How Airbnb, Pinterest, IBM, & the NYSE are cashing in on their own data

SAN FRANCISCO — Capitalizing on your company’s data is no longer optional. It’s expected.

That’s because all types of companies, from startups to tech titans, are generating more revenue in less time by implementing data technologies.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":1476127,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,","session":"B"}']

Big data nerds from Airbnb, Pinterest, IBM, and the New York Stock Exchange spoke today at VentureBeat’s DataBeat conference to tell us how their data practices achieve real, bottom-line results.

Here are their stories:

Airbnb

One of the ways Airbnb uses its data is to shave down its customer support work and make its customers happier. Early on, it noticed that nearly every customer was contacting its customer support center, a trend that was not only an overall bad sign but also something it could not sustain as it sought to scale up.

“We had lost that feedback loop between customer support and product,” said Airbnb head of data science Riley Newman.

So the company defined a problem, set up a clear success metric, and got a small team to turn data from frequent customer queries into product improvements.

“By spending the time up front by making sure the data was clean and reliable, we were able to move much faster later on,” said Newman about one of his biggest takeaways.

Last month, Airbnb closed a massive new funding round of just under $500 million, led by TPG Capital. Last week, the company introduced last-minute bookings for its San Francisco and Los Angeles markets.

New York Stock Exchange

The New York Stock Exchange (NYSE) works with more than 10 petabytes of data. That data spans all kinds of geographies, business areas, and platforms. Data scientists regularly need to access various chunks of those data sets, but that task is more difficult than it sounds.

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":1476127,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,","session":"B"}']

Prior to 2006, the NYSE relied on IT to push that data to the data scientists, but that was burdensome and expensive.

“We started recognizing that we needed to change our data platform model,” Steven Hirsch, the NYSE’s chief data officer, said on-stage at DataBeat.

“The old model just doesn’t work.”

So the NYSE made some changes. It now follows several key rules: No data lives in a single physical place; no single technology is used to store data; and multiple toolsets are used to analyze data.

[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":1476127,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,","session":"B"}']

It’s not easy to load all that data into a relational model. So the NYSE built a tool it calls the “orchestration engine,” which gives data scientists the toolset to be able to move all that data themselves without bothering the IT folks. They can filter by data type and date and push all the data into whatever app or environment in which they want to work.

“A data scientist makes a single request, and all the data magically appears in their environment,” said Hirsch. “It’s all about empowering the end user.”

And at the end of the day, he said “This saved us a ton of money.”

Pinterest

Pinterest’s data scientists are using Amazon Web Services’ RedShift to run interactive analysis on the company’s data instead of using more traditional tools such as Hadoop and Hive.

[aditude-amp id="medium3" targeting='{"env":"staging","page_type":"article","post_id":1476127,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,","session":"B"}']

“It’s way more efficient to interact with data directly instead of building reports every time,” said Pinterest data infrastructure team member Jie Li.

“As a result, a lot more people at Pinterest can look at data,” he added.

Pinterest has been using its data to run A/B tests and various experiments to optimize its site to achieve more growth and engagement. Some of the areas it has been able to improve include its signup page (it found the page didn’t explain Pinterest well enough); its social sharing rates, which it increased by 150 percent; and its non-U.S. user engagement with recommended content.

This last one had a result only data could have yielded: It found that while localizing content works in non-English speaking countries such as Germany and Japan, it doesn’t work as well in English-speaking ones such as Australia. English-based content from the U.S. is just better, and English-speaking users prefer it.

[aditude-amp id="medium4" targeting='{"env":"staging","page_type":"article","post_id":1476127,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,","session":"B"}']

In the last few weeks, Pinterest has not only beefed up its search and discovery offerings with its new Guided Search and Custom Categories features and improved Related Pins, it finally introduced its first monetization source, “promoted pins.” It also confirmed last week it is raising a new round of funding in the amount of $200 million.

IBM

IBM’s Watson supercomputer, famous for its Jeopardy winning streak, may soon revolutionize your health care.

Rob High, chief technology officer of IBM’s Watson Solutions group, explained how Watson simulates human thought and behavior by embracing probabilistic models rather than strict rules (like most computer programs use). That has profound ramifications for an industry like health care, in which doctors and nurses have far too much to read with too little time and context.

In determining treatment options, for example, Watson can scan through your health care history and compare it to millions of other patients’ to find the dozen that almost exactly resemble you across all sorts of variables. Then, it can propose treatment options that might seem dangerous or unexpected at first glance but entirely reasonable when you consider mountains of patient data. Humans just can’t do that, not by themselves.

[aditude-amp id="medium5" targeting='{"env":"staging","page_type":"article","post_id":1476127,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,","session":"B"}']

It’s still early, of course, but this is a huge business opportunity for IBM — and a profound social opportunity for humankind, whether this tech comes from IBM or a competitor.

“This is the future of cognitive computing,” said High.