Measuring What Really Works on Twitter: Post Timing and Headlines

53 Comments

picture-34
(Photo: da100fotos)
“What gets measured gets managed.”
-Peter Drucker

I like data and enhancing performance through following the numbers.

I use half a dozen tools to track metrics on this blog, and I have similarly used tr.im to track click-through on Twitter links, demographic and geographic splits, etc.. I find retweets interesting, but only to the extent that they attract meaningful attention (not just impressions), which can be approximated with clicks on embedded links. In the last two weeks, I’ve found bit.ly to be more reliable and robust than tr.im.

Below I’ve included my top-30 most-clicked Tweets from 12/21/08 to 2/13/09…

I’ll present a number of findings first that I’ve observed across all the data gathered thus far (about 8 months across several tools), recognizing that the sample size for this approximate two-month period isn’t sufficient in and of itself.

First, what are the best times to Tweet for maximum impact?

12 midnight – 4am PST: 6, all between 12 midnight and 2am = 3 per hour
4am – 8am PST: none (at least partially due to the fact that I’m sleeping)
8am – 12 noon PST: 8, all 10am – 12 noon = 4 per hour
12 noon – 4pm PST: 11, 9 between 1:30 – 3pm = 6 per hour
4pm – 8pm PST: 9, evenly spread = 2.25 per hour
8pm – 12 midnight PST: 3 = 0.75 per hour

Conclusion: Based on this limited data set, if you were to post in one 1.5 hour window during the day, the most effective time frame would be 1:30-3pm PST (4:30-6pm ET), which averaged 6 top-30 hits per hour. It’s possible the 4-8am PST period could beat this, but — alas — sleep is more important to me than Twitter.

Second, do the most-clicked headlines/Tweets have common characteristics?

For headlines, here are a few observations:

Top 10 Tweets – 4 questions (4 opening questions)
Top 11-20 Tweets – 2 questions (1 opening question)
Top 21- 30 Tweets – 3 questions (1 opening question)
Conclusion: 70% of the most-clicked Tweets are statements, but 4 of the 4 questions in the top 10 were opening questions. If you include a question, using it to begin the Tweet will increase click-through.

The most common words in top-30 most-clicked Tweets: how to…, best, most, worst, great.

The top-30 most-clicked Tweets in descending order of popularity

1—-Total Clicks—-Date—-Time (PST)—-Actual Tweet

2 4724 1/13/09 11:38 AM The best job in the world – seriously – is taking applications now: http://tr.im/62yw Thanks, MJ! (www.minjungkim.com)
3 4180 1/25/09 10:35 PM Want to remember everything you’ll ever learn? Use this algorithm: http://tr.im/bnsa Thanks, Gary Wolf + readers, for reminding me :)
4 3838 1/26/09 4:20 PM The most incredible fat-loss photos you will ever see: http://tr.im/cwlw
5 3640 1/21/09 12:49 AM The “worst Twitter post ever”? http://tr.im/b9a1 I doubt it.
6 3627 1/12/09 12:18 AM Markus Frind works <1 hr per day and makes $10 million per year: http://tr.im/4scm Good illustration of keeping it simple.
7 2602 1/20/09 10:21 AM What happens when one of the world’s best violinists plays in a metro station for change? http://tr.im/ave3 using a $3.5-mill violin
8 2514 1/7/09 12:24 AM Some great thoughts on the topic of boredom: http://tr.im/32me Thanks, @gmc!
9 2230 12/22/08 11:18 AM The most impressive 27-year old you’ve never heard of? Barack Obama’s speechwriter: http://tr.im/2k3k
10 2139 2/4/09 8:56 AM How to fold a t-shirt in 2 seconds (video): http://tr.im/ekl4 First saw this method in Tokyo.
11 2080 12/22/08 5:11 PM Newest Wired mag feature now online ‘Flirting with The 4-Hour Workweek’ – freaking hysterical: http://tr.im/2kjs

12 2016 1/20/09 1:43 PM How to learn any language in 3 months: http://tr.im/ayrt
13 1953 1/2/09 4:50 PM Is the Aeron chair worth it? http://tr.im/2uxd Do you have any fave chairs for extended sitting and writing?
14 1864 2/4/09 1:49 PM For those who prefer undressing to folding. How to undress in 2 secs (video): http://tr.im/emf7 Thanks, @1rick!
15 1833 12/31/08 12:30 AM The best infomercial ever: http://tr.im/2rqy This guy is great – “you’re gonna love my nuts” Check out the fine print S&H.
16 1728 1/26/09 10:42 AM Today on Techcrunch – Kevin Rose: 10 Ways to Increase Your Twitter Followers – http://tr.im/cuaj
17 1695 2/5/09 5:15 PM How to open a bottle of wine w/ a phone book: http://tr.im/eul2 Tree or car tire also work quite well.
18 1667 2/2/09 11:28 PM How to do push-ups without your feet touching (video): http://tr.im/eawc
19 1576 1/31/09 5:50 PM My favorite article about redesigning (and explaining) the Internet: http://tr.im/dwze
20 1574 12/30/08 3:29 PM Thinking of setting up online companies in Vermont. Anyone tried this? http://tr.im/2rd0
21 1541 1/19/09 11:26 AM Great chart – The Air Force’s Rules of Engagement for Blogging: http://tr.im/a98z Thx @gmc!

22 1534 1/27/09 5:30 PM The basic movements of Parkour: http://tr.im/d687 Wear ankle support.
23 1413 2/2/09 12:38 AM How to build an upside-down fire. Fun stuff: http://tr.im/e47t
24 1394 1/28/09 12:39 PM RT @devincheevers Great write up on Verdasco v. Tsonga. No time? Just read paragraph 3: http://tr.im/dct7 My fave paragraph in a long time.
25 1360 1/8/09 2:32 PM When insane pen tricks go wrong: http://tr.im/39qq From @jpdefillippo You’ll shoot your eye out!
26 1349 12/29/08 11:52 PM Wow. The real Star Wars hologram projector, from USC: http://tr.im/2qjq Thx, @gmc!
27 1348 12/24/08 2:58 PM Trial by Fire is here! http://tr.im/2mme Takes a min to load, but thx to akman and @rostendorf for pointing it out :) More X-mas coming
28 1339 2/6/09 1:55 PM Looking for successful case studies and new tools/sites/resources for updated and expanded 4-Hour Workweek: http://tr.im/ezw3
29 1290 2/2/09 11:08 AM RT @mashable How much is your Twitter account worth? http://tr.im/e771 Also, added reader fire-building tips: http://tr.im/e47t
30 1260 1/25/09 2:10 PM Giving away brand-new Fujitsu color travel scanner: http://tr.im/cmhc Do sth #nice for others today and link or describe. Have fun with it!
31 1251 1/1/09 1:43 AM Drunk awareness test. How many passes does the team in white make? http://tr.im/2rrv Happy 2009! It’s going to be the best yet :)

Can you find any patterns in the data that I missed, or have you found patterns among your own most popular Tweets?

Related Links:
Follow Tim on Twitter
How to Use Twitter Without Twitter Owning You – 5 Rules
Top 10 Twitter Tips for Beginners
5 Ways to Use Twitter for your Business or Career (NY Times)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Comment Rules: Remember what Fonzie was like? Cool. That’s how we’re gonna be — cool. Critical is fine, but if you’re rude, we’ll delete your stuff. Please do not put your URL in the comment text and please use your PERSONAL name or initials and not your business name, as the latter comes off like spam. Have fun and thanks for adding to the conversation! (Thanks to Brian Oberkirch for the inspiration)

53 comments on “Measuring What Really Works on Twitter: Post Timing and Headlines

  1. Tim,

    Thank you for doing this. Because you have so much more traffic potential, this kind of analysis on your Tweets is priceless. The sample size is so small for most of us that it would be a lot harder to draw these kinds of meaningful solutions.

    Cheers,
    Adam

  2. Tim,

    I love your evangelism of great new tools and useful metrics for tracking productivity – I had no idea that twitter analytics like this existed, and am always grateful for another way to track how effective I’m being (or not, as the case may be).

    Keep up the laser-focussed and highly-quantified work!

  3. Hi Tim,
    Wow, amazing data! I was just listening this morning to Ed Dale (et al) on Internet Marketing This Week – They were talking about twitter a lot. Might be worth checking out their podcast if you have not already… they mentioned you a couple times already :)

  4. Hey Trim,

    Great observations on what works and why. For me, I’ve found that adding a photo (in our case a photo from whatever part of the world we happen to be in) increases our CTR substantially. What I didn’t realize was how effective questions we’re. Great Work.

    Best,
    Rob

  5. @Tim

    I bet that you can increase your click-throughs by changing the color of your Twitter links on your blog. The blue is very difficult to read against the dark background.

  6. I guess since the posts are so short if you treat it like a headline of an ad you can get a good response . Good copywriting skills would come very handy for this .
    what do you think tim ?

    oh and by the way i am really interested in your views about learning techniques , methodologies , hacks .could you write more about that and also Could you recommend and books , websites , blogs or any resources for more info on that .
    Thanks in advance .

  7. I guess since the posts are so short if you treat it like a headline of an ad you can get a good response . Good copywriting skills would come very handy for this .
    what do you think tim ?

    oh and by the way i am really interested in your views about learning techniques , methodologies , hacks .could you write more about that and also Could you recommend and books , websites , blogs or any resources for more info on that .
    Thanks in advance.

  8. Hi Tim, great info. with your research regarding Twitter. Quite honestly I am having difficulty finding the value in the Twitter platform from my business perspective. I find it to be way more socially based than business and continue to search for how my company, Sweet Spot Golf, can use this outlet.

  9. Interesting. The data looks polynomial in terms of rank order – so the 20th most popular tweet is 3/4 as popular as the 10th but the 10th tweet is 1/2 as popular as the first etc. Which tells me that popularity has much less to do with time of day than the attention grabbing nature of the tweet. It also indicates that you probably have a pretty consistent audience at any given time throughout the day (waking hours of course). Makes sense if your audience is across multiple time-zones. Time of day of tweet will be less important.

    The question position is compelling. I think a question is a good call to action in a tweet. Makes a lot of sense. You should keep collecting all this, compile it up, throw it into a stats package and run a model on it. Might be interesting.

    • Hi All,

      Here is a very cool observation on click distribution by Paul:

      “Interesting. The data looks polynomial in terms of rank order – so the 20th most popular tweet is 3/4 as popular as the 10th but the 10th tweet is 1/2 as popular as the first etc. Which tells me that popularity has much less to do with time of day than the attention grabbing nature of the tweet. It also indicates that you probably have a pretty consistent audience at any given time throughout the day (waking hours of course). Makes sense if your audience is across multiple time-zones. Time of day of tweet will be less important.”

      Thanks, Paul!

      Tim

  10. Are you marketing or servicing your following? The purpose of analyzing your tweets is to learn what you can do better, but it looks like you’re doing great! Your top click through tweets appeals to one sense of $$$$. FREE! How much is your Twitter Account worth. Not Surprising given the economy.

    Were you able to track how many new followers? Qwitters?

  11. These are awesome analytics (and the first of its kind as far as I know). The 4:30-6pm ET is intriguing but I am surprised niche specifics weren’t brought up in this post. I have found that particular target audiences are on at different times. Obviously this is important if one would like to use twitter as a marketing tool (or even just to connect with the like-minded). I have made my own observations and have incorporated Tweetlater in an effort to maximize “tweet impact.” This could be deemed important for those of us with (much) less than 20,000 followers. i.e., tighter niche.

    Indeed the exponential nature of the data is fascinating. Trying to see what else can be ascertained here…

  12. Hi Tim,

    I would find this more impressive if you were an unknown. But since you aren’t and have hordes of “fans” who follow you, I imagine that the numbers are a bit skewed….

    What would be more interesting would be if you opened an anonymous twitter account, befriended the masses and then analyzed your open rates. I imagine that they would be far different than what you are getting now…just my .02 though.

  13. Thanks for the great info Tim! You say that you have used tr.im & bit.ly to track your click throughs from twitter, but it looks like you send most of your tweets from ping.fm. Are you able to tie ping.fm into one of the tracking services? Thanks!

  14. Interesting analysis.

    Couple of questions.

    Does your work consider which of the clicks are real, ie: bots vs. human?

    In terms of Tweets for impact, what’s your thoughts on geo-distribution of followers?

  15. Tim,

    I’ve always found through my own analysis via Twhurl links about Twitter (i.e. tools, tips, news) always have a much higher click rate. I guess that is down to people on twitter being interested in twitter.

    Using this theory, the tighter your following is focus to your niche (something to consider when trying to grow a reciprocal following), and the more niche based your tweets = higher CTR.

    Interesting analysis – surprised your Porn one didn’t make the top ten…

    Cheers,

    Chris

  16. Tim

    Impressed by your book and your deconstructivist (is that a word?), first principles approach to life and learning. Am working on internalising what I can use. Have to say, though, that twitter seems to me to be the antithesis of the silence and space that many of your other ideas would help to promote. No offense, but I have no interest in following another person on their day-to-day experience of the world. It seems to me as paradoxical an unattractive as sitting in front of a reality tv show to watch someone else living. It places too much emphasis on someone elses experience of life, and not enought on ones own adventure. That is not to say that what you are twittering is not intelligent and potentially useful information. But information is not what we are all lacking. Thanks for the book and the blog, which are great (and enough). I will go outside and listen to the birds twitter.

    Thanks

    Scobie

  17. “What get’s measured gets managed.”
    -Peter Drucker

    That unnecessary apostrophe makes me feel ill. Good post, though.

    ###

    Tim: Just a typo. Corrected.

  18. Thanks for the data Tim. This is interesting stuff and valuable information for how Twitter might help businesses. There is so little data right now so anything helps. Oh by the way, the fact that you have time to do this is amazing all by itself.

    JP

  19. Wow Tim,

    This stuff is over my head at the moment. I have been using Twitter about 6 weeks. I consider myself fairly techy, been around computers a long time.

    Staying on the cutting edge is more of a challenge.

    Thanks for the information about Twitter RT’s. More to learn now!

    Your Pirate Lifestyle Guru,

    Rob

  20. I think you missed one incredibly important variable, the number of followers you had at the time of each tweet. That would make things easier to analyze and compare. If you had 1000 more followers at the time of a particular tweet compared to another it could make a big difference. All in all a great look at the data side of things, looking forward to seeing more!

    • @Doug,

      I totally agree on the variable of increased followers. This would be a great correlation to include, perhaps putting a “% of total followers” number next to the total number of clicks.

      Best,

      Tim

  21. “5 3640 1/21/09 12:49 AM The “worst Twitter post ever”? http://tr.im/b9a1 I doubt it.
    6 3627 1/12/09 12:18 AM Markus Frind works <1 hr per day and makes $10 million per year: http://tr.im/4scm Good illustration of keeping it simple.”

    Both link to the “fat-loss” story in no. 4 (the text is correct, the address is not – if you hover over those links you’ll notice them directing to http://tr.im/cwlw )

  22. Not too surprising that most likely time to be RT’d is the most likely time for social networks to be active among the demographic of Twitter. Combine with days of the week too for pinpoint accuracy. Running social networks you get this kind of data pretty easily.

  23. I love this kind of case study. I’ll talk to the other engineers at bit.ly and see if we can do a similar case study across a day’s worth of bit.ly links to get a more significant sample size and post results back here if we can mine this data.

    Glad you’re onto bit.ly now, btw. Reliability and speed have been our focus of late, and it appears to be paying off. Stop by next time you’re in NYC and we’ll hook you up with a pufferfish and t-shirt.

  24. Tim,
    I’m attempting to go from utter obscurity to somewhat well known through a combination of Twitter, blogging, networking and a website. I’m the painter turned Behavioral Coach! I’m at 19.5K for February.

  25. Tim,
    This is really fascinating stuff! You are definitely on the cutting edge in all facets of life – I’ve leard so much about Twitter and it’s value in 5 minutes!

    Thanks!

  26. Hmm..the only way to generalize this beyond one datapoint is to have a sample across a large number of other people who track this as well as you did. Otherwise, while the results are interesting, they apply to only one person..namely, you.

    • @Allen Weiss,

      You wrote:

      “The only way to generalize this beyond one datapoint is to have a sample across a large number of other people who track this as well as you did. Otherwise, while the results are interesting, they apply to only one person..namely, you.”

      I totally agree 100%.

      All the best,

      Tim

  27. Hey Tim,

    As others have mentioned, you have hordes of fans and followers, which impacts the strategies that work for you on Twitter (a “How to” tweet from RandomDude might not be so much more effective than RandomDude’s other tweets, for example).

    For context, I would love to see which of your tweets perform the WORST – say, the ones that made you lose followers – and analyze those against the data you have here.

    All my best,
    Susan

  28. Hi Tim – great info, thanks. But you say that to tweet, “the most effective time frame would be 1:30-3pm PST (4:30-6pm ET)”. Ha! That’s all very well for ppl who live in North America, but on the other side of the world, where I live, in Hong Kong, that’s 4:30am-6am the next morning.
    Kay

  29. Don’t you find that most successful tweets are very egocentric? ie. asking to test ‘your’ ‘ mine’ own skill, ability, intelligence, looks. Surelly we are better than the best and not worse than the worst out there…

    After I think about it for a minute, I’m not too surprised. You made me think. I like that…Oh damn, here comes the ego again:)

  30. Wow–and the clicks just keep on coming. I can see why your tweets pull well. Undressing in 2 Seconds link shot (easy-to-find vid, though), maybe others as well; haven’t hit ‘em all. And, yes, I plan to read every post. (Perhaps I’ll load them into SuperMemo for future call-up…)

  31. Tim I wonder how difficult you find it to balance twitter posts that contain links with posts that don’t. I follow you on Twitter and, if I’m not mistaken, you make a lot of tweets that contain links. I wonder if you have had any problems with that from twitter. If perhaps they have scolded you at any point?

    Thanks.

    Love the blog. Keep up the great work.