UI/UX Design · Product Design

Hodo — Travel App

An AI travel companion that matches places to how you actually travel.

There are more places to visit in a new city than anyone can sort through, and the person who wants a quiet morning gallery is handed the same list as the person who wants the loudest bar. Hodo learns which one you are first.

Hodo — Travel App — the finished product
Role
Co-founder, User Experience Designer, Product Designer
Toolkit
Figma, Miro Board, InDesign, Adobe XD, Asana
Team
Soumya Sharma, Nikita Gokhale, Charmi Chovatia
Duration
Dec 2021 – May 2022

01 The problem

A new city, a thousand options, and a list that doesn't know you.

Solo travellers arriving somewhere new, who want a place worth going to and don't have the days it takes to find one.

Everybody gets handed the same city.

There are plenty of blogs and reviews about anywhere worth visiting, and plenty of them are misleading, biased, or simply written for somebody else. Sorting the useful ones out takes research time that people travelling for a week don't have, so they do the safe thing and go where the city is already famous for going.

Somebody who likes to party can't work out which pub in a new city is the right pub. Somebody who wants a quiet morning is handed exactly the same list. That was the challenge: to say anything useful, the app had to know something about the person first, and it had to have real places to say it about.

  • Solo travellers
  • A city they don't know
  • No time to plan it
  1. G-01

    Learn the traveller

    Collect the right information from users to help them discover their own travel personality.

  2. G-02

    Recommend simply

    Make a simple place recommendation based on what actually interests them.

  3. G-03

    Somewhere to keep it

    Give travellers a wish list, so a place they like now survives until the trip.

A recommendation is only as good as the question it starts from. Ours started with a personality quiz.

02 Research

I asked a lot of questions before I found the right one.

Interviews and surveys came first, then an impact-versus-effort graph, so the first build contained only what was worth building.

The question board was a map of everything we didn't know yet.

The interviews ran from how somebody spends a free weekend to how they'd plan a trip on somebody else's behalf. They gave us the pain points and the requirements in the participants' own words, and those answers set every decision that followed.

Plotting each feature on an impact-versus-effort graph turned all of it into a build order: we could see what a feature was worth against what it would cost us in time, and put only the highest-value ones into the first round of user testing.

The interview board

every question we went in with, grouped by what it was trying to learn

  1. The core script

    19 items

    asked of every traveller

    • Tell us a little bit about yourself.
    • What do you like to do in your spare time?
    • What are your thoughts about travelling?
    • How often do you travel?
    • What is your purpose behind travelling?
    • What kind of a travel planner are you?
    • How do you decide the budget?
    • What does your trip itinerary usually look like?
    • How do you decide what places you want to visit?
    • Tell us about your process to plan a group vacation.
    • Have you ever had a bad experience while travelling? If yes, what was the reason?
    • How do you react to a situation where nothing is going as planned?
    • What makes a good travelling experience for you?
    • Tell me about a time you have used technology to assist you in trip planning.
    • What would your successful experience for planning a trip look like?
    • What are your thoughts about an app that plans a trip for you?
    • What would you ideally have such an app do?
    • What would be some concerns you may have about this app?
    • What information would you like to see in this app?

  2. The probes

    9 items

    follow-ups, asked when an answer opened a door

    • Are you planning to travel somewhere soon?
    • Do you always plan your trips?
    • What are some limitations while traveling?
    • Do these limitations decide where you want to go?
    • How is individual travel different from travelling in a group?
    • Have you taken help from a travel agency?
    • What is the ideal way to travel, according to you?
    • Does Google fulfil your need for information about travel planning?
    • If you think that it already exists, can you share a particular app that has helped you, and in what ways?

  3. The agency side

    9 items

    asked of the travel agents

    • Does your travel company have pre-planned trips, or customized trips?
    • What kind of travellers do you interact with?
    • How many custom trips do you plan on an average every year?
    • What are the things that you take into account while planning a trip for a client?
    • What kind of research goes into planning such a trip?
    • How do you cater to their overall needs and make sure that they have a well rounded experience?
    • How is your process to plan a trip for a large group different from that for planning for an individual or a couple?
    • What are some problems that you come across when planning a trip for others?
    • If you could do this whole process in a different way, how would you do it?

  4. The process walk-through

    5 items

    a planner's day, step by step

    • What different kinds of travellers do you interact with?
    • What are the important steps of your trip planning process?
    • Do you want to describe any most important step in detail?
    • Do you use any online software to plan the tour, or do it manually?
    • Walk us through your trip planning process.

The probes were the ones that turned out to matter: how people decide where to go, whether they trust what a search engine tells them, and what a good trip felt like the last time they had one.

I am lazy to find a good place to visit during weekends
User research interview
I would love to explore a new place in my city
User research interview

Two of them are kept exactly as they said them. The first one is the product: the effort is the problem, not the appetite.

Why anybody goes anywhere.

six reasons, from the first round of interviews

Travel the word at the centre of the board, and the six things it branched into

  1. Opening mind

    People try new things that they are not familiar with.

  2. For work

    Some people travel to a new city, where work takes them.

  3. Learn something new

    People travel to learn new skills and different cultures, which include history and food.

  4. Explore

    Many like to find places on this planet that are secluded.

  5. To relax

    A lot of people want to get away from their regular life.

  6. Spend time with family

    People also bond with family and friends.

There are a lot of reasons to go somewhere new, and no two travellers rank them the same way. Grouping the ones that behave alike is what made the data collectable, and it is also what kept the categories broad enough that nobody falls outside them.

Five travel personalities, not twelve.

We combined the personalities that behaved similarly, which did two things at once: it made the dataset possible to collect, and it kept the categories inclusive enough that a real person recognises themselves in one of them rather than being sorted into a box.

  1. P-01

    Culture Creature

    Art, history and the facts behind both — festivals, museums, fairs, theatre, old forts and castles, and the local spots where a place keeps its heritage.

  2. P-02

    Active Adventurer

    Walking over driving. Hikes, treks, trails and adventure sports, up early, out all day, and comfortable choosing adventure over comfort.

  3. P-03

    Avid All-Rounder

    A bit of everything: art and history, the city's culture, food and nightlife, pubs and clubs, adventure spots and hikes, and shopping.

  4. P-04

    Leisure Lover

    The family vacationer — time with the people they came with, hotels and activities that actually relax them, sightseeing worth a photograph, and their own pace.

  5. P-05

    Party Person

    Dressed up and out late — the pubs, clubs and nightlife, wineries and breweries in the daytime, live bands and concerts, and somewhere good to eat after.

Two people the app is for.

the two personas the interviews produced

  1. Carla Jonas

    Senior Software Engineer
    Leisure Explorer
    Age
    42
    Family
    Married
    Location
    Boston

    I love travelling! I wish I had an assistant to plan my trip as per my preference.

    • Busy schedules
    • Hard working
    • Loves travelling
    • Semi-organised

    Carla is a mom of two little kids. She is very ambitious and loves to stay fit and healthy. Her husband and her try to organise 2 family trips per year. They love spending time together, traveling, exploring new places and doing family-friendly activities. They are willing to spend more on activities and experience.

    Personality
    • Indoorsy Outdoorsy
    • Planned trip Impulsive
    • Touristy places Hidden gems
    • Leisure traveler Active traveler
    Goals
    • Planned trip itinerary and best service
    • Create memories for family
    • Discover new destination
    Frustrations
    • Too much time spent on searching and planning
    • Time consuming to find places based on our preference
    Preferred channels
    • Chrome
    • Mobile
    • Email
    • Traditional ads
    Motivation
    • Price
    • Comfort
    • Convenience
    • Safety

  2. Andrew Scott

    Project Manager
    Adventurous
    Age
    28
    Family
    Single
    Location
    San Francisco

    I just want to have a reliable recommendation system.

    • Organised
    • Productive
    • Active
    • Extrovert

    Andrew is a single young adult who likes to travel to different cities and explore the new places. He loves nature and likes to do a lot of adventure sports in his free time. He is a very busy man, does not have any time to plan his own trips, and he doesn't rely on the internet for the right information.

    Personality
    • Indoorsy Outdoorsy
    • Planned trip Impulsive
    • Touristy places Hidden gems
    • Leisure traveler Active traveler
    Goals
    • Visit all the hidden hiking trails in a city, and the sunset point
    • Save time in finding the right places
    Frustrations
    • Hates misleading information on the internet
    • Wants to explore nature but couldn't find a good place
    Preferred channels
    • Chrome
    • Mobile
    • Email
    • Traditional ads
    Motivation
    • Price
    • Comfort
    • Convenience
    • Safety

Carla plans two family trips a year and resents every hour of the research. Andrew wants the hidden trail, and doesn't trust the internet to find it. Different lives, same complaint: the planning costs more than the trip is worth. Both of them would take a recommendation over a search result, provided the recommendation knew anything about them.

03 The design

Quiz, then flow, then screens, and the flow changed twice.

The first version explained itself constantly. Every iteration after that took something out.

The flow that survived

second iteration — after the faults in the first one showed up in testing

Hodo onboarding, then sign up or log straight in

  1. F-01 The quiz a handful of questions
    • Onboarding explains what the quiz is for
    • What matters most to you while travelling
  2. F-02 The result one of five personalities
    • Active Adventurer
    • Culture Creature
    • Leisure Lover
    • Party Person
    • Avid All-Rounder
  3. F-03 Home recommended for you
    • Destinations matched to the result
    • Or search a city and country yourself
  4. F-04 The place not a list of links
    • What's there, and the best season for it
    • Like it to keep it
  5. F-05 Saved trips the wish list
    • Everything liked, held for later
    • Open an individual trip to see it again

This is the second flow, not the first. Testing found faults in the plan my own trip branch, so we simplified and deliberately reduced the number of affordances in the app; after the second round we changed the flow again and added the detail about each location in the dataset, which is what gave people a reason to stay on a screen. The most ambitious idea died here too: the first design sent travellers straight from the quiz to a screen of recommended places, the way Netflix suggests films. The AI model behind it would have taken far longer to produce results in that shape than we had.

Drawn by hand first, so we could throw them away cheaply.

sketch Two hand-drawn Hodo frames side by side: the quiz screen, Lets take your travel personality quiz, with the question what is the most important thing while travelling and picture answers for food and touristy place drawn under it, and the answer screen, Wow! We have found these amazing places for you, listing Yellowstone with mountains drawn beside it and Dubai with a skyline, over a full-width click to see more row, with detailed plan written in the margin
sketch — the quiz, and the answer it gives back
These were drawn to be thrown away, and most of them were. The rough set carried every feature we thought could earn its place, and the recommendations were laid out the way a streaming service lays out films. Low fidelity is where an idea that expensive should die, and this is where one of ours did.
sketch Two more hand-drawn Hodo frames: Welcome back! Let’s do this again — the same question, what’s the most important thing while travelling, with picture answers drawn for nature and shopping — and the result screen, Here is your result!, holding three rows of recommendations each with a thumbnail beside it
sketch — the returning traveller is asked again, and the result screen that answers
Same pen, further into the set. The quiz is also the way back in for somebody who has already taken it, and the result is the screen the personality colours below were designed for.

Five personalities, five colours.

the personality colour system

  • Party Person

    pubs, clubs, live bands, somewhere good to eat after

    #F42254

  • Leisure Lover

    family time, hotels, sightseeing, their own pace

    #85BDFF

  • Active Adventurer

    hikes, treks, trails and adventure sports

    #1FA93D

  • Culture Creature

    museums, festivals, forts and local heritage

    #FFC701

  • Avid All-Rounder

    a bit of everything, inside one trip

    #484689

The rest of the system is ordinary and deliberately so: one type ramp from a 28px heading down to 12px support text, and a four-colour core palette. The part that had to be designed was this: a colour per personality, carried through the result screen and every recommendation after it, so the match is legible before anybody reads a word.

The fidelity where the flow gets argued about.

The Hodo log-in wireframe: an email field, a password field with a show-password eye, a remember-me checkbox and a forgot-password link, a Log in button, a sign-up line, and four social sign-in marks under an OR rule
log in — and the four social sign-ins under the rule
The Hodo quiz wireframe: the heading let's take your travel quiz, the question what's the most important thing while traveling, and four picture answers two across — food, hot spots, culture, nature
the quiz — four picture answers, two across
The Hodo recommendation wireframe: Recommend a trip, New York for four days and four people, with Day 1 and Day 2 each holding four places, a thumbnail and two lines per place, over a home, itinerary and profile tab bar
the recommendation — New York, day by day
The Hodo travel-history wireframe: New York, Jordan and Grand Canyon, each with its dates, and a See More button at the foot of the list
travel history — where they had already been
wireframes — four of the working screens, exactly as they were exported. Working screens, not finished ones. Mid-fidelity is where we could still change our minds cheaply, and where we cut the affordances people didn't use: the app got quieter at every round, and easier at every round.

04 Final screens

The screen we drew, and the screen we shipped.

By the third iteration, the quiz flow and the personality result had finally lined up with each other.

The final design screens.

Eight finished Hodo screens in two rows of four: create account over a blue skyline, the trip-duration question with three answers, the Culture Creature personality result, the Hey Traveler home screen, an Italy destination page listing places, the Salotto42 place page, saved trips, and a profile page
the set as it was published — sign up, the quiz, the result, home, a destination, a place, saved trips, profile
Eight screens, in the order somebody meets them. The personality result is the hinge: everything after it is coloured by which of the five it lands on.

The final design stopped explaining itself and started asking one thing at a time.

The aim was a clean flow with only the right information on each screen and, above everything, a home that matched the personality result to the quiz that produced it. The design flow had changed for the third iteration, so the last pass was about making the two halves agree.

Which is why the finished home screen asks a single question, gives one primary action, and then shows you places rather than filters.

The screens we drew.

The Hodo home wireframe: Welcome Traveler, a destination field with a search button, the line here are some trip suggestions for you based on your personality, nine placeholder tiles named New York, Paris, Italy, Zion, Highway 1, Spain, Brazil, Hawaii and Japan, and a home, itinerary and profile tab bar
home — nine placeholder tiles, and a sentence explaining them
The Hodo menu wireframe: a username with an avatar at the top, then profile, upcoming trips, recommend a trip for me, plan a trip manually and settings, sliding over the home screen
the menu behind it — including plan a trip manually, which testing later cut
wireframes — the home screen and its menu, exactly as they were exported.

And the screen we shipped.

final The finished Hodo home screen: a photograph taken from an aeroplane window behind the greeting Hey Traveler, a destination field holding Italy, a red Recommend button, and a Destinations for you row showing Thailand and Spain
final — one question, one action, and places you can already see
Same screen, two fidelities. The grid of placeholders became two photographs, the explanation became a greeting, and the search field got the only red button on the page.

And the app in motion.

The Hodo prototype running in a phone mockup on the app's purple cloud field: the onboarding screen Take a quiz, with its illustration, the line about finding your traveler personality, and Skip and Next under it
the prototype, recorded — onboarding, sign up, the quiz, home, a place, saved trips, profile — 3:14
The clickable prototype, run end to end in one take. It is the prototype and not a shipped build — the recommendation model never got the runtime it needed — but every screen in it is one of the eight above, and the flow is the one that survived testing.

05 Testing

I put it in their hands and watched it get simpler.

The first flow tested badly. Every round after that removed something.

The design started out complicated, and people told us so straight away.

Testers didn't understand several of the features the first time through, and the flow looked complex before it looked useful. Over further rounds of testing and iteration we narrowed the scope until the app matched the way people actually think about a trip, and the results at the end of that were genuinely good.

Three questions decided every change: what did the testing say, how long would the change take, and could it be built at all. That last one is what killed the most ambitious idea in the project, and the app is better without it.

A user testing session: a participant sitting on a sofa working through the Hodo prototype on her phone, with printed and highlighted test notes beside her
session one — the prototype, the phone, and the notes
A second user testing session: a participant in an armchair reading a Hodo screen on her phone with her hand at her chin
session two — the pause that tells you a screen is doing too much
Watching somebody stop is more useful than any of the things they say afterwards. Most of what we cut, we cut because of a pause.

Every round of testing took something out of the app. That is what finally made it usable.

06 Next steps

What I'd build next.

We proved it for five personalities. The version after this one is twelve.

Having tested the validity of the project for 5 personalities, the future scope is making it work for all 12. The dataset is Italy only, so the same groundwork has to be repeated country by country. The foundation of the recommendation algorithms is laid; the next step is making the content-based and collaborative-filtering approaches work together rather than separately.

And the app recommends, then stops. Whether we keep building it or somebody takes it over from us, the next features are the ones that turn a recommendation into a trip: a geolocation feature that enables route planning, a plan-a-trip flow, and booking.

  • 5 personalities tested → 12
  • Italy dataset → more countries
  • content-based + collaborative filtering
  • geolocation and route planning
  • plan a trip, then book it

A recommendation is only half a trip. Hodo got the half nobody else was solving.

The next one

Bean & Beyond

I designed a coffee-shop app for exploring the menu and customizing every order.

Read the case

Back to the work