How We Turned a Bespoke Bot into ‘Business as Usual’

BBC News Labs
BBC News Labs
Published in
4 min readJun 8, 2017

--

By Labber Rachel

Blimey, we’ve won an award, the 2017 Online Media Award for ‘Technical Innovation’.

We’re obviously a little bit proud when Labs’ innovation is publicly recognised, but in some ways we were also a little surprised. To be honest, we have moved on to other projects, and the bot has moved on as well. What started as a one-off project for the EU Referendum was repurposed for the U.S. election, and then handed off to another team within the BBC for use for tomorrow’s general election.

Instead of reflecting on the Twitter bot’s success, I thought this might be a good chance to talk a little bit about the journey of a successful prototype, and in the case of the bot, about a successful transfer to business as usual.

E.U. Referendum-Bot => U.S. Election Bot

Back in 2016, Labber Jacqui Maher wondered if we could re-use the EU referendum result graphics that were being produced for broadcast TV for online audiences. She wrote a quick proof-of-concept script that:

  1. Reformatted the live TV graphics for Twitter
  2. Composed a tweet by agreeing a filename format with the graphics team
  3. Posted the tweet to Twitter

For more on the technical details on the Twitter bot, please see Jacqui’s post on our blog.

The referendum bot wasn’t just a prototype of a tool — it was also a prototype of co-ordination, communication and process. News Labs worked together with the BBC’s graphics and Visual Journalism teams to join up several production systems, define an appropriate BBC presence on Twitter, and co-ordinate rehearsals before the big night.

The two different graphics cards generated by our U.S. election bot.

That meant that by the time the U.S. election rolled around a few months later, we had already built a relationship with the right people and defined a well-running process. When we were asked to iterate the bot for this vote, we focused on generalising the referendum-bot and extending its capability. Instead of just posting constituency-level graphics, we updated our code so that it could generate visuals showing the number of points each candidate had after states announced their projections as well. We also generated both English- and Spanish-language graphics, for posting on our World Service account.

U.S. Election Bot => ‘BAU’ Bot

When the US election passed successfully, we’d proven that this prototype was well-received and useful enough to be used regularly by the BBC’s Visual Journalism team. It was ready to fly the messy Labs nest into the blue sky of ‘Business as Usual’.

For the 2017 general election this year the bot has been lovingly handed over to an engineer in the Visual Journalism team. This team will modify the code from now on and develop it further for ongoing use.

Because the results bot is still a prototype in essence, it needs a lot of understanding of the scripts, the data structure and process of preparing the system for a new election. The Visual Journalism team have future plans to improve the user interface to prepare for an election or event, and allow the system to be started and monitored by a journalist rather than a developer.

Prototypes are sometimes not just about the technology, but about building relationships with people outside our team.

But even if Visual Journalism chooses to start again afresh, we are most proud that the results-bot has done what a good prototype should do: “test a concept or process or to act as a thing to be replicated or learned from”.

We find transferring experiments out of Labs and into production challenging at times. The American Press Institute recently published a study on newsroom innovation and found that this is true for other Labs-like teams, who build tools and services in isolation but don’t always have the resources for maintaining them as products or services.

Our Twitter bot is a success in our books because we’ve been able to collaboratively manage the transfer. We’ve demonstrated that prototypes are sometimes not just about the technology, but about building relationships with people outside our team. Our thanks again go to Paul Sargeant in the Visual Journalism team who acted as our “people and process” prototyper extraordinaire.

You can follow the 2017 General Election results at @BBCElection.

--

--

BBC News Labs
BBC News Labs

Driving innovation for BBC News. Powered by @BBC_Connected.