When we explain the Atorie style profile to new members, the most common misunderstanding is treating it as a static result. People complete the quiz, get their profile summary, and assume that is the finished version of what we know about their taste. It is not. It is the starting version, and there are specific things the profile updates on over time that are worth understanding if you want to use Atorie well.
I want to be honest about both what the profile does and what it does not do. The product works best when people understand its actual mechanics rather than imagining something more sophisticated than it is or writing it off as too simple.
What the initial quiz actually captures
The quiz establishes four things. The first is your occasion map: how you are actually spending your time across different dressing contexts, roughly how often each context appears in your week, and what the formality register of each tends to be. This is not about what kind of person you want to be. It is about what occasions your wardrobe actually has to serve.
The second is your material preferences. We ask about fiber preferences (natural versus synthetic, weight preferences, texture sensitivity) because material is a primary driver of whether someone actually reaches for a garment. A person who finds wool uncomfortable will not wear a beautiful wool coat regardless of how much they like its appearance. Material preferences are also more stable than aesthetic preferences, which shift with exposure and mood. Your fiber preferences from two years ago are more likely to still be accurate than your aesthetic preferences from two years ago.
The third is your silhouette orientation: whether you prefer garments that fit closely, garments with volume, or something in between, and whether that preference varies by garment category. Many people have silhouette preferences that are category-specific: they prefer a closer fit in tops and more volume in trousers, for instance. The quiz tries to capture that specificity.
The fourth is your color relationship: your anchor colors, your statement color range, and whether you tend to operate in high-contrast or tonal combinations. We ask this in terms of your actual wardrobe as it exists today, not your aspirational wardrobe. Aspiration is useful context but it is not the same as what will actually integrate with your current clothes.
The signals the profile updates on
After the initial quiz, the profile refines itself based on three types of signal.
The first is save and dismiss behavior. When you save a piece to your collection, we note the attributes of what you saved: its construction type, material composition, silhouette category, color placement, and occasion suitability tags as we have labeled them from the label's intake information. When you dismiss a recommendation, we note those same attributes. Over time, patterns emerge that are often more specific than what people articulate in a quiz. A member might describe themselves as broadly interested in minimal aesthetics, but their save history reveals a consistent preference for a very specific kind of structural minimalism that is different from the organic minimalism they had also marked as interesting. That specificity shows up in the save data before it is legible enough to put into words.
The second is the wardrobe context tags you add to saved pieces. When you tag something as "actually wearing this" versus "aspirational save," that distinction affects how we weight the piece's attributes in your profile. A pile of aspirational saves in one direction and consistent wear-pattern saves in another direction tells us something important about the gap between your stated preferences and your practical habits. We update toward the practical habits.
The third is explicit feedback. When we surface a recommendation that is genuinely off, there is a feedback option that lets you tell us specifically what is wrong with it. We do not use this to disqualify categories entirely. We use it to adjust weights within categories. If someone tells us a recommended piece is "too formal for how I actually dress," that is not a signal to stop showing them formal pieces. It is a signal to recalibrate what level of formality we are treating as baseline for their profile.
What the profile does not learn
There are things the profile cannot pick up that are worth knowing about.
It does not learn from browsing without engagement. If you look at something for a long time without saving or dismissing it, we do not treat that as a signal. We made a deliberate choice not to use passive attention as a data point, because passive attention is too noisy: a person can spend three minutes looking at a piece because they are confused by the price, trying to figure out if they have something similar, or considering buying it as a gift. Without an explicit action, we do not know which.
It does not learn from purchase data. We do not have visibility into what you actually buy. The profile is entirely built from what you tell us in the quiz and what you do within the Atorie interface. This is a limitation: if you buy something from a label through Atorie's discovery and then dislike it when it arrives, we have no way to register that negative outcome. What we can register is if you come back and dismiss pieces from that label, which is an indirect signal.
It does not update very quickly. We built the profile to be stable rather than reactive. A single save or dismiss does not meaningfully change your recommendations. It takes a pattern of consistent behavior across a reasonable number of interactions before the weights shift enough to change what you see. This is intentional. People have mixed days, make exploratory saves, and change their minds. A profile that swings dramatically in response to a few actions in a single session would not accurately represent anyone's actual taste over time.
Where the profile currently falls short
I want to be direct about a genuine current limitation: the profile handles garment attributes well but outfit assembly less well. We know your preferences for individual pieces with reasonable accuracy after a few months of active use. We are much less good at understanding which pieces from your saved collection actually work together, because we do not have enough information about your existing wardrobe to assess compatibility at that level.
The wardrobe-upload feature we are working on now is aimed at this gap. If we know what you already own, we can start filtering recommendations toward things that extend your existing wardrobe rather than just matching your taste in isolation. That is a meaningfully different kind of recommendation and it requires a meaningfully different kind of input from you. We are not there yet. What we have now is accurate for individual piece discovery and less accurate for understanding how a new piece would actually integrate with what you own.
We think that transparency about this limitation is more useful than leaving people to discover it by noticing that our outfit context suggestions sometimes feel disconnected from their actual wardrobe. The profile is a useful tool for discovery. It is not yet a complete wardrobe management tool. Those are different things, and we want to be honest about which one we are right now.