Twitter analytics – @ONS over the last 9 months

(By Lauren Bradford, Social Media Manager)

Twitter Analytics was updated last month allowing the export of data back to October 2013 (or more specifically the last 3,200 tweets posted since then). I exported all our data from October 2013-July 2014 and looked at the highlights for impressions, retweets, url clicks, embedded media clicks and total engagement which show some interesting patterns. Before I get to that, here’s a bit of background.

The social media team work to three different deadlines which reflect when people interact most with our content.

  1. 9.30am is when all of our releases are published, so our followers expect content from us at this time and it gets shared widely as it’s breaking news.
  2. There’s a peak around lunchtime and we find content that works best here are interactive data visualisations and infographics, giving people opportunity to explore and play with the data.
  3. The third peak is around 5.30pm which catches the commuters.



When we look at impressions, i.e. the number of times a tweet is served, our high profile headline releases come up top, unsurprising as these are breaking figures and shared immediately by the big news agencies and journalists with large followings. The breaking GDP preliminary estimates tweets take up five of the top eight posts, with the highest impression figure of 231,702 for our post in April.

We supported ONS’s appearance on Steve Wright’s BBC Radio 2 show in February, and our tweet on how many babies were named Sherlock has the second highest impressions at 175,925, helped out by our friends at QI retweeting.


We published an ad hoc article in January which looked at falling real wages, a hugely topical subject.


Behind the GDP preliminary estimate and our Sherlock posts above this was our next most popular. After that are stills of interactive maps which always work well for us.


Another popular post looked at UK healthcare spending compared to G7 countries with a combined total of 271 retweets.


URL clicks

In November we lifted a map from our Internet Access release of where people had never used the internet.

Although a relatively low key release the url in the tweet to the release had 1,357 clickthroughs. The next highest clickthroughs were for a suite of 12 regional profiles we released in October.

An article on graduates in the labour market had four tweets of the top 20 most url clicks.

Some of the tweets with high url click figures are for lower profile releases which shows the interest is there. An interactive wheel we published in November looked at which sectors made up the UK-Non Financial Business Economy, and the post had 539 clickthroughs.


Embedded media clicks

When we look at our tweets with the most embedded media clicks (how many times a photo, video or GIF is clicked within a tweet) it tells a slightly different story, as the posts were tweeted alongside external events rather than scheduled release days. Eight out of the top 20 embedded media clicks were for #MindTheGap.

The rest of this top 20 is made up of tweets asking questions linking to graphics and interactives to explore.



Total engagements

Total engagements looks at the total number of times someone has interacted with a tweet. This includes all clicks anywhere on the tweet including hashtags, links, avatar, username, and tweet expansion, retweets, replies, follows, and favourites. When you combine all these together the top ones were

  • Sherlock with 7,581 engagements,
  • Overall engagement for all #MindTheGap posts with 6,235 engagements,
  • GDP preliminary estimate with 3,275 engagements,
  • The map exploring the average earnings where you live with 2,727 engagements,
  • Workday migration map with 2,700 engagements
  • Map of where people have never used the internet with 2,046 engagements.


What can we learn from all this?

It’s clear that people generally interact most with things that drill down and compare geographies, and have an element of interaction. Most, but not all of the top posts include an image. Some include static maps, some are headlines and facts, and some ask questions letting you explore the data through interactives and infographics. All this reinforces work we’d done on optimising social media content which concludes the polish you add to social media content can help its performance, it’s the content itself that will determine if it does well. That and a good amount of luck that the right people see it.

And it’s interesting that republished content that ties in with external events appear frequently, showing that although our statistics are talked about when they’re newly released, there is a long shelf life for them and they can be even more popular when repackaged to tie in with relevant and topical events.

Since I exported the data up to the end of July there have already been several posts that would completely change the top 20 rankings.


And with our annual baby names release coming out on Friday 15th August and our second feature on Radio 2 the following Monday 18th, all the above will change I’m sure! We’ll have a look at Facebook analytics as well to see if things are the same there.