Twitter recently extended the Gift of Good Data to every registered user. And that data should serve as a great big eye opener.
The good news? You no longer have to estimate Reach.
The bad news? Your tweets aren’t reaching nearly as far as you thought.
The twist? Unlike Facebook, on Twitter, a lower reach is a good thing.
Calculating Tweet Reach Has Always Been a Dark Art
Reach is a term used to explain how many people were exposed to a piece of media – be it a Facebook post or a random tweet. This metric is similar to Unique Visitors on a website and is important when calculating conversions or effectiveness for a specific piece of content.
Until this week, Twitter users could only accurately determine Reach for promoted tweets. Regular tweets went without metrics leaving users with accurate means to show how effective a tweet was, and no way to predict the Reach for a future tweet. The best we could do was rely on two methods of estimating Reach.
The first Reach estimation method is called Potential Reach. This method is based on the understanding that Actual Reach exists between 0 and 100% of a user’s Follower count. However, since a more accurate number doesn’t exist, Potential Reach merely provides that last accurate piece of data – the total number of people who could have potentially seen a tweet.
That is, if you have 1000 followers, this method would put a tweet’s Potential Reach at 1000. If a user with 500 followers happens to retweet you, that tweet’s Potential Reach jumps to 1500.
Twitter’s savvier users can see how Potential Reach might sound great, but doesn’t account for how real people use Twitter. That is, Twitter users skim, dipping into Twitter Streams every now and again, for extremely brief periods. The majority of users have little interest in consuming the Twitter firehose, and after a user follows around 150 people, that type of use becomes impossible.
That’s why Recurve has always used the second method, Probable Reach. Through a couple of assumptions, we put together a Reach Chart  that accounts for things like follower habits and geographic appeal. The end result is that the more followers a person has, the less likely a tweet is to be seen. Our best case assumption was 16% of Followers would see an average tweet.
With either method, users were dealing with an estimate. Worse, users were dealing with an estimate that how no means to verify how accurate that estimate was.
Reach Estimates and Engagement Rates
While each of these estimation methods are fine on their own, they run into trouble when you try to gain actionable insight from them.
In particular, the primary insight most users want is Engagement Rate – the percent of people who saw a tweet and then did something. This number allows you to compare the same piece of content across channels, dialing on which channel actually performs the best for you, and thus which channel deserves most of your effort.
When trying to calculate Engagement Rate, the big numbers of Potential Reach turn into a big negative. As you can see below, 1 engagement on a Potential Reach of 1500 returns an Engagement Rate that is abysmally low. Even based on a Probable Reach of 240, that tweet still performed in the realm of a good banner ad.
And that’s the problem with Estimated Reach – be it television views, magazine subscribers, or billboard impressions – it leads to data that is automatically suspect. When creating a rate from an estimate, it is damn near impossible to make an informed decision.
Goodbye Reach Estimates, Hello Impressions
And, as mentioned in the intro, Twitter has destroyed the need to estimate Reach. Instead, the social network has added the Impressions metric right into their analytics tool. Twitter defines Impressions as:
Number of times users saw the Tweet on Twitter
And Twitter now gives you this number. This means that any user can now tell exactly how many people saw a tweet. No estimates needed.
And this also means that those previous estimates are grossly low. At least according to hacking through my personal data.
To put this data in perspective: as of yesterday, I had 3225 followers. 72% are of my followers are in the United States, and about half are in my timezone. I am active on Twitter, regularly engaging, and average close to 600 tweets per month.
Under previous methods, my Potential Reach for a tweet is 3225 and my Probable Reach for a tweet is 516.
The average reach based not on estimates, but real data? 165
And that number is a bit high, thanks to a few outliers. A better average is 106.
Based on my personal data set of 3006 tweets, I’ve started calculating a new metric Averaged Impression Estimate (AIE). AIE is the Average Impressions of the middle 69.2% of Tweets (based on ordered performance) divided by the Average Follower Count from when those tweets were posted.
While that might sound overly arduous, this metric provides a good rule of thumb for predicting how much attention a tweet should get based on Follower Count, must the same way Probable Reach did.
And my Average Impression Estimate? 3.61% of Follower Count
That number is, of course, way below Recurve’s upper bound on Probable Reach. And it’s drastically smaller than Potential Reach. But, it does provide a good, conservative estimate of a tweet’s reach before the tweet is sent.
Average Impression Estimates Are Lower Than Expected…And That’s a Good Thing?
It kinda stings that an average achieves a very low Impression count. But there are two very important upsides.
1. Unlike Facebook, Twitter is not hampering Organic Reach
As noted by many (Recurve included), the content of Facebook Pages must outperform the mighty EdgeRank algorithm in order to reach end users . This algorithm artificially restricts what users can see, which artificially limits Reach. In order to gain Reach on Facebook, user’s need to engage with a post. The more people who engage, the more impressions a piece of content gets.
However, there is only weak correlation between Engagement and Impressions on twitter. That is, only engagements that increase Reach (Retweets and Mentions) have an impact on Impressions. This means that, on Twitter, Organic Reach is determined not by the whims of Twitter, but by the availability of the users.
Or to put it another way, Twitter is Impression Neutral. All Tweets are delivered to all users in the order they were created, and it’s up to the user to read them.
2. Actual Impressions Mean Better Engagement Rates
How many people a Tweet reaches is arguably less important than how many people interact with that Tweet. And having actual Impressions allow users to calculate actual engagement rates. Or you would have to, if Twitter didn’t also do this for you.
Accurate Engagement rates are necessary to know what works, and what doesn’t, for your customers. And Twitter has given you those. In fact, they even do the math for you. Also, bear in mind that Twitter tracks everything, and that everything from Profile Clicks through Favorites, is reflected in a Tweet’s Engagement metric.
Better Data Makes For Better Decisions
The new Twitter Analytics is a case of not just more data, but better data. Twitter Analytics now provides actionable insights. You can, with a large degree of accuracy, determine your own Average Impressions, and predict future ones. You now know what Tweets are getting engagement, and how your Followers are engaging with them. And you can pull patterns from them. And make decisions based on data, not estimated data.
And if you want, you can strip out the outliers and get a little a little predictive by calculate your own Average Impressions Estimate.
If you’re interested, I’ve opened up the last 9 months of my personal data for you to play with. You can find the data set here.