Kyle Campbell • keyone guest portal

Case study • keyone guest portal

The keyone
guest portal

End-to-end design of a web app to stand in for a reception desk: the right information at the right moment, whatever shape the data arrives in.

Role
Sole product designer
Timeline
Spring 2024 to August 2026
Outcome
8.5k monthly users

The challenge

Without a receptionist, how can you make guests feel supported?

keyone runs hotels and holiday homes across the Austrian Alps, often with nobody on site. The guest portal had to do the receptionist’s job: get guests checked in and through the door, and make sure they never felt alone during their stay.

Constraints

  • Guests used it for one stay, so it had to work without onboarding.
  • Operations, check-in, and deposits each ran on their own service, so data arrived in different shapes through different channels.
  • Organisational silos meant that a shipped design was unlikely to be revisited, so it had to be fully specified at handover.
Guest & business goals

For guests, the app had to work like a reception: help them check in, find their room, and learn about their stay, so they never felt alone or like they were missing information.

For business, the app had to reduce operational overhead - support calls about check-in, deposits, and missing keycodes. Damage reports, early checkout reporting, and upsells were secondary goals.

Where it started: a one-season MVP

Before the portal, I designed a small web app for two business problems: getting guests to register early checkouts, giving cleaners a head start on changeover days, and getting them to report damage, giving us a paper trail and faster repairs. Each unit got its own page, reached by a QR code stuck behind the door.

Both functions looked simple, and both hid states I only found while designing. It ran for one summer: damage reports came in, early checkouts less than hoped. It made the case for expansion into one app for everything about a booking. I named it the guest portal, because who wants to download another app?

Basic home screen with early checkout and damage report buttons
One page per unit, two jobs.
Early checkout screen with a warning
A warning before an action that can’t be undone.
Damage report form with tap targets marked in red
Tap targets specified for developers. Every interactive element in the portal later got room for at least 44px.

Information architecture

A layout for content that varied by property

An audit of everything a guest needs across a stay showed that property details were entered by individual account managers, in whatever format each preferred. I couldn’t assume what any field would contain, or whether it would contain anything at all. The unpredictable content meant a constrained or a static layout would fail.

Pencil sketches of layout options in a notebook
First sketches. I took the two top-right layouts into Figma, then dropped the far right once real content showed it was too text-heavy.
Options considered
An accordion of section titles, which I judged dull and slow to scan for six categories of content, or a grid of icon buttons that works without reading (the app launched in English and German only) and feels like a phone’s home screen.
Solution
A hero card for the guest’s current tasks, with the grid below. Each button opens a slide-up panel built from three components (title, text, and image), filled from backend fields in a set order. Empty fields are skipped, so any mix of content renders cleanly.
App screen with a grid of five buttons
At launch: homepage card with five buttons.
Slide-up panel with details about a ski race
Panels pull set fields from operations, or carry a one-off event.
  • Gradient between a property photo and the text over it

    Property photos varied as much as the text did, so the gradient under the title was specified to keep its contrast whatever the image.

In hindsight: the accordion
  • Grid of fourteen buttons

    The grid grew as the business added features, past fourteen buttons at some properties, and every one had to be absorbed. The accordion I’d vetoed would have handled that volume better. I didn’t give it enough consideration up front, and by the time I saw it as the better solution, the product couldn’t change.

Information architecture

The right task at the right point in the stay

A stay has phases, and what a guest needs changes with each one. The hero card only ever holds the most important information for the moment. The buttons below it cover everything else, and stay disabled or hidden until they’re relevant.

Hero card asking the guest to finish check-in, with a processing-time warning
Before arrival: check-in leads, and Getting in waits.
Hero card with the unlocked keycode, breakfast service, and local guide
Mid-stay: the keycode, breakfast, and the local guide.
Hero card with review and book-again prompts, and fewer buttons below
After checkout: two last tasks, fewer buttons.

A card that changes with the stay

The hero card shifts at every stage: before arrival, 72 hours out, check-in day, mid-stay, the last 24 hours, and after checkout. Three examples:

Before arrival
Check-in leads, with a warning that processing can take up to 90 minutes, so nobody finds out at the front door.
72 hours out
Once check-in and deposit are done, the keycode appears locked, with its release time, so guests know when to expect it.
Last 24 hours
On multi-day stays, early checkout becomes a primary button, now that it’s relevant.
States that stay readable
  • Check-in checklist with completed, active, and locked steps

    Every state of the check-in steps keeps text contrast of at least 3:1, even the locked one, where WCAG doesn’t require it.

  • Locked keycode showing when it becomes available

    The locked keycode uses a deeper version of the brand teal (7:1 on off-white) at an opacity tuned to still meet WCAG AA for large text.

  • Alert with an icon, colour, and text

    Statuses carry more than one cue, so understanding never depends on colour alone.

Protecting the hero card

The card only works if guests can trust it to be accurate and to be about right now. Both got tested.

Edge case
Operations couldn’t always confirm a check-in instantly, so a finished task could still show as incomplete. Instead, the card warned guests up front that processing could take up to 90 minutes. When the delay was later cut to seconds, the warning went with it.
Pushback
The business asked for a book-again prompt in the card mid-stay. I pushed back: mid-stay, the card is for the keycode and the stay itself, not the next booking. We compromised on a button in the grid, one of the ones that later crowded it.

User testing

Testing the check-in kiosk

As keyone’s hotel line grew, we put check-in kiosks in the lobbies: a tablet version of the portal, cut down to check-in, payment, keycodes, and Wi-Fi. I designed a payment flow around our provider’s API, and guest registration ran in an iframe, so nobody could wander off onto the open internet.

What testers saw after logging in with an incomplete booking.
Self check-in tablet on a stand in a hotel lobby
The kiosk in a keyone hotel.
Test
At a test event hosted by Empatic UX (opens in a new tab), I ran moderated usability tests with eight participants. Each got a booking code, a check-in date, and a few minutes to get their keycode.
Findings
Almost everyone stalled at least once, unsure what to tap. Critically, the iframe made the registration form look finished when it carried on below the fold. Only three of eight checked in within the allotted time.
Fix
I made smaller UI changes, based on the friction points, and the developers reformatted the iframe.
Result
A re-test at the company Christmas party confirmed guests could get through. Kiosks launched in six hotels that winter, then rolled out to more properties in the summer.
More kiosk screens
The horizontal view of the completed check in screen - a part of the adaptation was the decision to use a vertical tablet in the kiosks.
An adaptation of the payment screens that worked with our provider's API folded this process into the tablet.

Live-iteration

Fixing the five-login booking

A year after launch, a small but significant share of bookings had a problem. Multi-room bookings were split into decimal booking numbers (1000.1, 1000.2), so those guests got two to five times the emails and a separate login for every room, and had to track tasks and keycodes across all of them.

What changed

Login
Any of the booking numbers now opens the whole booking, which also cut the duplicate emails.
Summary
The hero card became a booking summary: dates, guests, and rooms.
Rooms
Each room got a collapsible card with its own keycode, getting-in instructions, and open tasks. The first card with an open task opens by default; the rest stay closed, so a five-room booking doesn’t become a 2,000px screen.
Booking summary at the top of the main screen
The booking summary, with payment as the primary action.
Room cards, one marked with an open task
The highest room card with open tasks expands.
Open room card with its keycode and buttons
Room cards hold their own keycode and buttons.
The overhaul it triggered

Some of the grid buttons were unique to each room, so they had to move onto the room cards. That broke the original layout and triggered the first (and only) overhaul of the design, which was my chance to fix some old mistakes. I revised margins, padding, and type hierarchy across the app. Most screens in this case study are from after the overhaul.

Winter 2026 data

Did it work? Yes and no.

  • 8,500+ monthly tracked users*
  • 17,000 monthly sessions
  • 2 sessions per guest, on average

The main goals landed. Guests completed the critical tasks, and support calls about check-in, deposits, and keycodes dropped, a celebrated success internally, though we never put a number on it. The secondary goals didn’t land.

What the data showed

44% of users opened Getting in and 43% Getting here, so the buttons for that stage of the stay were doing their job. Upsells, early checkout, and damage reports together reached under 5%. With 17k monthly sessions over 8.5k users, the average guest came twice: most likely once to check in and once for the keycode.

Share of tracked users who opened each function*
  • Getting in 44.0%
  • Getting here 42.8%
  • Partner offers 3.0%
  • Other features 1.9%

*Guests who rejected cookies could not be tracked.

Which devices guests used

The share of desktop users surprised me. With journey-stage data (recommended in my handover), I would have explored how to serve them better.

Share of tracked users by device*
  • Mobile 70%
  • Desktop 30%
  • Tablet under 1%

What the data couldn’t show

I couldn’t tell whether the secondary features had a discovery problem or a content problem. We couldn’t count sessions per booking or see whether guests came back mid-stay, and scroll depth was useless when page height varied by property. Without those answers, any redesign would have been a guess, so closing the gaps became the heart of my handover.

Conclusion

What I left for my successor

I analysed the winter data on my way out. My last act as keyone’s product designer was a set of recommendations, starting with tracking changes made quickly enough to catch the last of the summer guests, so my successor would have the data to design a fix before next winter.

  1. Keep phone users logged in for 48 hours. Auto-logout only makes sense on the shared tablet, and discourages coming back on mobile.
  2. Track more: journey stage attached to sessions and events, more detailed button interactions, and sessions counted per booking.
  3. Let the data pick the next step. If guests aren’t seeing the content, show it during an artificial check-in delay; if they see it and don’t use it, the content needs rework.

What I’d do differently

Give the accordion a fair hearing at the start, and build better analytics in from day one, then check them often. Both would have got us to a more useful hero card much sooner.

I still don’t believe a receptionless hotel can offer the same level of service as a traditional one. That scepticism is why I put so much heart into this: it was a problem I thought genuinely needed solving for our guests.

If anyone from keyone ever reads this (hi Mobert, Pali, Helle), cheers, and thanks for trusting me to build this.

Let’s
talk

If you came here via my CV - my contact details are there, otherwise please reach out to me on LinkedIn (opens in a new tab) or at campbellk70@gmail.com.