primary language (typically English).
Yazi auto-translates using Google Translate into your target languages.
Manually edit any translation for accuracy before publishing.
Participants select their preferred language at the start of the study.
All questions render in their chosen language.
Responses are captured in the original language and translated back to your primary language.
For multi-day studies, language settings are configured once at the study level and apply across all activities — participants don't need to re-select their language for each stage.
Publishing & Sharing
Publishing
A study must be published to WhatsApp before it can be accessed by participants. While in draft, you can build, edit, and preview — but the study link will not work and broadcast invitations cannot be sent until you publish.
Two Ways Participants Can Access Your Study
Link sharing — Share a link via email, social media, QR code, or any other channel. Participants click the link, WhatsApp opens with a passcode, and they tap send to begin.
Broadcast invitations — Send a WhatsApp message directly to a list of participants. They receive the message in their WhatsApp chat and tap a button to begin.
Live Link vs. Test Link
Every study generates two links:
Live link — For real participants. Responses are recorded permanently.
Test link — For previewing the experience yourself. Resets AI interview memory so you can test the full conversation repeatedly.
Always use the test link when testing AI interviews. The AI remembers previous conversations on the live link — if you've already reached the maximum question count, it won't ask further questions on subsequent attempts.
Sandbox vs. Live
You can build and test your entire study in the sandbox before going live:
All work in the sandbox carries over when you publish with your dedicated WhatsApp number — nothing is lost.
Test the full participant experience using Yazi's test number.
Sandbox data exports are limited to 10 rows (full export available on live account).
Fraud Prevention
WhatsApp-based studies carry inherent fraud protections that apply across all study types:
Phone number authentication — Each participant is tied to a verified phone number, making it difficult to create duplicate respondents.
Device fingerprinting — WhatsApp is tied to a single device, preventing bot farms from scaling easily.
Voice note verification — Requiring a voice note response confirms a real human participant.
Engagement patterns — Response timing and interaction patterns are visible, making rushed or automated completions easy to identify.
Last updated