How to finish your thesis data collection in 48 hours (instead of 3 weeks)

Why most master's theses stall at the same place

Ask any grad student where their thesis got stuck and most will give you the same answer: data collection.

Not the research question. Not the lit review. Not writing up. The part where you have to go out and get real responses from real humans who don't know you, don't owe you anything, and have better things to do on a Tuesday night than fill in your Google Form.

You post it in three WhatsApp groups. Your friends fill it in. Your friends' friends don't. You hit 11 responses and need 50. You ask your supervisor for an extension. They give you the look. You start considering a career in anything else.

I get DMs about this every week. This guide is the setup I give them.

What's actually possible in 2026

The old options for a master's student doing primary research were:

1. Convenience sampling — post your survey in personal networks and pray. Slow, biased, usually not defensible in an exam committee.
2. Institutional panel — your university's psychology research pool or equivalent. Limited to undergrads, narrow demographics, and usually locked to psych department theses.
3. Commercial panel providers (Prolific, CloudResearch, Qualtrics) — decent quality, but typically R1,000 — R10,000+ for the volumes a thesis needs, and you have to manage the study yourself.
4. Research agencies (Nielsen, Ipsos, Kantar) — R400,000+ and a 6-week timeline. Not happening for a master's budget.

What's new in 2026 is a fifth option: Cookiy, which lets you generate a screener survey directly inside Claude Code (or Cursor, or Claude Desktop), target verified respondents through a global panel it's connected to, and pull results back into your terminal. For a typical thesis sample of 50 adults, you're looking at under R600 all in.

Disclosure up front: I'm partnered with Cookiy. I only recommend tools I'd use, and the guide below is honest about what it can and can't do.

When this setup works (and when it doesn't)

Before you spend a rand, read this carefully. If your thesis falls in the left column, the setup below will save you weeks. If it falls in the right, you'll need a different pathway.

Use Cookiy for	Don't use Cookiy for
Adult consumer behaviour	Research involving children under 18
Attitudes, opinions, preferences	Clinically-diagnosed patients where diagnosis is the inclusion criterion
Self-reported habits (sleep, exercise, spending, social media)	Medical vitals, lab measurements, or drug-trial data
Product and concept testing	Purposive samples needing rare ethnic or socioeconomic quotas
Willingness-to-pay / pricing research	Qualitative research where YOU must moderate for methodological reasons
Job market, workplace, employment research	Studies where your ethics committee has stipulated a specific recruiter
Mental health, wellbeing, stress (self-report, not diagnosis)	Forensic or legal populations

Rule of thumb: If your research question starts with "how do adults who X think/feel/behave about Y," Cookiy probably works. If it starts with "I need patients diagnosed with X by a clinician," you need a medical recruiter and your institution's research ethics committee (REC).

The DM that kicked this guide off — master's student researching hypertension patients — sits on the edge. The honest answer: she can use Cookiy for "adults who self-report as managing high blood pressure" and study their adherence behaviour, pharmacy experience, or knowledge gaps. She cannot use Cookiy to study clinically-confirmed stage-2 hypertensive patients at a specific blood pressure threshold. Same topic. Different research question. One works here, one needs a hospital recruitment agreement.

The 5-step setup

Budget roughly one hour for the full setup. After the first run, future studies take 10 minutes.

1. Install Claude Code

Claude Code is Anthropic's agentic coding tool. Free to download, works in your terminal. If you've never touched a command line, this is a good first reason to.

• Download: claude.ai/code
• Sign in with your Anthropic account (free to create, same one you use for Claude.ai)
• Open your terminal and run claude — you'll see a chat prompt

If you're on Windows, install WSL first (there's a one-click Microsoft installer). If you're on Mac or Linux, it just works.

Tip: Your .ac.za email doesn't unlock anything Claude-side, but if you're a student you qualify for Cursor Pro free for a year — Cursor has the same agentic features and supports the same Cookiy skill.

2. Install the Cookiy skill

The Cookiy skill is an open-source plugin that teaches Claude Code how to run the full research workflow (create study → preview guide → run AI simulations → recruit real participants → export report).

• GitHub: github.com/cookiy-ai/user-research-skill
• In Claude Code, paste the repo URL and say: "Install this skill."
• Claude Code clones the repo, registers the skill, and confirms it's active.

You'll also need the Cookiy MCP server, which handles the actual backend (study creation, recruitment, billing). The skill's install instructions walk you through it — takes about two minutes.

3. Get your free credits

Cookiy gives new users $30 in free credits when you sign up with a legitimate email. That's enough for:

• ~60 screener survey responses, or
• 3 full 15-minute AI-moderated voice interviews, or
• A mix — generate a discussion guide + run 2 interviews + 20 survey responses

Apply your credit: Sign up at cookiy.ai with the email you'll use in Claude Code. Credits apply automatically.

If you're a student: comment THESIS on the ATB post that ships this guide and I'll send you a top-up code good for an additional $30 in credits. Between the two, you have enough to finish a typical master's thesis sample.

4. Write your research prompt

Inside Claude Code, with the Cookiy skill installed, you describe your study in plain English. Cookiy converts it into a proper screener + discussion guide.

Template prompt for a quantitative survey (50 respondents):

Run a Cookiy screener survey.

Topic: [your research topic in one sentence]
Target respondents: [adult demographic description — age range, country,
  any self-report qualifiers like "self-report as managing high blood pressure"]
Sample size: 50
Screener questions: [list 2-3 yes/no filters that qualify respondents]
Main questions: [list 5-7 research questions, each measurable]

Output format: anonymous responses, CSV export, sentiment summary.

Example (the hypertension thesis from the opening):

Run a Cookiy screener survey.

Topic: Medication adherence habits of adults managing high blood pressure
Target respondents: Adults 35-70 who self-report as currently taking
  prescription medication for high blood pressure. South Africa or
  global English-speaking.
Sample size: 50
Screener questions:
  - Are you currently taking prescription medication for high blood pressure?
  - Have you been on this medication for at least 3 months?
Main questions:
  - How often do you miss a dose in a typical week? (never / 1-2 / 3-5 / daily)
  - What is the main reason you miss doses?
  - Have you ever stopped taking your medication without telling your doctor?
  - Where do you get your prescriptions filled?
  - On a scale of 1-10, how confident are you that you're taking your
    medication correctly?

Output format: anonymous responses, CSV export, sentiment summary.

Cookiy will generate the questionnaire, show you the recruit preview, and tell you the total cost. Review it, tweak any question that's leading or double-barreled, then approve.

5. Run it and export results

Once you approve, Cookiy launches the study. For a 50-person screener, expect results within 24 to 48 hours. Some complete inside 6 hours depending on the targeting.

When the study finishes, Cookiy drops a report with:

• CSV of anonymous responses (ready for SPSS, R, Excel)
• Summary statistics per question
• Sentiment analysis on open-text responses
• Breakdown by demographic variables

Export the CSV, drop it into your analysis tool, and start writing your results chapter.

Mistakes that will burn your credits

I've seen three mistakes wipe out someone's free $30 in an hour. Avoid them:

1. Vague screener questions. "Are you interested in AI?" lets anyone through. "Have you personally used ChatGPT, Claude, or Gemini in the last 30 days?" filters for your actual population. Tight screeners save you money.
2. Leading research questions. "Don't you think AI tools are making students lazy?" is a bad question. "What effect, if any, has using AI tools had on your study habits?" is a good one. Your examiners will catch leading questions — get rid of them before the data is collected, not after.
3. Too many questions. Above 10 substantive questions you start losing completion rate. Respondents drop off, and you pay for attempts that didn't finish. Keep it under 10 questions plus screeners.

What to do if your ethics committee asks for recruitment details

Your REC (or IRB, depending on country) will want to know how respondents were recruited, how consent was obtained, what data was collected, and where it's stored. Cookiy handles most of this automatically:

• Recruitment: Connected panels (Cint, Prodege, Lucid and others) with pre-consented respondents
• Consent: Respondents opt in per study before questions are shown
• Anonymisation: Responses delivered without PII by default
• Data storage: Hosted, encrypted in transit and at rest

For your methodology chapter, include a line like: "Respondents were recruited via Cookiy, an AI-moderated research platform connected to global consented respondent panels. All participants opted in per study and provided informed consent. Responses were delivered anonymously."

If your REC needs a data processing agreement or GDPR documentation, Cookiy's compliance team will provide it — email compliance@cookiy.ai.

Qualitative option: the AI-moderated interview

Surveys give you breadth. Interviews give you depth. If your thesis needs qualitative data — pilot interviews, deep-dive case studies, or open exploratory research — Cookiy also does 15-minute AI-moderated video interviews at $9.90 per participant.

The AI moderator follows your discussion guide, asks follow-up questions, reads facial expressions and tone, and shares the full recording + transcript + sentiment analysis with you afterwards. You watch live from the observation room.

For a typical qualitative thesis chapter needing 8-12 interviews, you're looking at under R1,600 total. Compare that to an SA market research agency quote of R80,000-R120,000 for the same work.

Timeline to compare

Stage	Traditional pathway	Cookiy pathway
Draft survey	3 days	20 minutes (Cookiy generates, you tweak)
REC approval	2-6 weeks	2-6 weeks (no change — still needed)
Recruit 50 adults via WhatsApp / friends	3-6 weeks	24-48 hours
Clean + analyse data	1 week	1 week (no change)
Total active research time	8-14 weeks	4-6 weeks

The REC approval step still takes as long as it takes — ethics committees don't go faster because your tool is smarter. But the recruitment step goes from the slowest part of your thesis to an overnight task, which usually collapses 4-8 weeks out of your timeline.

Summary

• Data collection is where most master's theses stall
• Cookiy + Claude Code turns a 3-week recruitment grind into a 24-48 hour pull, for under R600 per study
• It works for adult consumer behaviour, attitudes, self-report, willingness-to-pay, and product testing studies
• It doesn't replace your REC, a medical recruiter for clinical populations, or your moderator skills for depth-interview research where methodology demands you moderate
• Student free credit: $30 standard, ping me for an extra $30 code

Next steps

• Install Claude Code and the Cookiy skill before your next thesis meeting
• Before you run anything paid, draft your research prompt and run it as an AI-simulated interview first — that's free and lets you tighten the questions before spending on real respondents
• If you want a second pair of eyes on your questionnaire before you launch, DM me — I'll review it for leading/double-barreled questions
• Related: The best AI tool for every task — what to use for lit review, writing, and citation chasing

This guide powers the "Comment THESIS" CTA.