Dress for Thursday, Part 1: Start With the Goal, Then the Color and the Closet

What to Wear (W2W) is a theoretical model for a platform that helps people dress for what they need to do, and feel at ease doing it. Appearance shapes how others judge us and, perhaps, how we act, yet few people are ever taught to use it, and professional image advice is out of reach for most. This first of two articles lays the foundations. The model starts from a goal (“credible in Thursday’s negotiation”) and builds outward: a measured color profile, a digital closet, a daily plan, and a human stylist for the moments that matter. Each piece is examined against existing platforms and analogies from other industries, and each is given a free layer and a paid layer, with the reason for the split. Part 2 covers buying, renting, reselling, and caring for clothes, plus privacy, honest advice, and what it would take to build. It is a model to be implemented and validated step by step, not a business plan.

As You’re Seen, You’re Treated

When I was a teenager I flew to the United States by myself, and before I left, my father suggested I wear a suit. I wasn’t going to an interview or a wedding. I was going through immigration. “The way they see you is the way they treat you,” he told me.

For years I filed that under things fathers say. Then I started noticing it at work. At a company event, we bought a very good pair of shoes for a senior executive who was going to speak to the whole organization, and he never wore them. The size was fine; the style and color simply weren’t him. Another executive declined a shirt made for a brand launch because it “didn’t flatter him.” Around the same time, a friend was promoted to a C-Suite position, and the company offered him an image consultant as part of the role. It was optional, but nobody doubted it belonged to the job.

I was the suit person. A suit is a uniform, and a uniform means never deciding anything. Eventually, on my wife’s recommendation, I worked with an image consultant. I kept most of my wardrobe and changed little, yet the people on my team seemed to become more open: they shared ideas more freely and came to me for advice, sometimes about more than work. I can’t prove the clothes caused it, but it feels like it, whether because of how they see me or because of how I feel and act.

The research points in a similar direction, with caveats. In experiments at Princeton, identical faces were rated about two points higher in competence, on a ten-point scale, when paired with clothing that judges perceived as “richer,” and the bias persisted even when participants were warned about it and told to ignore the clothes.^[1]

A second line of research asks what clothes do to the person wearing them. In the best-known experiment, people wearing a white coat described as a doctor’s coat did better on attention tasks than those wearing the same coat described as a painter’s.^[2] A 2025 meta-analysis, co-authored by the original researchers, raised some doubts about the early studies while finding value in later work.^[3] Now think of the executive and the shoes that never left the box. The shoes were fine; they just meant nothing to the person who was supposed to wear them. If clothes carry meaning for the one wearing them, the meaning you choose probably matters. That is why the model begins with what you need to show up as, and not with what is fashionable.

Most people never get this kind of help. An hour with an image advisor in Mexico typically costs between roughly 750 and 2,300 pesos depending on the format,^[4] and a full process costs far more. The people who need it most, such as the newly promoted manager, the professional changing industries, or the person about to present to a large room for the first time, rarely have an employer paying for a consultant. Or even worse, an introvert who needs to feel comfortable doesn’t get this type of help. Yet the point of dressing well is not to impress. When what we wear fits who we are and what we have to do, we feel better about ourselves, our self-esteem gets a lift, and we can be more natural and closer to the people around us. That is the help this series wants to put within reach of more people.

Thursday at 10:00

Picture a manager with a client negotiation on Thursday. It is Saturday afternoon, and they are in a fitting room holding a jacket they like but aren’t sure about. In the model proposed here, they open W2W and photograph it, and a few seconds later the app would answer three questions: the color sits inside their palette; combined with pieces they already own it makes nine outfits, three of which fit the goal set for Thursday (“credible, calm, prepared”); and it overlaps with a navy blazer not worn since March.

They put the jacket back. Or they buy it, knowing exactly why. Either way, the decision would take thirty seconds instead of a week of doubt and return.

None of this is science fiction. Background removal, color extraction, and outfit generation already exist as separate tools, and closet apps are starting to move toward the moment of purchase; Whering, for instance, markets “shopping your own wardrobe” to curb impulse buys.^[5] What makes the scene possible is everything that has to come before it, and that is what the rest of this article, and Part 2, are about.

The Question Before the Camera

The first screen would not ask for a photo. It would ask one question: what do you need to show up as this month?Confident, credible, approachable, energetic, relaxed. It would then ask about the workplace dress code and how often the user meets clients. This goal becomes the thread running through every module, because the premise is that the meaning of what we wear matters more than the label on it.

Only then would the app build the color profile, which is the subject of the next section. In the model, the profile is free, and so is almost everything in what I’ll call the daily loop. Why give so much away? Research on freemium offers a hint. In a longitudinal study of Spotify users, people deciding whether to upgrade leaned on how enjoyable they expected the premium tier to be and whether it was worth the price, yet because they hadn’t used the premium features, they relied on secondhand information about them. Retention among paying users, by contrast, was driven by access across places and devices and by discovering new content.^[6] It is only an analogy, but it suggests three design rules for the free tier:

• It should be a real habit, not a crippled demo.
• It should let the user taste the paid layer once, so the upgrade is not a leap of faith.
• The paid layer should be where a real cost appears (human time, logistics, heavy computation) or where the value per user is highest.

This is the same logic as the subscription funnel in my laundry paper: earn the repeat customer first, then offer the subscription, using their own numbers.^[7]

Reading Color Without a Studio

A color profile should be measured, not guessed. The one proposed here speaks the seasonal language people already know, but it is built from measured numbers about skin, hair, eyes, and contrast, read from a live camera view that is never saved.

The best-known color systems sort people into seasons. The lineage runs from Munsell’s hue, value, and chroma (1905) through Itten’s Bauhaus teaching to Suzanne Caygill’s consulting method and Carole Jackson’s mass-market Color Me Beautiful (1980), which later practitioners expanded to twelve sub-seasons.^[8] The system is popular, and many people enjoy it, but its foundations are softer than its reputation. A vendor white paper that sells an AI (artificial intelligence) season tool concedes that the warm and cool categories “have not been validated against dermatological measurement standards in peer-reviewed literature,” and that draping, the method most consultants trust, rests on practitioner experience.^[8] A patent for a color-analysis kit calls the four-season approach an oversimplification that depends on subjective judgment, in which even experienced consultants make mistakes, although that source also has a product to sell.^[9] For a free feature meant to build trust, this matters: if two apps tell the same person two different seasons, the hook becomes a liability. Seasons are still worth keeping, because they are the shared vocabulary of the field and people already understand them. The weakness is treating them as boxes. Part of the reason consultants disagree may be that many people sit between two seasons.

The way to make seasons more reliable is to measure the dimensions underneath them: how light or deep the coloring is, whether the undertone leans warm or cool, how muted or clear it is, and how much contrast there is between skin, hair, and eyes. The first three are Munsell’s axes in modern dress. Contrast has some perceptual evidence behind it. Researchers measured 289 faces aged 20 to 70 and found that contrast between facial features and the surrounding skin declines with age, and digitally raising that contrast made faces look younger to more than 100 participants.^[10] That study was about faces, not clothes, so it supports measuring contrast, not any particular palette. On harmony, a 2014 study of 239 online participants found that outfits were rated most fashionable at moderate color coordination, not at maximum matching, for both women’s and men’s clothing.^[11] The sample was crowdsourced and the colors came from four palettes, so I use it as a design hint: a palette should be a core plus a flexible range, and the app should reward moderately coordinated outfits instead of penalizing every deviation.

The proposal is to keep the seasons as the language and use the measurements to place each person between them. Instead of one label, the app would report a blend, for example 60% Summer (cool, soft) · 40% Autumn (warm, muted). The dominant season supplies the core colors, and the secondary season widens the accent range. Because the blend comes from continuous measurements, two people are rarely identical, so each profile is close to unique while still using names anyone understands. The percentages would come from how close the measured undertone, depth, clarity, and contrast sit to the classic twelve season profiles.

Five ways to produce a profile compare as follows.

Method	How it works	Strength	Weakness	Role in the model
Consultant draping	A person holds fabric swatches against the face under controlled light	Trusted, personal, explains why	Subjective; results can differ between consultants; costs money	Paid human review
Self-report quiz	Questions about veins, hair, eyes, sun reaction	Free, fast, needs no image	Depends on how people see themselves	Confirmation step
AI season from a photo	A vision model assigns a 4- or 12-season label	Scalable, easy to use	Still quite new, so there is no evidence yet on how it behaves; could be considered with constraints	Never used alone
Colorimetric measurement	Converts pixels into device-independent values (CIELAB, a standard color space, and ITA, the Individual Typology Angle, a skin-tone index)	Objective in principle; can be tracked over time	Phone and lighting error	Core of the profile
Hybrid (proposed)	Measured dimensions, blended season percentages, and user feedback	Credible, explainable, improves with use	Needs engineering	The proposed profile

Phones are the weak link. In a preprint built on 14,475 matched images of 965 Korean participants, raw smartphone color differed from a DSLR (digital single-lens reflex camera) reference by a mean ΔE (delta E, the standard measure of perceived color difference) of about 7, and no image came within the ΔE 2 threshold the authors treat as a close match. A linear color correction brought the smartphone down to about 2.3, yet a common skin index, ITA, stayed unreliable afterward (ICC, the intraclass correlation coefficient, a measure of reliability, of 0.455, rated “poor”) because it leans on the noisiest color channel.^[12] That work is not peer reviewed, covers one population under controlled conditions, and did not test lighting. A published proof of concept from Brown and Morgan State researchers got phone-derived ITA readings to track a professional colorimeter in four participants, measured on fingers, but best with the flash on, the room lights off, and exposure settings controlled.^[13] The lesson: don’t promise precision the hardware can’t deliver, and design around the error. The model therefore proposes a calibration that asks for as little as possible, since every extra step is a reason to quit:

1. Light check. The live view shows whether the person is near a window in daylight, with a small indicator. Under artificial light, the app asks what kind of bulb and marks the result provisional.
2. White reference. They hold a sheet of blank white paper beside their face for about two seconds in that same live view. There is no separate photo; the paper gives the software a known white to correct against.
3. Done. The app extracts the numbers and discards the frames.

The artificial-light reading would happen only when it is needed, the first time the person photographs garments at home, with the paper in the first shot of that session. Paper whiteness varies by brand, a limit the calibration would have to account for. The Monk Skin Tone Scale, a ten-shade scale developed by Google with Harvard sociologist Ellis Monk and released openly, would serve developers as a yardstick for testing fairness across skin tones, not as a step for users.^[14]

Photo or live view? Live, as a matter of privacy by design. The app reads the camera preview on the phone, extracts a few numbers, and discards the frames, so no photograph is stored or uploaded and there is no image archive to leak. Face-landmark tools built for this, such as Google’s MediaPipe Face Landmarker, are optimized for mobile GPU (graphics processing unit) inference and offer a live-stream mode that returns a 478-point face mesh, which the app could use to find the regions to sample.^[15] What the profile stores is small: ranges for skin, hair, and eye color, a contrast value, a confidence level, and the conditions of the reading. That reduces what could leak and simplifies consent, but derived values are still personal data, and whether they count as biometric data under the laws that matter is a legal question I take up in Part 2.

People change, and the profile should too. A person who goes gray, changes hair color, or tans over the summer should update it with a quick re-read, since contrast shifts with each of these and declines with age.^[10][16] The re-read and a short summary would be free; a full report on how the update changes the closet, and a human review, would be paid.

This is also how the profile fits the market described earlier. In the apps I reviewed, closet apps start from the clothes, and standalone AI color tools start from the face and stop there. The model connects them: the measured profile becomes a filter that scores every garment in the closet and every item in a store, and a first reading would take about a minute. That is the first piece. The next one is the problem every wardrobe app runs into: the blank page.

Fifteen Pieces, Not Five Hundred

The hardest moment in any wardrobe app is the blank page: photographing a few hundred garments before getting any value. The existing platforms attack it in two ways. Cladwell skips the problem by starting users from pre-built capsule wardrobes, not from their own closet.^[17] Indyx sends a trained archivist to your home to catalog for you, and its styling team starts work once at least 25 items are in the app.^[18]

The model borrows both ideas. The user would start with 15 to 25 anchor pieces: what they would wear to Thursday’s meeting and what they wear most. The app would build around their goal from the first day and score each piece against their color profile. The closet itself would be free, and generously so. Acloset already lets users catalog up to 100 items at no cost,^[19] and Indyx describes its digital wardrobe as always free.^[18] Giving the closet away costs little, and it creates the asset that makes everything else valuable: a record of what the user owns, which no competitor can copy and which they would be reluctant to rebuild elsewhere.

Free would also cover the outfit builder, a “shuffle” for fresh combinations (Whering’s “Dress Me”),^[5] and basic cost-per-wear tracking, which Stylebook, Acloset, and Whering all offer in some form.^[5][19][20] Paid would start where human labor does: sending someone to catalog the rest of the closet, and deeper analytics, the kind Indyx reserves for its paid membership.^[18] With a profile and fifteen pieces in place, the app could begin to answer the question people ask every morning.

From Closet to Calendar

Next comes the daily question. Cladwell gives free users one algorithmic outfit per day,^[17] and the model proposes the same, tied to the goal the user set: today’s outfit is chosen for “credible” or “relaxed,” not for novelty. A manual calendar, where outfits are dragged onto days, would also be free.

What the user would pay for is removing the decision altogether. Automatic planning would read the calendar and the weather, rotate what has not been worn recently, and avoid what is at the cleaner’s. It would also build multiple outfits for a day that includes a morning meeting and an evening dinner, and pack for trips (packing lists already exist in Whering).^[5] The charge belongs here because this layer needs a calendar connection and constant computation, and because it delivers the most value to the busiest users. In keeping with the second design rule, the user would see one full week planned automatically before deciding whether to subscribe.

DRESSED offers the feature that matters most for this series. Its AI stylist explains why it chose each outfit, and checks in with a simple question: “How does this make you feel?”^[21] If the goal is confidence, that question is the product. W2W would add a follow-up after the event: how did Thursday go? Those answers are what teach the platform what actually works for them. (How that learning loop works, and where computer vision can replace hand-built rules, I take up in Part 2.)

A Human, When It Matters

Algorithms handle Tuesdays. Thursday’s negotiation may deserve a person. Indyx is, so far as I found, the only closet app that combines a digital wardrobe with human stylists, each of whom sets their own price.^[18] In the model, the human layer would be a paid option, because human time is the one cost that doesn’t fall with scale, and it would be reserved for high-stakes moments: a 20-minute review before an important event.

The human layer also suggests a way to reach people who would never buy it themselves. My friend’s company paid for his image consultant when he became a C-Suite member. Employers could sponsor this module for the people they promote or hire into client-facing roles, at a fraction of the cost of a full image consultancy. A platform that pairs a measured profile and a digitized closet with short stylist reviews can serve a whole cohort of managers, which a consultant working one client at a time cannot.

What Comes Next

By now the manager has what most people never get: a color profile built from measurement, a closet that has been cataloged once, a daily outfit tied to the goal they set, and, when the stakes are high, a person to ask. Almost all of it is free, and all of it starts from what they need to show up as on Thursday.

But no closet is complete. Once the platform knows their palette, their closet, and their calendar, a series of natural questions opens up. What is missing from the closet, and what should they buy and from whom? What should they sell or swap when they outgrow a piece? Nuuly shows the appetite for renting: it generated about 568 million dollars in 12 months with roughly 420,000 subscribers.^[22] What if they could rent next week’s clothes, and not just a gown for a wedding? And who washes, presses, and returns all of it? That last question is the one I tried to answer in the laundry paper. Each of these is a module that plugs into the same closet, and each is optional.

Part 2 takes them one at a time. It also asks the harder question: how can a platform that earns from brands, rentals, and data still give advice you can trust?

Sources

1. Oh, Shafir & Todorov, “Economic status cues from clothes affect perceived competence from faces,” Nature Human Behaviour (paper; summary in Princeton Alumni Weekly).
2. Adam & Galinsky, “Enclothed cognition,” Journal of Experimental Social Psychology, 48(4), 2012 (abstract).
3. Horton, Adam & Galinsky, “Evaluating the Evidence for Enclothed Cognition: Z-Curve and Meta-Analyses,” Personality and Social Psychology Bulletin, 51(2), 2025 (record).
4. Cronoshare México, personal shopper prices. A services marketplace, so an order-of-magnitude reference.
5. Whering on Google Play.
6. Mäntymäki, Islam & Benbasat, “What drives subscribing to premium in freemium services?,” Information Systems Journal, 2019 (PDF).
7. The Outsider’s Advantage: Building a Laundry Business on Lessons Ten Other Industries Already Paid For, Pluris.
8. HueCheck, “The Science of Seasonal Color Analysis” white paper. Published by a vendor of an AI color-analysis app.
9. US Patent 5,924,426, Cosmetic personal color analysis method and kit. Written by the inventor of a competing method.
10. Porcheron, Mauger & Russell, “Aspects of Facial Contrast Decrease with Age and Are Cues for Age Perception,” PLoS ONE, 8(3), 2013 (summary in ScienceDaily).
11. Gray, Schmitt, Strohminger & Kassam, “The Science of Style: In Fashion, Colors Should Match Only Moderately,” PLoS ONE, 2014 (full text).
12. Kang, “The Color-Clinical Decoupling: Why Perceptual Calibration Fails Clinical Biomarkers in Smartphone Dermatology,” arXiv preprint 2512.21988v2 (preprint). Not peer reviewed.
13. “Smartphone tristimulus colorimetry for skin tone analysis at common pulse oximetry sites,” Biophotonics Discovery, 2(3) (report in Optics.org).
14. Google, Monk Skin Tone Scale.
15. Google AI Edge, MediaPipe Face Landmarker.
16. Schmalwieser et al., study of seasonal skin color in Austrian farming families, Photochemical & Photobiological Sciences, 2025 (article). Exploratory, small sample.
17. Stylebook vs. Cladwell, published by Indyx, a competitor, and Cladwell on the App Store.
18. Indyx: how it works, Clueless closet, stylist FAQ, homepage. Prices differ between pages.
19. Acloset on App Pricing Lab.
20. Indyx vs. Stylebook, published by Indyx, a competitor.
21. Think Outside the Closet, review of DRESSED.
22. eMarketer, Nuuly’s rapid growth signals momentum for fashion rental.