Using data to build a better business – advice from LinkedIn

Lutz Finger, director data science/data engineering at LinkedIn, spoke about getting your data right at the Digital Innovators' Summit in Berlin today

Lutz Finger, director data science/data engineering at LinkedIn (© Ole Bader/Sandwichpicker)

Finger began by showing an analysis of attendees’ LinkedIn profiles, and showing how connected they are to each other. "People who have more connections are our ‘hubs’," he said. "They are the key to knowledge sharing and growth of the platform."

Getting your data right

In terms of getting your data right, Finger shared the example of Alta Vista – an old search engine claiming to have data on "all of the websites on the internet". Finger said it took 11 months for them to get wiped out. And why? Because although Google could not compete with Alta Vista’s claim (AV had more resources and data centres), it asked the right questions, according to Finger. "Google looked at a sub-set and data websites and figured out what are the most important sites for you. This led to a different, relevant metric. Alta Vista could have done this, but didn’t ask the right question – it’s not rocket science."

Finger said that forming the right question to achieve the right "value function" can be extremely difficult in a set system, for example in media companies who have a lot of legacy.

"When you form a question you should know the value function," said Finger. "What value will the answer bring? If it’s not good enough, revisit the question. Ask the right question, measure the right data, take actions and learn from it."

The world is full of wrong metrics, according to Finger – one of which is measuring success of a particular story via the number of retweets it receives. He said that research shows many people retweet based on headline alone, and if they’d read the whole article, it may not have led to the same action. "We always need to back-test our assumption," said Finger.

Keys to data success

  1. Make your data accessible – the ideal editor was once one who understood readership and we let them please our audiences alone. All that knowledge is decentralised. You need to make that data accessible and bring it all together.
  2. Build a content profile for each reader –try to figure out who your reader is. Understand the activity of their mouse and how they move over the screen, so you get an understanding of what and how far he is reading and what he likes/doesn’t like.
  3. Build a knowledge profile from your journalists – what is this journalist good at – you need a database to figure this out. When something is happening, you need to know who is the right person – a subject matter expert.
  4. Hire a data team – also known as data scientists.
  5. Envision data products distribution networks – this is nothing new for the media industry. Always think about how to build it. Companies like Recode built theirs quickly and did this with data. They analysed how to reach channels and with which medium – a data driven exercise.
  6. Ad placement – if you go to an RTB engine, you have 25 milliseconds to decide which ad to place – the media industry is already doing this and we need to figure out how to monetise it.
  7. Content recommendation – always fails utterly when there is not enough information on a single reader. We need to ask a lot of questions of the reader. The media industry has that profile - Axel Springer and Burda are doing it well and can build good content recommendations.

Story by Amy Duffin, FIPP


Photo of Alexander  von Reibnitz

Print und Digitale Medien

Alexander von Reibnitz
Tel.  +49 (30) 72 62 98-152
Photo of Lutz  Drüge

Print und Digitale Medien

Lutz Drüge
Tel.  +49 (30) 72 62 98-187
Photo of Lolita Lara Dueck

Print und Digitale Medien

Lolita Lara Dueck
Tel.  +49 (30) - 72 62 98 – 123
Photo of Anja  Mumm


Anja Mumm
Tel.  +49 (30) - 72 62 98 - 152
Photo of Gabriela  Wiencek


Gabriela Wiencek
Tel.  +49 (30) - 72 62 98 - 111
Fax:  +49 (30) - 72 62 98 - 112

Social Bookmarking