By Dennis Clemente
NEW YORK–Joseph Essas of Open Table, the world’s leading provider of online restaurant reservations, opened the talk at the Data Driven last June 16 at Bloomberg’s offices. It was Data Driven’s last monthly meetup as it takes a well-deserved two-month summer break.
Open Table has become a household word everywhere. It has over 32,000 restaurants worldwide, with more than 760 million diners seated since 1998. It has represented more than $30-billion spent at partner restaurants.
Today, Open Table seats over 16 million diners every month.
“Our product is real time availability,” Essas said.
Essas explains Open Table’s success. It’s about understanding the diner… building a profile of you as diner from explicit and implicit signals, information you have provided, reviews you have written, places you have dined….”
Essas says ratings are very important. Open Table has generated 30 million reviews.
The basic data ingredients for Open Table: Diner-restaurant interactions, restaurant metadata (what kind of price range /hours/topics), user metadata and user metadata (user profile). Reviews are reportedly rich and verified, and come in all shapes and sizes.
“Our system sits inside a restaurant,” he said, guaranteeing diners’ presence in a restaurant.
When figuring out trends, it uses dish tags and a bit of linear algebra to easily detect what dish is trending. Right now, artichokes are reportedly trending.
Open Table also creates diner profiles. For sentiment, it uses ratings as labels for positive and negative sentiments. “People used to be very generic. Now they (may ask for a) waiter with a ponytail,” he said.
“Our job is to optimize restaurant business as much as we can,” he concluded.
David Guleck spoke next about Bonobos. Founded in 2007, it is reportedly the largest US apparel company originating from ecommerce. Its data science and engineering team was founded in 2012.
This mobile-only enabled site gets results from automation and informed strategic decisions.
Its data team focuses on 3 goals in its structure: data engineering (data acquisition and data tools); bilI/reporting (democratization, self-service) and data science (deep analysis and predictive algorithms).
Bonobos’ learnings include use of email (relevant content matters,frequent optimization is hard) . For the latter, you ask yourself, how many emails should you be sending?
Other key learnings rested on recommendations (algorithms are the easy part, inputs and outputs are lots of work) and promotions.
“We’d like to do product similarity scores, but we’re not there yet,” he said.
Another presenter, Cockroach DB talked about its scalable, survivable, consistent and open source offering. Started last February, it received $6.26M series A funding.
Its goal is to make data easy and grow to any scale –horizontal scaling, commodity hardware. One way it thinks it can solve problems are by making apps agnostic
It has a layered architecture, monolithic sorted map, distributed transactions, RocksDB for storage, and raft for consistency.
Last speaker was Gideon Mann of Bloomberg, showing the sentiment analysis the company sells to Wall Street.