Archive for January, 2011
The nominees for the 83rd Academy Awards were announced earlier this week. The winners will be chosen on February 27. In the meantime, we thought it would be interesting to use TweetReach data to predict who might win.
There are a number of data points we could use to predict the winners. For this initial experiment, we’re going to look at four metrics: reach, exposure, tweets, and contributors. We’ll start with a baseline in today’s post and check in on the numbers every Friday until awards weekend. Then we’ll conduct a more thorough analysis and see how we did after the Oscars are handed out.
Reach. Reach is the number of unique Twitter streams that have had tweets about a particular topic delivered to them. Our Oscars reach hypothesis: The movie/actor that has reached the most unique people on Twitter will win the award.
Tweet Volume. The simplest predictive metric is overall tweet volume. Our hypothesis: The movie/actor that is tweeted about the most will win the Oscar.
Contributors. The number of unique contributors could tell us something about a movie’s chances for success at the Oscars. Hypothesis: The movie/actor with the most different people tweeting about it will win the Oscar.
Reach:Exposure. The ratio of reach to exposure gives us an idea of how diverse the Twitter audience for a topis is; higher R:E ratios indicate a wider and more diverse group of people received tweets about a topic. Hypothesis: The movie/actor with the highest R:E ratio will win the Oscar.
The Nominees Are…
We’ll look at the big awards, since they’ll generate the most Twitter traffic and give us the most data to analyze. This year’s nominees are:
And The Winners Are…
Interpretation and Other Thoughts
In the first week, the frontrunners are The King’s Speech and Inception for Best Picture, Colin Firth and Jesse Eisenberg for Best Actor, and Natalie Portman and Michelle Williams for Best Actress. We also wouldn’t discount True Grit in the Best Picture category, or James Franco and Annette Bening in the Best Actor and Actress categories.
One thing we notice immediately is that some of these queries are pretty noisy. For example, James Franco gets tweeted about a lot and many of those tweets aren’t specifically about the Academy Awards or his performance in 127 Hours. But for others, almost all the tweets about them are related to the Oscars or the movie. Jennifer Lawrence is a good example of this – nearly all the tweets about Jennifer relate to Winter’s Bone or the Academy Awards.
Basically, some of the nominees are so famous that it’s difficult to sort through general tweets about them to find only the ones related to the awards. Some of the numbers above reflect this, particularly when it comes to tweet volume. This applies to James Franco, Natalie Portman, and Nicole Kidman – they were in top spots for several metrics that relate directly to popularity. As we get closer to February 27, we anticipate that a higher percentage of tweets about the actors and actresses will be related to the Oscars, which will help with the noise. In addition, we’ll work on filtering out more non-relevant tweets.
The four metrics used in this post (reach, reach:exposure, tweet volume, and number of unique contributors) are just our first step in predicting this year’s Oscar winners. Next week, we’ll get into a more sophisticated analysis and see what else we can learn from how Twitter is talking about the Academy Award nominees.
No matter the size of your TweetReach report, the report’s reach to exposure ratio (R:E) can tell you a lot about the impact of your tweets. Whether you’ve measured 50 tweets or 50,000 tweets, you can use the ratio of reach to exposure to understand something about how those tweets have spread.
The reach:exposure ratio* represents the depth of penetration of tweets about a topic. A lower R:E suggests that people are seeing tweets about a topic over and over, while higher R:E numbers suggest broad but shallow penetration of that topic. However, a higher R:E does not always indicate success. Depending on the type of tweets you’re measuring and your goals with those tweets, bigger is not always better.
Low R:E = 0.0-0.19
A low reach:exposure number (anything 0.19 or lower) suggests a large percentage of contributors are tweeting multiple times about the hashtag or keyword you’ve measured, which means the message is limited in scope and is not spreading far beyond those people’s followers. A number of people are receiving many tweets about this term.
Low ratios are fine in many cases, however. For example, regional issues and local news events, conferences, and Twitter chats, parties and contests are likely to fall in this range. And that’s perfectly reasonable for these types of events; they’re relevant to a smaller or localized audience and that’s who will see it.
If you’re aiming for a larger or more diverse audience, then start thinking of ways to get tweets about your topic out to a wider audience. What can you do to encourage more (and different) people to tweet? You should also be concerned if your R:E ratio is below 0.05. That suggests that someone is tweeting a lot about your keyword, to the point of being spammy.
Average R:E = 0.2-0.59
An moderate or normal ratio will be anywhere between 0.2 and 0.59. This suggests a normal distribution of tweets, retweets, and amplification. In this case, some people are tweeting multiple times, some influencers are tweeting to lots of followers, and most people are tweeting once or twice to their smaller set of followers. Most people will receive 2-5 tweets about your message.
Many general brand mentions and large product launches or announcements (Verizon iPhone, Chevy Volt) will fall in this range. Larger media events with a wide, popular appeal (like the Academy Awards and Super Bowl) will fall in this range, as well.
High R:E = 0.6-0.99
The closer this reach:exposure number gets to 1, the more different contributors are represented in this report. Ratios of 0.6 or higher indicate that a wide variety of different people are tweeting to a number of diverse followers, spreading the message far and wide. Most people are not seeing more than one or two tweets about the hashtag or keyword. This is ideal if you want a large amplification of your message.
Twitter trends and popular hashtag memes (#LessAmbitiousMovies, #FollowFriday), significant national or international news events, and very big-name products and brands will fall in this range.
Depending on how many contributors and tweets you measured, high ratios could be somewhat misleading. If your report includes just a few hundred tweets and you have a R:E ratio of 0.8, then one or two people with large followings can account for a great deal of this ratio. The smaller the dataset, the larger the impact a few people can have on the R:E ratio.
You should also be concerned if this number is higher than 0.85 or so (which is incredibly rare). This indicates that most people only received one tweet about this message, which might not bode well for retention. How likely are people to remember something if they only read about it once? And if only one tweet about a topic is delivered to a person’s Twitter client, it’s very possible that person did not even see it at all.
*To calculate the reach:exposure ratio, the formula is reach divided by exposure. You should get a number between 0 and 1 (if it’s higher than 1, then you probably used exposure divided by reach). For example, a recent report reached 147,425 people and had an exposure of 763,506 impressions. The R:E ratio is 147,425 / 763,506 = 0.19.
Note: We’ve written up a comprehensive set of full report benchmarks if you want more info about other metrics besides R:E.
Photo credit: Loud Speaker by paparatti.
On Sunday, January 16, 2011, we tracked tweets containing #goldenglobes and it told us a lot about how users watch TV together on Twitter. We thought we’d make it into a nice infographic. Click here to view the full size version of our 2011 Golden Globes infographic.
TweetReach Trackers now have contributor detail pages! Take a look:
The new contributor page includes influence metrics like Klout scores, number of followers and Twitter lists, as well as basic profile information. We’ve also included share of voice stats for the Tracker – both the number of tweets and overall impressions contributed. Finally, you can view the tweets that person contributed to the Tracker, and the retweet, reply and exposure information for those tweets.
To view the detailed contributor stats, just click on any contributor’s name in your Tracker.
For months – even years – people have been speculating about when the iPhone would be available on Verizon. Today, we finally learned that the Verizon iPhone will officially go on sale in February.
We’ve been tracking tweets about the Verizon iPhone for a while. And it just so happens that we also tracked tweets about CES. So, how do these two events compare? Did the Verizon iPhone announcement really overshadow CES, at least as far as the tweets are concerned? Let’s see.
Here’s the tweet volume for January 11 (times displayed in PST):
Verizon iPhone tweets peaked at more than 56,000 in one hour. As a comparison, the highest number of tweets per hour about CES was 9,641. In one week, 136 thousand people tweeted 443 thousand times about CES. In less than one day, 114 thousand people tweeted 199 thousand times about the Verizon iPhone. Tweets about CES reached 42 million people in a week; tweets about Verizon iPhone reached 33 million people in a day.
Over the past few weeks, speculation about the Verizon iPhone really heated up. Many people thought the announcement would come at this year’s CES event. Others joked that Apple, who didn’t attend CES, was intentionally waiting to announce the partnership, in order to overshadow CES. And then last Friday, January 7, we learned of a Verizon press event scheduled for Tuesday, January 11, just two days after CES ended. The Wall Street Journal confirmed that the Verizon Apple partnership would be announced at this event. People got very excited. On that day alone, our TweetReach Tracker monitored more than 45,000 tweets about the Verizon iPhone, with a reach of 19.3 million unique Twitter accounts.
Given how many anticipatory tweets we tracked, we were very excited to see the tweets on the actual day of the official announcement from Verizon. The announcement was scheduled for 11 a.m. ET/8:00 a.m. PT on January 11, 2011. During the announcement hour, tweets about the Verizon iPhone spiked, as people posted updates from the announcement and their opinions on the news. In just that one hour, we tracked 56,303 tweets from 39,275 different users, reaching 21,576,495 unique Twitter accounts.
Last week, we tracked tweets about CES, the Consumer Electronics Show, which ran from January 6 – January 9 in Las Vegas, Nevada. More than 140,000 people from around the world attended this enormous technology and electronics event.
And when we say enormous, we really do mean enormous. We used the TweetReach Tracker to monitor tweets for one week around CES (the two days leading up to the event, the four days of the event, and the day after the event). And during that week, we tracked:
from 136,738 contributors
generating 1,112,409,883 impressions
reaching 42,200,045 people
That’s more than 1.1 billion impressions delivered to a potential unique audience of more than 42 million people. Nearly half a million tweets were posted about CES, from more than 135 thousand different Twitter accounts. That’s pretty enormous.
The number of tweets about CES reached a high point on January 7, the second day of the event, resulting in more than 130,000 tweets posted that day.
During the main hours of the event on January 7 (from 8 a.m. to 7 p.m. PST), an average of 7,162 tweets were posted every hour, with a maximum of 8,429 between 10 a.m. and 11 a.m. During the four days of the trade show, an average of 3,700 tweets were posted every hour, with generally higher volumes in the afternoons.
We expected to see a lot of tweets from – and retweets of – major tech and electronics brands. And while there were definitely plenty of tweets about CES from accounts like @BlackBerry (and @BlackBerryHelp), @SamsungTweets, @kodakCB, and @Sony, most high-impact tweets came from other sources. The most influential contributors in this Tracker were mostly mainstream media outlets, tech blogs, and geeky celebrities, with only a couple tech companies making a big impact. Here’s a list of the top 12 most influential contributors to the CES Tracker. These 12 accounts contributed the top 50 tweets by overall exposure (our impressions metric) and accounted for 148 million of those 1.1 billion total impressions.
This was definitely one of the biggest events we’ve ever tracked tweets about, especially in terms of overall impressions generated. We’re curious what will top it. Maybe the Academy Awards? Guess we’ll see next month.
PS – If you’re interested in how we calculate reach, exposure and our metrics, we explain it all here. Also, we’ve been tracking tweets about the Verizon iPhone and wrote up an analysis of those tweets here. If you think this CES data is impressive, check out the iPhone data.
One of the questions we’re asked every day is, “How do I know if my TweetReach report numbers are good enough?” Raw metrics can be useful, but to be truly valuable they need to be evaluated in context.
Of course, we generally don’t recommend that you compare your social media results to others’. There are so many variables that play into a tweet’s or campaign’s reach that there is no single benchmark that applies to every situation. However, it is helpful for you to know what sorts of TweetReach numbers are appropriate in general, and how to interpret your particular results. Additionally, we recommend that you run regular reports and compare your metrics over time (you can also do this automatically with a TweetReach Tracker).
So, on to the numbers. Below are several benchmarks you can use to understand the impact of your full TweetReach reports (some of these metrics will also apply to the free 50-tweet reports, but we’ll be publishing another post soon that specifically addresses free reports). An individual full report will include up to the most recent 1,500 tweets from the past week. We used data from a large set of reports that hit that 1,500-tweet maximum in the past 12 months for the following metrics.
Reach and Exposure
In a full report (1,500 tweets), you’ll likely see reach and exposure numbers like the following:
Low: 100,000 or less
High: 1,000,000 or more
Low: 300,000 or less
High: 5,000,000 or more
The reach and exposure numbers will vary from report to report, depending on context. A report for a one-hour Twitter party will be different from a report for a three-day conference, which will be different from a report for a press release or a news event. But the above numbers are a good approximation of an appropriate baseline for any report.
If your numbers are low, look for contributors who were tweeting many times to a small amount of followers; this could be perceived as spammy to their followers. Higher reach numbers indicate that more unique people were potentially reached with tweets about your message.
More important than the reach and exposure numbers on their own is the number of people reached compared to the overall impressions generated. No matter how many tweets are in your report, this reach:exposure ratio is an important one to interpret.
A low reach:exposure number (anything 0.2 or lower) suggests a large number of contributors are tweeting multiple times about the hashtag or keyword, so the message is not spreading far beyond those people’s followers and a number of people are possibly being inundated by tweets about this term. Low ratios are fine in many cases (for example, reports about regional issues will likely be between 0.1 and 0.2), but you should be concerned if you see numbers below 0.05. That suggests that someone is tweeting a lot about your keyword, to the point of being spammy.
An average or moderate ratio would be anywhere between 0.2 and 0.4. This suggests a normal distribution of tweets, retweets, and amplification. In this case, some people are tweeting multiple times, some influencers are tweeting to lots of followers, and most people are tweeting once or twice to their smaller set of followers. Most people will see 2-5 tweets about your message. This is exactly the range we expect most reports to fall in and it’s totally appropriate.
The closer this reach:exposure number gets to 1, the more diversity in contributors to this report. Ratios of 0.6 or higher indicate that a wide variety of different people are tweeting to a number of diverse followers, spreading the message far and wide. Most people are not seeing more than one or two tweets about the hashtag or keyword. This is ideal if you want amplification of your message. You should only be concerned if this number is higher than 0.85 or so (which is incredibly rare). This indicates that most people are only seeing one tweet about this message, which does not bode well for retention. Most people will likely forget about something if they only see it once.
Low: 200 or less
High: 1,000 or more
A report for a Twitter party or hashtag chat will generally include tweets from fewer than 300 contributors. This is perfectly fine – over a short time period like an hour or two, a few hundred people in an organized chat or party will easily generate 1,500 tweets. If you’re tracking more general brand or product mentions and your 1,500-tweet report includes fewer than 200 contributors, then you should look at who’s talking (and how much), and what other things you can be doing to encourage your message to spread. In this case, it’s quite possible a few people are talking a lot about one thing in particular, and you might want to investigate that.
A larger brand or a longer-term or high-visibility event should generate higher contributor numbers. If your report include recent general mentions of a common brand name, for example, you’d expect to see at least 1,300 contributors in a 1,500-tweet report, suggesting that most people mention the brand once, then move on. A big product launch or popular press release should generate at least 1,000 contributors.
Low retweets: 200 or fewer
Average retweets: 300-500
High retweets: 650 or more
Most reports will include mostly regular tweets, like the pie chart on the left. In fact, most reports include 50% to 75% standard (non-reply, non-retweet) tweets. That’s roughly 700-1,100 regular tweets in a report.
A report for a Twitter party or hashtag conversation will likely include a large percentage of @ replies, like the pie chart on the right. This suggests the message is not spreading much beyond the people engaging in the conversation. To help to message spread beyond that circle next time, you could explicitly encourage participants to retweet or tweet directly, instead of @ replying.
For a useful general benchmark, look at the number of retweets your report includes. Most reports include a few hundred retweets; usually about 20% – 35% of all tweets in a report are retweets, as both the pie charts demonstrate. Very, very few reports have more than 50% retweets. If your goal is to spread a message as far as possible, aim for higher numbers of retweets.
Hopefully these approximate benchmarks will help you interpret your full TweetReach report. We’ll be posting more benchmarks for free quick reports and Trackers soon, as well as more detailed discussions about data for different contexts, so check back in for those posts.
We currently offer two TweetReach reporting formats – the individual snapshot report and the Tracker. One of our most frequently asked questions is when it’s appropriate to run an individual report and when it’s best to set up a Tracker. Depending on the type of data you’re analyzing, one of these two formats will better serve your needs. To decide if you need to set up a Tracker or run a snapshot report, just answer three quick questions about the tweets you’re measuring.
1. When are/were the tweets posted?
a) Recently posted
b) Will be posted in the future
If you answered a, you should run a report now. Twitter only keeps tweets accessible for about a week, so if your tweets are older than that, we can’t retrieve them for analysis. Don’t lose them!
If you answered b, move on to question 2.
2. How many tweets do you expect?
a) Fewer than 1,500 tweets
b) More than 1,500 tweets
If you answered a, then you can run a snapshot report after your event has occurred. The snapshot report uses the Twitter Search API, which searches through the most recent seven days worth of tweets. So run that report after, but within one week of, the time period you wish to measure. If you need to measure more than a week’s worth of tweets, see question 3. A snapshot report can include up to 1,500 tweets. If you have 50 or fewer tweets, your report will be free. If it’s between 51 and 1,500 tweets, it’s $20.
If you answered b, you’ll need to set up a Tracker, which can track more than 1,500 tweets. Set up your Tracker before your event begins, or as soon as you can, so we capture as many tweets as possible. The Tracker uses Twitter’s Streaming API, so it captures tweets in real time, as they are posted to Twitter.
3. What is the time period for your analysis?
a) 0-7 days
b) A week or longer
If you answered a, you can run a snapshot report or a Tracker, depending on your answers to questions 1 & 2.
If you answered b, you need to set up a Tracker before your event begins, or as soon as you can. Trackers run in real time, so they will find all new tweets as they are posted to Twitter, but they cannot retrieve old tweets.
In sum: if you answered a to all three questions, run a snapshot report. If you answered b to all three questions, set up a Tracker. For other a/b combinations, check the chart below.
*A few disclaimers about these particular combinations… Snapshot reports include up to 1,500 tweets from the past week, so if you have more than 1,500 tweets or data older than 7 days, we won’t be able to find all of your tweets for analysis. You can run an individual snapshot report anytime at tweetreach.com. Trackers monitor tweets in real time, so they will find all new tweets as they come in, with no limits on number of tweets or length of time. But Trackers cannot go back in time to include old tweets. You need a TweetReach Pro subscription to run a Tracker.