LinkedIn had one of the first data science teams. Now it’s breaking up the band

MOUNTAIN VIEW, Calif. — LinkedIn has divvied up its respected data science team, which created features like skills, “people you may know,” “people also viewed,” and “who’s viewed your profile,” VentureBeat has learned.

The shift, which occurred around five months ago, split up a team that had for years worked together in the company’s product division, sources familiar with the shake-up told VentureBeat.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":1581120,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,dev,enterprise,","session":"B"}']

Over the past few years, the data science team contained two subsections: the product data science team, where people came up with innovative new data-powered features that give off new data for analysis, and the decision sciences team, which analyzed key product metrics and explored usage data.

Now the social networking company has divided the subsections and stuck them in separate departments. The decision sciences team now reports to the office of the company’s chief financial officer, while the product data science team is now part of engineering.

“It was just really clear that there was just a lack of clarity around rules and responsibilities, which was frustrating for people on the team, as well as people who had to work with them,” said Laura Dholakia, the head of LinkedIn’s business operations and analytics team and leader of the former decision sciences team, in an interview with VentureBeat.

The company’s data scientists — numbering close to 150 now — had the option to go with one group or the other. But for the most part, their day-to-day work consists of the same types of tasks.

If anything, the reorg pairs up the analytics people, who focus on paid products like recruiting tools, with the data scientists, who look into the ways people use LinkedIn’s free “consumer” service for connecting with others. As for the product data scientists, working with the engineering staff reduces the potential for redundancy.

What’s more, the dispersal of the data science teams reflects the importance of analytics as a standalone process as the company matures.

Because LinkedIn employees need high-quality information on subjects like engagement, user growth, and revenue sources, it’s probably understandable that analytics shouldn’t be one more task for product-making data scientists.

“A lot of individuals depend on that business intelligence to be done super well,” Igor Perisic, vice president of engineering and the leader of the former product data science group, told VentureBeat. “I need to drill into lots of segments to figure out how a product is doing. That’s a full-time job. It’s not a part-time job. Now you need to choose between these two things.”

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":1581120,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,dev,enterprise,","session":"B"}']

Despite the advantages Perisic and Dholakia spoke of, several data scientists have left the company in the wake of the changes.

While this division doesn’t signal the end of some sort of golden age for LinkedIn or for data science, the move does carry a touch of symbolism: One of the first data science groups in the world no longer stands out on the org chart as a standalone unit. Instead, data scientists there now play a more integrated role than ever before.

“It’s become more of a role at LinkedIn instead of a team,” one source told VentureBeat.

Changes through the years make sense

The recent changes follow years of adjustments to what data scientists do at the company.

[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":1581120,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,dev,enterprise,","session":"B"}']

Under the leadership of DJ Patil — one of those credited with coining the term “data scientist” and LinkedIn’s former head of data products — data scientists came up with entire applications and features. It was an era of experimentation with the feel of a startup.

But after Patil left in 2011, collaboration across divisions became increasingly common.

Along the way, a few high-profile data scientists at the company left. Last year Monica Rogati went to wearable-device company Jawbone, and Pete Skomoroch became an advisor at venture capital firm Data Collective. This year, Joseph Adler left to become director of product management at data analytics startup Interana.

Then came the official bifurcation of the decision sciences and product data science groups in the middle of this year.

[aditude-amp id="medium3" targeting='{"env":"staging","page_type":"article","post_id":1581120,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,dev,enterprise,","session":"B"}']

Several people who know about the changes believe they make sense. It does seem reasonable to integrate people who think up products with people who build, maintain, and change LinkedIn’s code base, while pulling together the analyst types into one department, where they can then explore data for sales, marketing, human resources, finance, and other purposes.

But not everyone’s happy

Even so, data scientists Ahmet Bugdayci, Vitaly Gordon, Ali Imam, Gloria Lau, Leah McGuire, Patrick Philips, and Satpreet Singh have all exited since the rearrangement. So has Deep Nishar, who headed up product and user experience. (Nishar supported the reorganization.)

It could be that some of the data scientists preferred working closely with and learning from many other data scientists or that they simply didn’t like the work they were tasked with in the new order.

Maybe some people trained in statistics and domain-specific knowledge didn’t want to spend even a little bit of their time generating dashboards for executives or running A/B tests. In the past, data science leaders succeeded in recruiting new talent by promising a sort of oasis — a team of data-first product creation, a source told VentureBeat. The current structure might not lived up to that promise for some of the data scientists.

[aditude-amp id="medium4" targeting='{"env":"staging","page_type":"article","post_id":1581120,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,dev,enterprise,","session":"B"}']

But the reorg did aim to play to people’s strengths. The idea was “to create a structure that allows people to specialize in doing what they’re good at,” Dholakia said.

Then again, the turnover could merely be a function of the company growing larger in the past three years.

LinkedIn had 990 employees at the end of 2010, a few months before its IPO. On June 30 of this year, the headcount sat at 5,758. Many characteristics evolve when a company undergoes that kind of expansion; a typical employee doesn’t have the sway he or she would have at an early stage startup.

And some people who’ve seen the company grow were already planning to leave.

[aditude-amp id="medium5" targeting='{"env":"staging","page_type":"article","post_id":1581120,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,dev,enterprise,","session":"B"}']

“Some of it also overlaps with the tour of duty of four years,” Perisic said.

The reorg, with its implications for supervisors and responsibilities, might also just have presented the right circumstances for some data scientists to leave.

“For most people, it’s a natural time to start looking,” a source told VentureBeat.

LinkedIn’s website shows that the company is now looking to hire new data scientists for a variety of purposes.

To learn about how how data scientists currently work under the new arrangement, and how it compares with the structures in place at other companies with many data scientists on staff, see our follow-up to this article.