The best time to tweet if you want people to notice is … 4.01pm, according to my survey of nearly 120 Twitter users.
11am-12pm is another good time – but you’ve got more chance of being retweeted at 4.01pm.
The main findings of the survey are:
- Half of people read every tweet.
- But 20% read only the last 20 tweets or less when they check Twitter.
- At 11am-12pm, 51% of Twitter users are looking at tweets, but only 31% are sending them – a margin of 20%.
- At 4pm-5pm, 56% of Twitter users are looking at tweets, but only 39% are sending them – a margin of 17%.
- At 7-8pm, the margin is just 2% – people are more interested in tweeting than reading.
- Weekend Twitter users are much more likely to be tweeting than weekday Twitter users who often just browse tweets.
- As a proportion of all tweets, use of the term RT (for retweet) is highest at 4pm.
But before all that, thanks to everyone who took part in the survey – the ones I know about are listed at the bottom, together with an explanation of the survey methodology, and the issues with this that I’m aware of – as one person said: “I don’t think this will yield any useful results. Twitter isn’t an audience, it’s a platform. You can’t generalise Twitter behaviour. Happy to be proved wrong, but I’m sure I won’t be.”
How far back people read
This graph shows you how many tweets people read when they go to Twitter.
It shows that just over half of people read all the way back to the last tweet they looked at.
Many of the rest read back quite a long way, but 20% of Twitter users don’t look back any further than the last 20 posts (or fewer). (The ‘other answers’ mainly involve tweetdeck).
Weekdays: reading vs sending tweets
The graph below shows the % of Twitter users in the survey who said that they read tweets at a given time, and the % who send tweets at that time. (Here’s the data in table form).
As you can, see there is a broadly similar pattern of reading and sending tweets – the main peeks in reading seem to be as people arrive at work (9-10am), around lunch (12-1pm) and towards the end of the working day (4-5pm). There’s a smaller peak later – after dinner perhaps.
The interesting thing is when the biggest differences are – when people are more likely to be reading, and less likely to be tweeting. This difference is 20% at 11-12am, 19% at 12-1pm, and 17% at 4-5pm.
So if you want a tweet to have the best chance of being noticed, tweet at these times. Lots of people will be looking, and you’ll have less competition from other people tweeting.
But be careful of 2-3pm that’s when the margin is at its lowest (ignoring the middle of the night). 44% of Twitterers are reading but 38% are tweeting – a margin of just 6%. Your tweets may get missed if you send them at this time. The same is true of 7-8pm.
Weekends: reading vs sending tweet
This graph shows the same data as above, but for weekends. You can see people use Twitter much less at the weekend (the X axis for both graphs is the same) – and, relative to their reading, they much more likely to tweet. (Here’s the table.)
The difference between the %age who read and who tweet is much less at the weekends. So avoid weekend tweeting anything important – there are fewer people to see it, and relatively more of those who are around are tweeting so your tweet is more likely to get lost in all those other tweets. Weekend early afternoons are particularly bad – people are much more interested in tweeting at these times.
Getting a tweet noticed isn’t just about making sure people see it. You want people to retweet it. And there’s a clear pattern to retweeting.
This graph from Twist shows the proportion of tweets that contain RT (the shorthand for retweet). I’ll say that again – it’s the percentage of tweets containing RT, not the absolute number of tweets containing RT.
To me, it says that people are less inclined to RT at the weekend (although maybe there is less to RT at the weekend). But if you look at the weekday figures (mouse over the graph to see the times), you can see that the weekday peaks are at 4pm. So combined with the fact that a tweet is more likely to be seen between 4 and 5pm, this leads me to the conclusion that 4.01pm is the best time to tweet for maximum exposure.
To gather the data for the survey, I drew up 5 questions, set up a google spreadsheets form, and watched amazed as the data dropped in.
You can see the form here: http://tr.im/TweetSurvey. 295 people visited it according to my tr.im stats. Of these, 117 filled out the survey by the time I did this analysis. (For those who asked why the form couldn’t be more sophisticated, this is why! Only about 1/3 of people who clicked to see the survey filled it out. This is a fairly good response rate, but more questions would have lowered it).
Obvious problems with this methodology are that:
- There is no waying of knowing how representative the respondents are. I’m sure Britney Spears and Stephen Fry didn’t reply.
- The people who RTed it for me were mainly people involved in SEO or journalism. So this will definitely have skewed the data.
- I didn’t ask how many people respondents followed – so I’ve no way of knowing the proportion of people who follow a select few, hundreds or thousands of people. This will obviously affect the figure for how far back people read.
- The questions were geared to the web interface. Obviously there are all sorts of ways people can access Twitter, including watching for specific keywords etc in various API-based tools, grouping people you follow and treating groups differently. There were no questions about this sort of usage, as I wanted to keep the form simple. The respondent I mentioned above thinks this invalidates the results. He could be right …
- The questions are a bit vague. I just asked when people looked or tweeted more often than not. This relies on their recall, and it doesn’t address how much they do at a given time (EG they might usually send 100 tweeets at 9am and usually send 1 tweet at 3pm. These both get counted just as times when they commonly tweet).
- It assumes patterns where there may not be any. People may not have times when they usually do things – I neglected to add an ‘it’s all random’ option.
- If you have US and UK followers, you need to account for the time difference. Tweet at 8am UK time, and your US followers are in bed. Conversely, wake up at 8am UK time, and the last couple of pages are full of Graywolf doing some sort of annoying test, er Americans off to bed.
I also adjusted some of the data – EG I changed one response of 40-60 tweets to 50, to help with the analysis and graphing, and I shortened some of the longer replies (I should have put a 140-character limit!), or merged them
Original Article by: Malcolm Coles
Posted by: Shane Barker at 4:01pm 🙂
Shane Barker is a digital marketing consultant who specializes in sales funnels, targeted traffic, and website conversions. He has consulted with Fortune 500 companies, influencers with digital products, and a number of A-List celebrities.