Dual-Sided App · Marketplace + SaaS
GLAMIT is a dual-sided beauty app. One side is where clients find and book. The other is a smart business management system, where business owners run the calendar, the clients and the income, with an AI assistant working across both. I was the sole designer and owned both sides end to end, from the research to the design system that holds them together.
From the research
Survey of 60 clients and beauty business owners.
The Challenge
Client view
Urgent needs do not wait for business hours. A broken nail before an event, a fill that cannot slip another week. Finding availability on short notice means calling and messaging one business after another, and the answer arrives one reply at a time.
Business owner view
A last-minute cancellation is a hole in the day and in the month. Filling it means messaging regular clients one by one, hoping someone answers in time. Most owners do exactly that, and most of the time the hour stays empty.
Research
I surveyed clients and beauty business owners separately, to check whether the two problems I suspected were in fact one. They were: the hour one client cannot find is the hour sitting empty in someone else's calendar.
For clients
For business owners
From a finding to a feature
The window opens on its own, and the assistant invites the clients who searched for it.
The feed opens on what is bookable now, and every result carries its nearest open hour.
Appointments, revenue and costs share one day view, so the month adds up while it runs.
Competitive Analysis
I benchmarked the Israeli beauty apps a business here actually runs on, the general scheduling tools around them, and the global marketplaces that share GLAMIT's model. The Israeli tools manage a calendar and stop there. The marketplaces do fill a cancelled hour, by notifying whoever joined the waitlist first. The question none of them asks is who was looking for that hour.
| Capability | Israeli beauty apps Calmark · Plannie · Pink | Global marketplaces Fresha · Booksy · Treatwell | GLAMIT What I designed |
|---|---|---|---|
| Both sides in one account | One role | Two separate apps | ✓One account, one tap |
| Search by inspiration photo | Absent | Portfolio, one business | ✓The photo is the query |
| Filling a cancelled hour | Manual messaging | Waitlist, first in line | ✓Matched to who searched for it |
| The assistant's authority | No assistant | AI fills calendar gaps | ✓Green, amber, red, set by the owner |
| Client history | Client files | Client records | Automatic, on every candidate row |
Who it's for
Two people, opposite sides of the same hour.
A full-time job and three kids. She gives up appointments more often than she books them, because finding an open hour costs more time than the appointment does.
Her day is usually full. A cancellation is not a free hour, it is income she has no time to chase. She wants the calendar updated in seconds, and then to get back to the chair.
The AI assistant
On both sides the assistant collapses multi-step work into a single action. For clients it turns a photo and a sentence into businesses filtered on style, area, availability and price, and surfaces an earlier opening when one appears with the reason it fits. For business owners it matches the clients whose recent searches fit a cancelled slot, compares takings across eight weeks to find the weakest day, and composes the outreach to a client who has stopped returning.
Each capability was classified before it was designed, by one question: what may it do without being asked. Anything reversible runs on its own, anything prepared for the user waits at the button, and anything that commits money or closes an appointment is the user's, including the appointments the assistant found itself.
Design system
Many GLAMIT users are on both sides of it, running a business in the morning and booking for themselves at night, one tap apart. So the color is orientation: one component set, one accent token, and a theme per side.
Feature spotlight · Business side
92% of owners re-fill a cancelled slot by messaging clients one at a time. So the empty hour carries a price on screen: a cancellation arrives as 180 ₪ lost and 105 minutes free, and the recovery engine starts from that number. It opens the window, finds candidates and explains each one. The owner picks. The owner confirms.
The recovery flow
1 · Sign in
2 · The dashboard
3 · The morning
4 · Candidates
5 · Filled
6 · The day, closed
Feature spotlight · Client side
76% of clients want to see the work before they book, and 91% want availability first. So the search starts with a picture already saved and ends at a chair bookable tonight.
The full user flow
1 · Home
2 · The ask
3 · The answer
4 · The business
5 · Choosing a time
6 · Before paying
7 · Payment
8 · Confirmed
9 · It stays with the client
Beyond the appointment
Every two months the owner reports VAT, and every deducted expense needs a receipt behind it. The app already knew what the business earned, so I decided it should hold the expense side too. Now the whole period is handled in one place.
The screen opens on where the period stands, both sides of the ledger at once, so the owner knows what is settled before doing anything. Below it an expense can be added, what still needs attention appears with its cost and its deadline, and the VAT owed sits beside the two amounts it came from.
The period arrives already assembled, which is where the time is saved. Attention goes straight to the part that still needs a decision, and from there it can be closed and passed to the accountant.
The sheet opens over the screen underneath, so the owner keeps their place while checking it. It describes the bundle before anything is sent: how many documents, the single file they gather into, and how they split between clients and suppliers.
It also says the period locks and can be sent again at any time. Saying so at the moment of commitment is what makes the last step easy to take. Sending is the only filled action, and keeping a copy sits below it as a quiet link.
The same system, wider
A nail technician works from a phone. A salon owner with three chairs and a bookkeeper works from a desk. Same tokens, same components, same assistant, laid out for 1440: the bottom bar becomes a sidebar, and the raised AI action moves into it and keeps its violet.
Three screens, enough to prove the system carries. Not a second product.
The day and the month sit side by side instead of one under the other, and the appointment rows keep the phone's status badges and the assistant's line about the cancellation.
The KPIs answer whether the month is fine, the charts answer why, and the card points at what needs attention. The assistant reads the same numbers and advises on the business.
The assistant keeps its own entry point rather than becoming a widget in a corner. Same six questions, same answer, same line about drafts waiting for approval.
Happy to walk through the decisions, what each one cost, and what I'd change if I built it again.