A1. Asking Maven the Right Question
There's a common assumption that the hard part of working with a system like Maven is the system's intelligence. It usually isn't. The hard part is the question. A lazy prompt gets a mediocre answer out of even a genuinely capable system, and a sharp prompt makes that same system look brilliant, not because it got smarter in the meantime, but because the thinking was finally pointed somewhere specific.
Maven is a multi-agent research system, which means it's deciding what kind of work your question actually calls for before it does anything else. If a question asks for magnitude, it should route toward Quantitative Research. If it asks for meaning, emotion, or motivation, it should route toward Qualitative Research. If it asks for both, it should return both. That routing is only as good as the question makes it, which is why the real skill here isn't operating Maven, it's learning to write a question that signals what kind of answer you actually need.
Start by Naming What You're Actually Asking For
Before typing anything, ask yourself one thing: do you want the size of something, the reason behind it, or both? That single distinction changes everything downstream.
If you want to know how many people prefer something, how strongly, how factors rank against each other, or what share of a population behaves a certain way, you're asking for Quantitative Research. Words like measure, rank, compare, percentage, incidence, and likelihood signal that clearly. If you want to know why people feel a certain way, what makes them hesitate or trust, or how they'd describe an experience in their own words, you're asking for Qualitative Research. Words like why, what makes, how is it perceived, and what concerns signal that instead. And if you want both the size of a factor and the reasoning behind it, both signals need to appear in the question itself, not one implied and the other assumed. Maven is intelligent, not telepathic, and a prompt that sounds emotionally rich but never asks for a ranking won't produce one.
A Quant Question Needs an Implied Measurement
A strong Quantitative prompt doesn't need to read like a statistics textbook, but it does need to make clear that numbers, comparisons, or distribution are the actual output you want. "Tell me about brand preference for Walmart" has the shape of a question without the substance of one. "Measure brand preference for Walmart as a supermarket compared with competitors across grocery shoppers in the United States" gives Maven a defined construct, a comparison frame, a population, and an implied output, everything it needs to head down the Quant path with confidence.
A Qual Question Needs Room for the Mess
Qualitative Research is about meaning before measurement, so a good Qual prompt makes space for contradiction, hesitation, and interpretation rather than asking what happened in the flattest possible terms. "What do people think of American Express?" is too flat to work with. "What makes American Express feel premium, practical, aspirational, or expensive in the minds of U.S. consumers?" gives the system permission to go looking for status, self-image, and emotional trade-off, the actual texture a Qual answer is supposed to surface.
If You Want Both, Say Both
This is where most mixed prompts go wrong: someone writes a question that sounds emotionally rich, then is surprised there's no ranked list in the answer, or asks purely for ranking and wonders why there's no depth behind it. A strong mixed prompt asks for scale and soul in the same sentence: "Rank the factors that drive consumers to choose GEICO when price differences are small, and explain why those factors matter most in final purchase decisions." Ranking signals quantification. Explaining why signals qualitative depth. Together, Maven has no excuse to answer only one half of the question.
Define the Population Like It Matters
A question without a defined population is a letter with no address; it might be well written, but it isn't arriving anywhere useful. "Why do people choose Cleveland Clinic?" is broad to the point of being unusable. "Why do patients and families choose Cleveland Clinic over other health systems when making a major treatment decision?" carries an actual population, decision type, and stakes. "People" is very rarely specific enough on its own.
Let the Question Carry Some Tension
The strongest research questions aren't flat, they contain a real tension the research is meant to resolve: between price and trust, between admiration and conversion, between usage and expansion. "Understand GEICO's brand" has no tension and nowhere obvious to go. "GEICO needs clarity on what makes drivers choose one insurer over another when price differences are small" has an implicit contradiction built in, if price isn't the deciding factor, something else is doing the work, and that's a real research problem rather than a vague request for background.
A few words are worth watching closely here. "Understand" without a target ("understand American Express customers") is often a placeholder for a thought that hasn't finished forming; pair it with something specific instead ("understand why some consumers admire American Express but hesitate to apply"). "Best" invites the same problem: best by what measure, according to whom? Asking what drives preference, trust, or conversion instead gives Maven something it can actually evaluate.
A Formula Worth Defaulting To
When a prompt isn't coming together, this structure tends to work: name the brand or category, state the business problem, specify the audience, signal whether you need magnitude, meaning, or both, and include the tension if there is one.
For example: "State Farm needs clarity on what makes drivers choose one insurer over another when price differences are small. Rank the factors that drive preference among U.S. drivers, and explain why those factors matter." Nothing about that is complicated. It's just intentional.