Behind the numbers: How we measure things

Tuesday, 2 July 2013

“132,000 tweets per minute,” “9,000 retweets,” “300 replies.” You have probably heard numbers around the huge events that shape Twitter – but what do they really mean? And what is the data that can tell you how engaged your audience is on Twitter?

These are among the main measures we use inside Twitter. At the moment they give us amazing insights into the way we tweet. As we get new methods of counting, we will report them here.

Where do I find the data?
Finding the big numbers is straightforward – you can do it either by: recording the changes in your own Twitter account; you could use our API; or you can work with one of our certified partners - companies such as Topsy, SocialGuide or Mass Relevance. They are part of our “ecosystem” of trusted organizations that can help you get more out of Twitter and help you access data in a wide variety of ways and across a whole load of different metrics. We will be featuring some of these partners soon in a guide to how that ecosystem works.

What is the data that can tell you how engaged your audience is on Twitter? 

What should you look for? This series will showcase the numbers that count – some are easy to get hold of, some more difficult but this is a good place to start. We will be adding more to our “Behind the Numbers” series in the next few weeks - and linking pieces from right here.

Tweets per minute (TPM)
A way to measure the highest number of Tweets reported within any given minute. Often used to record the activity around a specific hashtag or event. Read our post titled “Behind the Numbers: Tweets per minute,” for more on this measurement.

Tweets per second (TPS)
This shows how many people are tweeting in any given second. It is used to look holistically at user activity on Twitter, rather than looking at a smaller dataset related to a particular event or hashtag. Read this post on the Twitter Engineering Blog for more on Tweets per second.

Twitter is about engaging with audiences at a very direct level. Mentions are one of the simplest ways to do this. A mention is any Twitter update that contains “@username” in the body of the Tweet. @replies are also considered mentions.

Follower count
Is this the most crucial metric of all? The number of followers you have can determine how your account is seen by the outside world. It’s visible information – if you keep an eye on the number you can see how it’s changed over time. We will be writing more on this blog about the importance of followers but its not just about the raw number: who are your followers? How engaged are they? Do they retweet your content? Where are they from and which time zone should you tweet in to find them? Are they influential? Your followers are your most valuable asset – the more you have that stay with you over time, the more likely they are to share what you post. If you can work out how many followers your followers have in turn, then that information becomes even more useful.

Every time someone favorites one of your Tweets you can see the figure at the bottom of each Tweet. It tells you how much people like your Tweets and it becomes part of a saved collection of favorites on their profile page.

As more and more people tag their Tweets with a location, you can gain incredible insight into how people tweet in different locations globally. Read more about geotagging here.

Sentiment analysis of Tweets is still being developed by companies around the world. Can you tell what someone is thinking by the way they tweet? Companies think they can and Datasift, for instance, claims to be 70% accurate in its analysis of sentiment. However, sentiment analysis as a whole is at its beginnings. Recognizing when someone is being sarcastic or ironic is something humans often have trouble with. It’s the difference between saying something is “bad” in the traditional way and saying something is “bad” in the Michael Jackson sense. Expecting an algorithm to do it may be asking too much right now. But in the future?

Read more: Tweets per minute

What would you like to see measured? Have you done something interesting with our data? Let us know at [email protected].