How Washington Post’s data-driven product development engages audiences - Reuters News Agency

How Washington Post’s data-driven product development engages audiences

Several tools focused on design, speed, and audience engagement have helped The Washington Post reach 100 million monthly unique visitors in the United States.

Mar 8, 2018

Shailesh Prakash, CIO and CTO of The Washington Post, kicked off a series of speakers at the Big Data for Media Week [1] conference by sharing some of the Big Data tools that can help digital publishers succeed.

The American publisher currently has 100 million unique visitors per month just within the United States and publishes 1,200 stories per day.

While content is at the heart of the company, The Washington Post recognises that product — especially the design, speed, engagement level — has become its key to success, Prakash told the audience of 200+ participants from news media companies around the world.

Some of the Big Data tools the company has built internally and used are:


The first tool built by The Washington Post, Clavis is a suite of audience targeting technologies The Post has successfully monetised. Clavis automatically analyses everything the media company publishes and arranges them into topics, while at the same time tries to understand who the readers are. It then personalises content and brand messaging. and The Washington Post has seen click-through-rate (CTR) increase significantly through the years.


This tool tries to predict an article’s popularity. It allows editors to prioritise content, identifies under-performing articles, and eventually supports advertising opportunities.


This multi-armed bandit for content variation testing is a dynamic optimisation of variants using real-time user engagement feedback. It takes a combination of headlines and images, experiments on them, and finally exploits the best performing combination by automatically directing the winning variants.


This is a somewhat controversial tool, whereby The Washington Post attempts to automatically generate headlines based on the story content. It can also suggest different headlines for different channels and devices. The three algorithms used are: hedge trimmer, multi-sentence compression, and neural machine translation.


This is an intelligent, automated storytelling agent. It uses Artificial Intelligence to automatically write stories based on structured data and deliver them to specific channels and personalise stories for readers.

The Washington Post has successfully used these tools during the Olympics and the U.S. election. Prakash does not believe these tools will replace journalists in the newsroom; instead, they can allow journalists to focus on investigative pieces as machines take over the more mundane reporting pieces.

Some of the other tools that they use are Tau (an article-scoring tool), Loxodo (a real-time data analysis), Riveting (to understand how riveting a story is), and BreakFast (measuring how successful their alerts are).

The tools presented work across multi-media platforms, be they texts, audio, or video [2] . Prakash emphasised that all these tools rely on relatively new Big Data and cloud technologies — all these simply would not have been possible 10 years ago.

About Serla Rusli

Serla Rusli is a business and financial journalism MA student at City, University of London. She can be reached By Serla Rusli, Student, City University of London

This article was originally published on the INMA conference Blog and does not express the views of Reuters.