I've Seen This Movie Before
I've been in marketing research for more than 20 years. I know the grind firsthand—the trials of getting sample, creating questionnaires, managing projects, tracking quotas, building crosstabs, arguing with the stats. It's a long, arduous process, and honestly, it hasn't gotten any easier.
In 1999, I launched SurveyWriter—probably one of the first web-based survey software applications. I didn't make a single dollar for the first two years. Not because the software was bad (it was—and is—very good), but because the research industry is conservative and doesn't adapt to change quickly.
I heard the same arguments then about web surveys that I hear today about synthetic data: “There's no way the results will be accurate. Telephone interviewing cannot be replaced.”
We've heard some version of that for every leap—phone vs. in-person in the '60s and '70s, online vs. phone in the 2000s. Every time, the industry has had reasons to doubt the next thing. And every time, the next thing won.
The Pattern Is Always the Same
Remember when taxis dismissed Uber? When hotels waved off Airbnb? When retailers ignored Amazon and studios scoffed at Netflix?
The pattern is always the same: an incumbent industry with entrenched processes dismisses a new technology because it doesn't perfectly replicate the old way. Meanwhile, the new technology gets better fast, and the cost and speed advantages become impossible to ignore.
That's exactly where synthetic survey data is right now.
What We Actually Built
No, you can't just open ChatGPT, ask for a purchase-intent survey, and generate 1,000 stratified completes you'd stake a brand on. But it is true that carefully trained machine-learning systems can model human response patterns in reliable, useful ways.
That's what we built at Simsurveys. Over 24 months, we developed multiple domain-specific AI models—not one model that claims to do everything. We have specialized models for Healthcare, Consumer research, and Social and Political studies, each trained on millions of validated responses in their respective domains.
In Healthcare: We augment our models with a curated, physician-level database of every single practicing physician in the U.S., linked to prescription behavior patterns. That's 15 specialties with responses grounded in real clinical practice.
Three Ways to Use It
Not every researcher needs the same thing, so we built three modes of operation:
- Synthetic data: Generate complete survey datasets from scratch. Write your survey, set your demographics, get completes in minutes.
- Augmented data: Take an existing set of live-respondent data and augment the original answers with responses to new questions you didn't ask.
- Expanded data: Have 200 completed surveys but need 800 more by tomorrow? We expand your dataset using the original live surveys to condition the generation of additional records.
We wrapped these models in a full questionnaire-design and project-management application, applying AI not just to data generation but to the entire research workflow. Our technology is protected by U.S. Patent Application No. 18/784,418 and multiple provisional patents—this isn't just another survey tool with “AI” slapped on top.
Let's Be Honest About What This Is
Is this a drop-in replacement for a bespoke, in-person study with a pristine, hand-curated panel? Probably not—and let's be honest, most research isn't that gold-plated anyway.
In our validation studies, Simsurveys typically delivers KL divergence scores of 0.05–0.09 against live panel benchmarks. That means results that are better than sloppy web intercepts or low-quality panels, and often within striking distance of high-quality panel outcomes.
Here's what matters: you can write your survey, program it, set quotas and targets, and generate completes in minutes. Then guide the app to build your crosstabs, draft an executive summary, and assemble a complete report—charts included. All at about one-tenth the traditional cost.
This Isn't a Loss Leader
We're not giving this away, and we're not loss-leading for market entry. Cloud economics, smart automation, and a team of brilliant data scientists make it possible to be fast, good, and cheap.
I'm not here to change the world, but I don't mind pushing the marketing research industry to make life easier for researchers, product managers, and decision-makers. We're choosing to lead, not follow—to take smart risks so researchers can do more real thinking and less busywork.
If this pushes the industry to raise its game, great.
My challenge to you: Kick the tires. Hold us to a high bar. If Simsurveys saves you weeks of time and frees your budget for bigger studies, additional research, or better margins while getting you answers you can use—that's a win for the work and the industry.