Building a new fast web platform and launching it quickly to 30 news websites

Last week Trinity Mirror Digital launched our largest website, Mirror.co.uk, on our new fast responsive platform. Since early October we have now rolled this out to 20 websites, delivering pages that load in less than half the time to millions of users of some of the UK’s largest news products.

Speed matters for everyone who uses products like ours. Audiences won’t wait for slow pages to load, advertisers want quick results from their investments, and publishers like Trinity Mirror need to move fast when changing and improving digital products.

I’d like to share something about how we have achieved such rapid results for users, advertisers and our business, and delivered a hugely ambitious, transformative project very successfully.

There were lots of reasons to put this project off, and we didn’t start until we had squeezed the maximum value out of the old platform via A/B testing. When we did start, we set clear goals: faster page speed, better viewability of adverts and “speed to rollout”, meaning the ability to change all of our sites with single releases, not repeated changes. This ruthless focus on 3 things that really mattered contributed a lot to our success and made it easier to prioritise the many possible changes.

A test-and-learn approach was also critical, and we began as we meant to go on: small and lean. Late last year we launched a fully responsive prototype site in Birmingham. It showed positive results and built confidence.

We returned after Christmas resolved to change every one of our sites by the end of 2016. It was a big job. We were tackling 30 sites across a wide range of team and audience sizes, serving more than 125 million users each month, and bringing in a lot of new approaches.

This included organising our product teams to work together for the whole site, but with each team focusing on distinct and specific goals and features. Constant communication and cross-team prioritisation was essential.

We had to compromise on scope. We really wanted to rebuild our entire publishing chain to maximise the speed at which we deliver the news, but that would have added risk and delayed delivery, so we focused on the end product that users and advertisers would actually notice.

By May we had basic pages of content ready to test with a randomly selected 1% of the audience on two sites. Caution was necessary at this stage, because we didn’t want to risk existing revenue growth. The results and data from these tests gave us further confidence and in August, we started using Get Reading to launch more pages to 100% of its users.

At this point the exercise became just as much about working with our business, and particularly the teams focused on revenue, as it was about things like Javascript libraries. The minimal viable product (MVP) was ready, but would it deliver and protect the digital revenue growth we already had? And then accelerate it?

There is enough in the excellent collaboration with our ad ops and sales teams to create another blog post, but suffice to say it is a great example of the collaborative culture I believe every business now needs in the midst of digital disruption.

In September we put together a roadmap to get to the real MVP, and delivered a rallying cry to the teams and the business that we believed we could do “30 sites in 30 working days”. Achieving that goal would transform our major products and deliver the new experience to more than 80% of the group’s audience in time for Christmas — which starts in mid-November when selling and trafficking digital ads, and early December in terms of freezing releases.

This meant building a production line at the same time as we added the final critical features. And that in itself is another good subject for a different blog post the team should write, not me.

As I write, we have more or less made it. We still have 10 small community sites to do in the next few days, but we’ve established a well-honed ‘industrial’ process and I’m confident we will have our 30 sites by 30 November.

It’s been a big success and we have learnt a lot, and I expect my colleagues will share more of the details here shortly.