Nara aims to Pandorify travel with its Big Data discovery engine
Nara’s premise is that if you tell it a hotel you have enjoyed, it can to ply you with similar hotels it thinks you may like to stay at during an upcoming trip.
To create its database about hotels, it analyzes millions of reviews published on the Web. It then uses Big Data analytics –meaning, studying the choice patterns of other consumers and matching against past events in large datasets to make predictions.
The startup, founded in Cambridge, Massachusetts in 2010, revealed its first product in summer 2012 when it began recommending restaurants.
In 2012, the company completed a $7 million Series A round, giving support to its development of an artificial intelligence technology that recommends US restaurants, hotels, and other lifestyle services via Web and mobile apps for consumers.
In August 2013, it launched a “minimum viable product” for suggesting hotels in 50 US and Canadian cities.
In the first quarter of 2014, it plans to expand to the UK.
Other verticals it’s considering entering is tours and activities.
Since August, its hotel product has seen traffic growth “at a nice clip”, with “healthy return rates” and “a declining bounce rate”, according to a phone interview with CEO Tom Copeman, former co-founder of Lululemon, a yoga accessories seller that has been a consumer success story.
Copeman headquartered his company in Boston to be close to the local expertise in machine learning, artificial intelligence, and deep Web analytics. He tapped Nathan Wilson, a former MIT research scientist, to become chief technology officer.
He also hired Kristin Pados, a longtime product manager at TripAdvisor, to be a general manager for travel.
To help with customer acquisition and business development, Nara in November announced its first B2B customer, Singapore Telecommunications (SingTel), which will draw on Nara to support an online food and dining guides in Malaysia, Australia, and Singapore.
Similar B2B deals could be in the offing with other customers for its hotel service, the startup hinted.
The pace of product development will is increasing steadily. Says Kristin Pados:
“When I was at TripAdvisor, we released changes to product every single week.
We are implementing a similarly iterative approach, rather than enormous releases.
We take a very data-driven approach to development and shipping product.”
Whether the hotel results that Nara provides today are relevant lies in the eye of the beholder.
Nara is still experimenting with how many times a user needs to click “thumbs-up” and “thumbs down” on various hotels before it can start delivering results a user will consider relevant.
When it comes to restaurants, it claims it can deliver relevant predictions after a user has rated about seven eateries.
Nara’s system claims to be able to handle when people vary in hotel taste by situation, mood, and modality. One traveler, for instance, might tend to prefer boutique properties for business trips but like to unwind at a low-key surfing hotel on the weekend.
This reporter’s anecdotal test of the service didn’t wow him.
Searching for hotels in Washington DC, I gave a positive thumbs-up rating to a couple of Kimpton chain hotels. Unsurprisingly, that meant that other Kimpton properties popped up as recommendations.
Giving a thumbs up to an inn led to recommendations of other inns.
I could have done that kind of sorting on my own, using existing tools at my favorite travel websites.
Each hotel suggestion on Nara has a “Why was this recommended button.” In my case, I saw this message:
You like places around Downtown. You like Comfort hotels. People in Washington love it.
When I started a search for hotel stays in Montreal, Nara seemed to apply the same criteria in filtering 25 top selections. The list of hotels didn’t differ much from TripAdvisor’s top ranked properties for the Canadian metropolis. Again, not so suprising.
Yet the type of hotel I prefer when visiting DC differs from the type I like in Montreal or elsewhere. My tastes are dependent on the type of properties available locally, their relatively expensiveness, whether I’m traveling for leisure or business, and other factors.
None of these factors seemed adequately accounted for.
The executive team maintains that their service works for the first user the first time they use it. Yet Nara still has some machine learning to do.
Just as Pandora, the streaming Internet radio service, has seen its reputation for accurate predictions improve over time, Nara could also get better as it learns from years’ worth of consumer behavior on its site.
If you’re curious to learn more, here’s a promotional video explaining Nara:
Sean O’Neill had roles as a reporter and editor-in-chief at Tnooz between July 2012 and January 2017.