When we first started the COVID-19 Dashboard back in March 2020, it was just a map and a few charts reporting four metrics on cases and deaths.
How that has changed in a year. We are now publishing nearly 200 metrics at 4pm each day – which adds up to about 40 million individual figures – as hundreds of thousands of users across the country refresh their browsers in anticipation.
Alongside the London Stock Exchange, the value of the pound and the weather forecast, PHE’s COVID dashboard is the most important and reliable daily statistical publication in the UK.
Evolution of the dashboard from March 2020 to October 2020
In the early stages of the pandemic, we did much of the work manually. For what felt like a very long time indeed, it was just three of us.
But with the sheer volume of data involved now, we have grown into a team of 18 and built a (largely) automated data pipeline that can handle all the different formats and reporting schedules.
Sadly, that doesn’t mean the pressure is off. Delays or changes to any one of the datasets at any stage can potentially set back the whole UK daily update.
A small snapshot of the pipeline for daily deaths data
The daily publication schedule is unrelenting, and the pressure is high. But what has kept us going through nights, weekends, holidays and three lockdowns has been our close relationship with the dashboard’s users.
Each week there are about 19 million page views across the site, many of them postcode searches on the interactive map. That’s more than any page on the Government’s GOV.UK portal since it was created.
During the peak of the third lockdown there were a record 76.5 million hits in just 24 hours.
We are proud that the latest data are always available to members of the public at the same time they are seen in Downing Street.
And we’ve been constantly adding new features to make it easier for people to understand the data most relevant to their communities, and to the policies that affect their lives, in response to user feedback.
We’ve also been working with individuals and groups who may have greater difficulty using the site to make it as accessible as possible, including providing easy-read formats, enabling screen readers, and building appropriate colour palettes.
All the data we use (and more) are available for download, and for professional users we have built a suite of Application Programming Interfaces (APIs) to make the data as open and reusable as possible.
This seems to be paying off. Tens of thousands of people have contributed to dashboard surveys over the last six months.
Results from our most recent survey showed an outstandingly high level of user trust in the dashboard data at 90.3%, up from an already high level in November when 73.4% of users responded that they trusted the data. But it’s a remarkable turnaround since the summer, when less than half said they had confidence in the statistics they were seeing.
Do you trust the data that you see on the COVID-19 Dashboard?
We already know what our users say they like the most (they are most interested in data about vaccinations and cases and now less interested in deaths and testing). And we know what they don’t like – delays, changing definitions and big corrections – all of which we try hard to avoid.
The team are trying to get an even better understanding of this feedback by using novel techniques such as text classification and clustering to pick out key themes from free text comments so we can make sure they are factored into our design process too.
The dashboard has been a major contribution to transparency in the government’s approach to the pandemic, and we know from our surveys it has become an essential public good for many.
So, as long as the data are needed, we will keep publishing every day – and keep listening to our users to inform further expansion and improvement.
View original article
Contributor: Julian Flowers