Kira Vella Thinks Most Product Problems Begin Earlier Than Teams Realize.

Most product teams make reasonable decisions based on an interpretation of the problem that nobody has examined closely enough.

Most product decisions that go wrong looked reasonable at the time. When CB Insights read the post-mortems of 431 venture-backed companies¹ that shut down since 2023, 43 percent named poor product-market fit, and two-thirds of those were early-stage teams that never found a market at all. Between them, the 431 had raised $17.5 billion.

Kira Vella has spent the past year building Studio Marelle around the moment before that money is spent. Companies that survive carry a quieter version of the same problem. Pendo’s 2019 analysis of usage across 615 software products² found that 80 percent of features were rarely or never used, and somebody made a confident case for every one of them.

Vella’s view is that the case is rarely foolish. It is usually a sensible response to a problem statement nobody went back and checked.

What follows is her method for checking it: what to ask before commissioning any research, and the one question she puts to strangers about every product.

The Problem Arrives Already Decided

Teams do not usually come to Vella with a question. They come with a diagnosis.

“I don’t think most product teams make obviously foolish decisions,” she says. “Usually the opposite. They make completely reasonable decisions based on an interpretation of the problem that hasn’t been examined closely enough.”

She hears the interpretation in the first sentence of the brief. “We need to improve onboarding.” “Users don’t understand this feature.” “People aren’t ready to trust the AI.” “We need a redesign.” Any of those might be true. The trouble, in her words, is that “by the time a team is saying them confidently, an interpretation has already hardened into a problem statement.”

This is why she has never liked the most common piece of startup advice. Validate your idea, she has written, “frames the goal as confirming that your current understanding of the problem is correct.” In her own answers she is blunter: “It makes it too easy to look for reasons to keep going. I want to know what we might be missing and what would make us reconsider.”

The mistake isn’t having assumptions. You cannot build anything without them. The mistake is forgetting that they are assumptions.

Stop Describing the Solution

Her first request is a small one. “Usually I ask them to stop describing the solution for a minute,” she says. “If somebody says, ‘Our onboarding isn’t working,’ I want to know what ‘working’ was supposed to look like. Were people meant to reach an activation moment faster? Understand the product better? Come back the following week? Invite somebody else? Pay?”

Then she wants what the team has actually observed. “There’s often a surprisingly large gap between the problem a team initially brings me and the decision they really need help making.”

The four questions she keeps returning to are plain ones. “What did you expect people to do? What are they actually doing? Where did the expectation come from? What evidence would change your mind?” She has been asking versions of them since 2021, when she was the founding research strategist for an early-stage climate and agriculture platform, and through a year of contract behavioral research at Meta.

Those questions decide whether a study happens at all.

Before proposing a study, I want to know what decision the team needs to make and what evidence could change it. That determines the work. Sometimes a usability study is exactly right. Sometimes the team first needs to decide who the product is for. The interesting part is deciding what you need to know badly enough to deserve somebody’s time and a company’s money.

The instinct behind it is visible away from work. Vella walks cities for hours with no destination, photographing doors and noticing where “somebody has improvised around bad design.” She calls it flâneusing. She sees the connection to research in the habit of noticing what people do without being asked to explain themselves. “I’m interested in what people say, obviously, but I’m usually even more interested in the tiny inconsistencies around what they say. The workaround. The pause before an answer. The thing somebody insists doesn’t matter and then spends ten minutes talking about.”

Eight People, Eight Products

One engagement from the past year shows the gap between brief and decision in practice. Vella keeps the client general, as she does with all of them.

“One team I worked with had a pretty clear idea of what they were building,” she says. “Then I put the product in front of eight people, and somehow we ended up with eight different products. One person thought it was an ingredient scanner, another thought it was a delivery app, someone else expected price comparisons.”

These were not eight people confused about the same product. Each had formed a different idea of what the product was and what problem it solved, and each idea made sense on its own terms. “They were taking whatever looked familiar and working from there,” she says. The team saw a recommendation network. Users saw a scan button and, in her words, “filled in the rest.”

There was real interest in the underlying idea, discovering products through people you trust. But the experience was not making those people visible, and there was no useful way to save something you wanted to try. “Even if a recommendation caught your attention, what were you supposed to do with it?”

The budget was limited, which sharpened the recommendation. “I separated the things we could address through clearer language from the functionality that needed investment,” she says, “and recommended making the people behind the recommendations more visible, clarifying what scanning was for, and prioritizing a simple ‘want to try’ list.”

The project left her with a question she now asks about every product and describes as embarrassingly basic: what do you think this does?

“It’s very easy to forget how much you know about your own product that nobody else does.”

The Roadmap Is a Set of Beliefs

The line Vella wants to be known for sounds like a description and is really a warning. “A roadmap is one of the clearest expressions of what a team believes is true about its product,” she says. “Underneath every item is an argument: this is the problem, this is why it is happening, this change will help, and it matters enough to prioritize over something else.”

“Research is often treated as something adjacent to strategy. But those judgments about what is happening and what deserves investment are strategy. If the understanding is wrong, speed just gets you somewhere expensive faster.”

AI raises the stakes of that argument rather than changing it. A team can now execute against an assumption far faster than it can examine one, which makes the quality of the first interpretation more consequential, not less. “The hard part isn’t making sense of the data anymore,” she has written. “It’s knowing what to pay attention to in the first place.”

The second AI problem is one most product teams skip. In a post this month she drew a distinction: “capability and willingness are two different thresholds.” A system may be able to complete a task long before anyone is comfortable handing it over, and, she wrote, “Increasing capability doesn’t necessarily increase willingness at the same rate.”

capability and willingness are two different thresholds

This week she wrote that Studio Marelle is adding retrievability to its AI adoption framework, alongside capability, willingness and appropriate control: the question of what a person needs to remember at the moment of review, and what in the experience will help them remember it. “Oversight has to be designed for the reality of the workday.”

Being the Thing Under Test

The test of all this came after she founded Studio Marelle in August 2025, with a year of behavioral research at Meta and a founding research role behind her. She had a site, a clear positioning, “and very little proof yet that the market cared.”

“I knew I could do the work,” she says. “None of that automatically creates a business. That was uncomfortable because I’m someone who likes evidence, and suddenly I was the thing being tested.”

I wondered whether I should make the studio broader, cheaper, or louder. But if I changed the business every time the market made me anxious, I’d never actually learn anything. So I kept narrowing. I talked to more founders.

She watched which conversations had energy, changed parts of the offer when the evidence supported it, and “left other parts alone when I was simply impatient.” It is her own method, applied to its author.

“Building a research business has been a fairly effective way of discovering whether I actually believe my own advice.”

The market, it turned out, did care. One client wrote that Vella “looked at our existing business as we were reshaping it, and applied a multi-lens approach. Her perspective and insight guided us in directions we may otherwise have missed.”

Where Kira Is Now

Vella is based in the New York City area and works with founders and product leaders wherever they are. Studio Marelle takes on teams that need to understand what is actually happening before deciding what to build next. Outside work there are the cities: she spends a lot of time wandering through them, photographing doors and collecting small pieces of visual and cultural ephemera.

“I want to work with teams using AI on problems that genuinely matter,” she says, “where the stakes are high enough that understanding people well is part of building the technology responsibly.”

Kira Vella runs Studio Marelle at studiomarelle.com.