𝐒𝐲𝐧𝐭𝐡𝐞𝐭𝐢𝐜 𝐫𝐞𝐬𝐩𝐨𝐧𝐝𝐞𝐧𝐭𝐬: 𝐀 𝐖𝐚𝐜𝐤𝐲 𝐖𝐚𝐯𝐢𝐧𝐠 𝐌𝐚𝐧 𝐨𝐟 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡.

TL;DR: Synthetic respondents have a place in research, but only for a few select purposes and as an addition to research on real people, not a replacement.

Let’s be honest: in an ideal world, we’d always rely on real human respondents. They bring context, emotions, and real-life messiness. But reality isn’t always ideal. Clients don’t have time, they don’t have budgets, and sometimes, running a study just isn’t an option. Plus, let’s face it: some questions just aren’t worth troubling real people for. Does a human panel really need to validate that “fast delivery” matters in e-commerce? 𝑫𝒐 𝒚𝒐𝒖 𝒓𝒆𝒂𝒍𝒍𝒚 𝒘𝒂𝒏𝒕 𝒕𝒐 𝒔𝒑𝒆𝒏𝒅 𝒎𝒐𝒏𝒆𝒚 𝒕𝒐 𝒂𝒔𝒌 𝒑𝒆𝒐𝒑𝒍𝒆 𝒘𝒉𝒆𝒕𝒉𝒆𝒓 𝒕𝒉𝒆𝒚 𝒍𝒊𝒌𝒆 𝒕𝒉𝒂𝒕 𝒃𝒖𝒕𝒕𝒐𝒏 𝒐𝒏 𝒕𝒉𝒆 𝒓𝒊𝒈𝒉𝒕 𝒐𝒓 𝒐𝒏 𝒕𝒉𝒆 𝒍𝒆𝒇𝒕?

As a researcher, I would never ask a synthetic database for something that I need to hear from customers. What is my job if not connecting businesses to real people and letting their voices be heard? But I have seen some clients in the wild who I think are a bit too much to expose their research requests to actual human respondents. 😱😱😱 And so did you, dear research colleague. 𝒀𝒐𝒖 𝒘𝒂𝒏𝒕 𝒕𝒐 𝒕𝒆𝒔𝒕 𝑯𝑶𝑾 𝑴𝑨𝑵𝒀 𝒄𝒐𝒍𝒐𝒖𝒓𝒔?

For me, this is a use case for synthetic respondents. If someone needs fast, cheap input, serving as a proxy for generic, aggregated opinions – we can ask the robot to stand for humans, much like we place a Wacky Waving Man in front of a shop. Wacky Waving Men have a resemblance to a person and some of their behaviours looks human – with a large generalization. And they do a good job of looking agitated and attracting attention instead of a real human.

So can synthetic respondents, too. They can help in early-stage idea validation, small hypothesis testing, or stress-testing assumptions before investing in full-scale research. Early-stage customer development, narrowing down many options to a few prioritized ones, endless iterations of changes – this is their use case. To be frank, I think their main user base will be those who do not research customers at the moment. Maybe the idea that feedback can be good can land in some diehard engineer mind better, if the feedback is from the bot? 🙄

BUT – and this is a big BUT – I strongly believe that synthetic respondents should not be used when real customer feedback is a must. They cannot give you personal experiences, emotions, and unpredictable human reactions that no AI-generated response can match. 𝐒𝐨𝐫𝐫𝐲, 𝐚 𝐖𝐚𝐜𝐤𝐲 𝐖𝐚𝐯𝐢𝐧𝐠 𝐌𝐚𝐧 𝐢𝐬𝐧’𝐭 𝒓𝒆𝒂𝒍𝒍𝒚 𝐡𝐚𝐩𝐩𝐲 𝐭𝐨 𝐬𝐞𝐞 𝐲𝐨𝐮.