This is the first post in a series, in which we will discuss the motivation, methodology and metrics of our recommender project, Find My Ryokan.
motivation
Imagine you go to a grocery store to buy a bottle of red wine. You don’t usually drink, but this is for your best friend’s party. Looking at all the shelves around you, you are baffled by the wild variety of choices that feel all over the place! Why does this bottle from Chile costs $15 while the other one from Washington state costs $60? Even though you don’t drink, you have the impression that those from the Old World should be more prestigous and thus cost more. But why is this bottle from France only $30? And the ratings - what does 95 mean over 89? You begin to feel more and more frustrated and then decide it might be much easier to grab a fancy cake instead. At least you see the toppings, size and design clearly!
This is the same situation when we search for lodgings, when we are also confronted with too much or complicated information. With so many results returned, how are we going to compare apple to apple when we have no way to tell the strength of one lodging versus another? And forget it if we are planning for a honeymoon and want a place that feel a bit more luxurious, secluded and special. We just cannot make an informed decision in a reasonable time.
We are destined to fix this problem in the ryokan (Japanese traditional inn) space with our new recommender, Find My Ryokan, by surfacing lodging specialties like what Pandora is doing to music.
who we are
We are a small team of PM and engineer, with professional background in creating large scale commercial products. We have a combined 40 years of experience traveling in Japan, with 50 ryokan stays in various price buckets across the country. We know what makes/breaks a ryokan stay, and are also proficient in Japanese culture while staying current with travel industry trends.
Like other travelers, we also utilize both Japanese and global travel sites for planning and booking, and face the same info maze pain. This forces us to overspend research time while getting puzzled by the vast number of choices. We are confident that with our background in computer science and application development, we can solve this problem algorithmatically.
what is a ryokan (and why it makes this problem worse)
Japanese ryokans are vastly different from western hotels, which are more standardized offerings in terms of room type, dining choices, amenities, etc. Ryokans, in contrast, can be so different from each other in so many ways, including architectural style, types of rooms and meals offered, history and service level.
A closer analogy to ryokan in western terms is boutique inns situated in historic European villages.
For example, you can quickly grasp the difference of a USD70 motel room versus a US1600 Four Seasons suite in New York City. This is because the differences are very visible, and its price positively correlates with branding, room size, decor luxury, and location. However, things are not so straightforward in the ryokan world. Similarly sized and decorated ryokan rooms can vary from JPY25000-100000/per person/night. Why the huge gap? This is because unlike western hotels, a ryokan consists of so many variables. The cost can go into fine dining (dinners and breakfasts are included in the price), types of onsen (hot spring bath - whether they are public indoor/outdoor, or private in-room), service level (the revered Japanese hospitality), the ryokan’s history and prestige (which unlike Four Seasons or Ritz Carlton, are not known globally), and many more. These factors all contribute to its price in various levels and some can be visible or not.
Consider the following collage, which illustrates the major visible components of 4 different ryokans. Is the cost more influenced by the onsen, food, room or heritage?
photo credit: ryokan official sites
Therefore, if I am a foodie and only care about Wagyu or local seafood, I should know that a candidate ryokan charges a steep price because of its fine dining, not because it offers 7 types of outdoor therapeutic bath with a view of Mt Fuji. A ryokan search thus should return an easy to understand result list with strengths highlighted for my decision making. I can then happily choose a ryokan that fits my need.
Furthermore, even if we only consider a single factor such as room quality and styling, there is a wide spectrum of offerings, as illustrated by the following collage (again showing 4 different ryokans). How can I know by a glance a reasonable price range, without prior expertise in ryokan stays?
photo credit: ryokan official sites
what we want to change
Given the above problem, our goal is to help travelers quickly understand the strengths of each ryokan presented, and pick one that provides the most satisfaction (utility) for your preferences. This comes from our belief that research is tedious and error-prone by human terms, and our time should instead be spent on actually enjoying the item.
This appeals to the human tendency to maximize enjoyment, while minimizing cost (time, travel budget, etc.) and the risk of loss aversion.
our mission
To create happy travelers by giving them the right mix of information, so they can plan a ryokan stay that is memorable and just as they like.
FMR’s value proposition is in helping travelers navigate the cryptic information maze typically associated with travel-related websites, so users can spend less time researching, and make informed and satisfying choices. We achieve this by
- highlight the niche and specialty of each ryokan, so users can understand the different kinds of ryokan offerings, and find the ones they like
- recommend ryokans that align with user preferences
- inform users the criteria/attributes to look for in selecting ryokans
what do current travel booking sites offer?
generic container of a diverse set of lodgings
Even though most travel sites act as a container of various types of lodgings including western hotels, ryokan, hostels and vacation rentals such as cabins, they do not acknowledge the unique set of attributes of each lodging genre. Rather, they tend to serve the majority of their listings - western city hotels, and use a common set of attributes from this genre for recommendation and search criteria across the board. The emphasis on hotel amenities and date/price/location/availability only return irrelevant results for other genres.
For example, instead of informing the viewer that a particular ryokan is housed in a historical building, the attributes surfaced are just “wifi available”, “free parking”, “tv” which are common amenities standard in western hotels and shared by many ryokans. These cannot be used to distinguish one ryokan from another.
In the following screenshot, when searching for a ryokan near Yamanaka Onsen, a popular travel booking site gives me 20+ results across various price points. Doing a comparison gives me absolutely no insight on the huge price discrepancy.
The detail page of the expensive lodging (screenshot below) also gives no hint on where the money goes. In fact, this establishment was established by a decree from the Maeda daimyo (lord) in the Edo period as guardian of the local hot spring. The current owner is the 18th generation. It was also frequented by Kitaoji Rosanjin, a famous sculptor and art critic in the 20th century. You can experience this in the lobby which is lavishly decorated with Rosanjin’s art works. This ryokan is also reowned for its refined Kaga cuisine (a style using local delicacies from the Ishigawa (Kaga in Edo-era) prefecture). These are all “ingredients” that can contribute to the price of a stay.
no meaningful recommendation
Very few travel booking sites have implemented practical recommenders. For those that have, they just recommend based on nearby hotels (geographical) as can be seen from the following screenshot. Nearby items might be vastly different in price, quality offered, food, view, etc. In ryokan particularly, geographical proximity is not a meaningful similarity measure.
Recommender systems play a significant role in improving ecommerce sites’ business, with 35% of Amazon’s sales and 75% of Netflix’s revenue from product recommendations. Why not delight and engage users with personal suggestions on what they prefer, thereby retaining them as long-term users?
what we are and not
We are your guide, to pinpoint things that you will like, shorten your research time and increase the likelihood that you will like something. We only show things you like and hide those we think you will not like. As we highlight the strength of each ryokan, we also curate a set of ryokans around a strength or theme so if you are really into something (e.g., wagyu), you have a one stop shop for all ryokans that specializes in this delicious food!
This is what sets FMR apart from other travel websites.
Since we are laser focused on high quality recommendation and discovery, wee are not going to find you the cheapest place on a certain date and place. That will be a traditional booking site’s job.
our targeted users
Our users are those who cherish traditional Japanese culture. They would like to have a holistic experience of staying in a classic ryokan, while savoring delicate Japanese cuisine and enjoying the beautiful nature outside.
travelers who do not know what a ryokan is
- first time foreign traveler (residing outside Japan), traveling solo or as a couple and want to try a ryokan stay for the first time
- people who are used to staying in standardized western hotels and are confused by the diverse ryokan choices
travelers who have stayed in a ryokan before
- people who have fond memories of a previous ryokan stay and want to try another ryokan similar to that one
- people who know what they are looking for (e.g., outdoor onsen with view of Mt Fuji) but cannot find such information in travel booking site search
FMR curates collections that appeal to our targeted users’ desire for nature, fine food and culture. We also sharpen their research skill and create serendipity for a memorable trip.
why now?
3 reasons why we launch FMR now:
- In this era, people are getting busier and more impatient with information overload. They want a fresh and brisk interaction while making choices.
- Japan is rebounding strongly from the pandemic and the government sets ambitious goals to boost both the number of tourists and spending per capita with massive campaigns. The goal is to make tourism the largest export industry in 2030. This means more information are created to cater to global visitors, creating even more information paralysis.
- For the 3 year inbound strategy, the government plans to redirect foreign travellers from the major trunk (Tokaido seaboard) to other regions. This strategy has two goals: to divert them from the busiest metropolises for better capacity control, and to boost tourism revenue in rural areas which suffer from declining population and reduced tax revenue. As a matter of fact, most ryokans are situated in rural areas. We see FMR to be a small help for this initiative.
conclusion
This cryptic info maze problem prevails in many areas, wherever you face a myriad of choices and cannot decipher qualitative product attributes due to inexperience. Shopping for sake immediately comes to my mind as the flavor profile already carries multiple variables, and terroir, craftsmenship, ingredients all contribute to a very complex product.
In conclusion, whenever a product commands multiple qualitative attributes (i.e., styles, service level, kind of cuisine served), instead of straightforward quantitative measures such as price, yes/no answers like wifi available, free breakfast, accept infants, etc., we need to employ a more thoughtful way to present the information for easy understanding and comparison.
We thus believe that our recommender can be scaled to multiple domains too.