Watch a great adviser in a first meeting and you'll notice something odd: they barely follow their own agenda. A client mentions, in passing, that their daughter has moved back home — and instead of proceeding to superannuation balances, the adviser puts the pen down and asks about that. Twenty minutes later it turns out the daughter's situation is the actual reason the couple is thinking about their finances at all.
The fact-find form would have captured the daughter — a name and date of birth in the dependants section. It would never have surfaced why she mattered.
I've spent most of my career building software for financial advisers, and for most of that time I treated discovery — the fact-find — as a data-collection problem. Get the fields filled. It took me embarrassingly long to understand that the industry's discovery problem was never about fields. It's that we replaced a conversation with a questionnaire, and those are not the same instrument. There is now a substantial body of research, spanning social psychology, survey methodology, and human-computer interaction, that explains precisely why — and it's worth taking seriously, because it points at what good digital discovery has to look like.
Follow-up questions are not politeness. They're the mechanism.
In 2017, researchers at Harvard published a study in the Journal of Personality and Social Psychology with a finding that sounds obvious until you see how specific it is. Across three studies of live conversations — including more than 2,000 speed-dating interactions — people who asked more questions were consistently better liked by their conversation partners. But the effect wasn't driven by questions in general. It was driven overwhelmingly by follow-up questions: questions that could only exist because the asker had genuinely processed the previous answer (Huang, Yeomans, Brooks, Minson & Gino, 2017).
The researchers traced the mechanism to what psychologists call responsiveness — the perception that the other party is listening, understanding, and cares. A follow-up question is hard evidence of responsiveness: you cannot ask "how did your daughter ending up back home change your thinking?" unless you actually heard the thing about the daughter. The same research found the inverse, too: self-focused conversation — redirecting to yourself, holding the floor — reliably decreased liking. And the effect has limits: a barrage of questions without give-and-take reads as interrogation, not interest.
Every experienced adviser knows this in their bones. What the research adds is that responsiveness isn't a soft nicety layered on top of discovery — it is the thing that makes people open up. Which is exactly the thing a static form cannot do. A PDF asks the same forty questions in the same order regardless of anything you write in box seven. It is, by construction, incapable of a follow-up question. It is architecturally unresponsive.
People answer standardised questions wrong — specifically when it matters most
The second pillar comes from survey methodology, and it should worry anyone who relies on self-completed fact-finds.
For decades, survey researchers assumed standardisation was the gold standard: ask every respondent exactly the same question, exactly the same way, and the answers will be comparable. Michael Schober and Frederick Conrad tested that assumption in a landmark 1997 study in Public Opinion Quarterly, and found a serious crack in it. When people's circumstances were straightforward, standardised questions worked fine. But when circumstances were atypical — when the honest answer was "well, it depends what you mean" — accuracy collapsed. Their canonical example is almost comically mundane: does buying a lamp count as a furniture purchase? Respondents guessed, and often guessed wrong. Allowing the interviewer to converse — to clarify what the question actually meant for this person's situation — substantially improved response accuracy, a result replicated across a series of laboratory and field studies over the following two decades (Schober & Conrad, 1997; Conrad & Schober, 2000; West, Conrad, Kreuter & Mittereder, 2018).
Now map that onto a financial fact-find. "Do you have life insurance?" I think so — something through work? Through super? Does that count? "What is your risk tolerance?" Compared to what? "List your assets." Does my share of Mum's house count? The loan to my brother? Nearly every meaningful question on a fact-find is a lamp-and-furniture question — the honest answer depends on clarification the form cannot provide. The research finding, translated: the clients whose situations most need to be understood accurately are precisely the ones a standardised form understands worst. Complexity is where forms fail, and complexity is where advice matters.
There's a compounding effect, well documented under what survey researchers call satisficing (Krosnick, 1991): faced with a long self-administered questionnaire and no one watching, people don't optimise their answers — they do the minimum required to be done with it. Straight-lining. "Not sure." "Retire comfortably." Anyone who has received a half-completed fact-find back from a client has seen satisficing in the wild.
The surprising part: conversation survives the machine
Here is where the research gets genuinely useful for anyone building digital discovery, because the obvious objection is that everything above describes humans. Great human questioners are responsive; human interviewers can clarify. Surely none of it transfers to software.
Except it does, and this has been measured. In a 2019 study presented at CHI — the leading human-computer interaction conference — Kim, Lee and Gweon compared identical questionnaires delivered as a conventional web form versus a conversational, turn-by-turn chat interface. The conversational version produced measurably higher-quality data: more differentiated answers, and far less of the minimum-effort answering described above. Strikingly, tone mattered too — a casual conversational style reduced it even further, but only in the chat format (Kim, Lee & Gweon, 2019). Related work has found the same direction of effect repeatedly: conversational interfaces with genuine dialogue capability elicit more accurate answers, more detail, and less survey fatigue than static forms (Xiao, Zhou, Liao, Mark & Chi, 2020; Conrad et al., cited in Kim et al., 2019).
Read the three research streams together and the conclusion is hard to avoid. The value of conversation in discovery was never mystical. It decomposes into specific, identifiable mechanisms — responsiveness signalled through follow-up questions, clarification matched to atypical circumstances, engagement sustained through interaction — and those mechanisms are properties of the format, not of human biology. A conversation that asks its next question because of your last answer, explains why it's asking, and clarifies what "insurance" means for your actual situation is doing the things the research says produce honest, accurate, complete answers. A form does none of them, no matter how well designed.
The honest boundary
None of this says software replicates a great adviser. The 2017 Harvard research is ultimately about human connection — responsiveness that leads to trust, the kind that lets someone say "we want to retire at 60" and be heard to mean "I'm frightened of my job." A machine can implement follow-up questions; it does not occupy the position of the person who has sat with you through a redundancy and a market crash. I've written elsewhere that nobody should try to automate that, and I hold to it.
But the comparison that matters for most Australians was never machine-versus-adviser. Roughly one in ten of us gets professional advice. For everyone else, the realistic alternative to a well-designed conversation is not two hours with a skilled human. It's the form — the PDF, the forty web fields, the instrument that fifty years of research tells us people misunderstand when their lives are complicated and abandon when nobody's watching.
Against the gold standard, conversational discovery is a humble second. Against the actual alternative, it isn't close.
The research has been sitting there for years, in journals advisers never read, quietly explaining why the fact-find form was always the weakest link in the advice chain — and exactly what has to replace it.
What we're building at Vecta
I'll be direct about the connection, because this research is not a curiosity to us — it's the design brief.
Vecta's discovery layer is built as a conversation, and the three mechanisms above are its three governing principles.
Responsiveness by construction. The next question exists because of the last answer. When a member mentions the daughter who moved back home, that's not noise to be filtered out on the way to the superannuation fields — it's the thread the conversation follows, the way a good adviser would follow it. The Harvard finding says follow-up questions are how a questioner proves they're listening; we've made them the basic unit of how discovery works, not an edge case.
Clarification where circumstances are atypical. The lamp-and-furniture problem is, we think, the central failure of the self-completed fact-find — so the conversation is designed to do what Schober and Conrad's interviewers did: explain what a question means for this person's actual situation, and why it's being asked at all. "Insurance through work" isn't a wrong answer to be rejected by a validation rule. It's the beginning of an exchange.
Engagement while attention exists. The satisficing research says people give you their real answers only while they're genuinely engaged — which is why the conversation shows a member the gap between what they want and what their numbers support while they're still paying attention, not in a document three weeks later.
And then the boundary, which matters as much as the capability: everything the conversation gathers flows to a licensed adviser, who reviews and approves any advice before a member ever sees it. The discovery conversation doesn't replace the human relationship — it means that when a member does reach a human, they arrive understood rather than as a half-completed PDF. And for the millions who were never going to reach that room at all, it means the instrument asking about their lives is finally one the evidence says they'll answer honestly.
The fact-find form had a fifty-year run. The research explaining why it never worked has been published for most of that time. We're building what the research says should have existed all along.
References
- Huang, K., Yeomans, M., Brooks, A. W., Minson, J., & Gino, F. (2017). It doesn't hurt to ask: Question-asking increases liking. Journal of Personality and Social Psychology, 113(3), 430–452. doi:10.1037/pspi0000097
- Schober, M. F., & Conrad, F. G. (1997). Does conversational interviewing reduce survey measurement error? Public Opinion Quarterly, 61, 576–602.
- Conrad, F. G., & Schober, M. F. (2000). Clarifying question meaning in a household telephone survey. Public Opinion Quarterly, 64, 1–28.
- West, B. T., Conrad, F. G., Kreuter, F., & Mittereder, F. (2018). Can conversational interviewing improve survey response quality without increasing interviewer effects? Journal of the Royal Statistical Society, Series A, 181(1), 181–203.
- Krosnick, J. A. (1991). Response strategies for coping with the cognitive demands of attitude measures in surveys. Applied Cognitive Psychology, 5, 213–236.
- Kim, S., Lee, J., & Gweon, G. (2019). Comparing data from chatbot and web surveys: Effects of platform and conversational style on survey response quality. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (CHI '19).
- Xiao, Z., Zhou, M. X., Liao, Q. V., Mark, G., & Chi, C. (2020). Tell me about yourself: Using an AI-powered chatbot to conduct conversational surveys. ACM Transactions on Computer-Human Interaction.