Just Hair

An AI hair-care assistant for women with textured hair: it explains your hair before it recommends anything, so years of trial and error become a routine you can trust.

Status

Concept, tested hi-fi prototype

Context

Solo project, ELVTR product design program

My role

Product and UX, end to end: research, prioritization, design and testing

Timeline

6 weeks

Segment

High-spending professionals with textured hair

Research

6 interviews, U.S. and Indonesia, ages 20s to 40s

Tested with

6 participants (mid-fi) and 7 unmoderated Maze testers

North star

Routine followed for 3+ weeks

Stack

Figma, Maze, Notion; LLM feasibility in Lovable

1. Background

Just Hair is my concept project from the ELVTR product design program. I chose it because textured-hair care is a deep, global problem that the industry underserves, and because large language models can finally read a photo and explain what they see in plain language. I owned every stage, and tested the LLM connection for feasibility in Lovable.

2. Project overview

Vision. Help women with textured hair understand their own hair first, then care for it with confidence and spend less getting there.

The market. Black consumers spent $6.6 billion on beauty in 2021, yet are 1.6 times more likely than White consumers to be dissatisfied with hair care products, and only 13% say it’s easy to find products that meet their needs at mass retailers (McKinsey, 2022). In 2017 they accounted for $54 million of the $63 million spent on ethnic hair and beauty products (NielsenIQ, 2018). High spend, low satisfaction: the gap is guidance, not supply.

Problem statement. Most people with textured hair don’t fully understand their own hair, and often take a decade or more to learn it. They rely on reviews, influencers and costly trial and error, so the industry puts the burden of expertise on the customer.

3. Ideation overview

Scope. A solo project: I made the product calls and tested them with users.

• 6 interviews over Zoom and phone: 2 with tightly textured hair, 3 medium textured, 1 straight, across the U.S. and Indonesia, ages 20s to 40s
• Secondary research on the hair-care market and on what LLMs can reliably analyze
• A competitive scan of curl quizzes and product lines participants already used (Prose, Cantu, Shea Moisture)
• Affinity mapping into six themes, then problem statements, POVs and How Might We questions

Target customer. Three personas came out of the research:

4. Customer segmentation

We prioritized high-spending professionals with textured hair, the Jasmine segment. Their spend is the business case: redirect it to the right products and the product earns trust and a path to revenue.

Scored 1–5, retrospectively from research findings.

Parents scored close behind and stay in the roadmap as household profiles; the same analysis and routines serve them once the core works for one person.

5. Customer pain point

We prioritized not understanding my own hair, the root cause behind years of trial and error.

Scored 1–5, retrospectively from research findings.

Every interview told the same story: “A lot of Googling, a lot of YouTube,” products bought and abandoned, and routines that broke when life or climate changed (“In Bali, I wash every day”). Product fit scored second, but recommending products to someone who doesn’t understand her hair only restarts the cycle, so understanding came first.

6. Prioritized solution

We chose photo analysis with a plain-language hair report, fed by a short lifestyle questionnaire.

Effort mapped from my feature backlog’s sizing (S = 4, M = 3, L = 2); scored retrospectively.

The analysis answers the top pain directly and was small to build on an existing multimodal model. The questionnaire and recommendations came with it, because a routine that ignores real life fails.

Key product decisions

What we cut, and why

7. Develop MVP

The MVP is one loop: photo → plain-language report → lifestyle questions → routine and matched products → reminders and rescans, with a thumbs up or down on every answer.

North star: routines followed for three or more weeks. An analysis is a one-time novelty; weeks of routine change hair and earn trust.

User stories and acceptance criteria

The test plan’s bar: complete each task with no errors.

User flows

The task flow is the happy path; the user flow shows every branch after the first report. Select a flow to open it full size.

8. Launch plan

Just Hair hasn’t launched; this is how it will go to market. An iOS version is in definition.

• First market: high-spending professionals with textured hair in the U.S., starting with the research participants and a waitlist.
• Promotion: natural-hair creators and communities on YouTube and social, where research showed women already go to learn.
• AI approach: ground an existing model in a curated hair-care knowledge base instead of training one. It ships fast, stays grounded in vetted hair science and updates without retraining. Fine-tuning comes once real thumbs up and down signals exist.

How it will make money

The risk to manage: affiliate money can bias recommendations, so they will stay tied to the hair profile, with sponsorship labeled.

Privacy, built in from day one

• Explicit consent before any data improves the AI
• Entries anonymized: names, location and photo metadata stripped
• A clear “don’t train on my data” option
• A plain-English explanation of how data helps the user

9. Customer feedback and iterations

Two rounds of testing shaped the final design. First a mid-fi test with the six research participants across three flows: complete a hair analysis, create a routine and find products.

Then an unmoderated Maze test of the routine flow, with no help: 17 people started and 7 finished.

Testers liked the routine questions and the clean layout but found the light-blue buttons looked disabled and the background too busy. The nine changes that followed: a stronger call-to-action color, outlines instead of background scribbles, more photos, guide text, simpler copy for all levels of English, a stronger blur behind pop-ups, clearer checkboxes and better contrast.