Twitter – Optimal Tweeting (according to Dan Zarella)
A while back, I got a book from my Skillsoft learning
library, with the following title: The
Science of Marketing: When to Tweet, What to Post, How to Blog, and Other
Proven Strategies. As a
statistician/analyst at a university, I was curious about how the statistical
techniques that I use on a routine basis are applied in the marketing and
social media world. The author, who
works with Hubspot as a social media scientist, examines several of the more
popular social media platforms, via large data sets, to see what really works.
The book was written in 2013, so some of the findings may no
longer apply, since the social media world is a fast moving world. Nonetheless, I found them interesting,
especially the Twitter chapter. I
summarize some of his results below, with my own thoughts in italics.
Naturally, if the book interests you, you should go to the
source. The Amazon link is given
above. The book sells for about 13
bucks, and about 20 bucks in paper. Note
that he also delves into many other social media platforms and strategies, so
the Twitter stuff is only a small part of the book’s content.
Below are some his results,
many of which are controversial by his own admission:
Regarding Tweets and Followers
·
Engaging in conversation does not generally lead
to more followers. In fact, his
analysis of many millions of tweets and accounts shows that “highly followed
accounts tend to spend a lower percentage of their tweets replying to other
accounts”.
o Accounts
with 1000 or more followers had about 8% of their tweets preceded by the @ sign
(i.e. were conversational). Accounts
with less than 1000 followers had about 16% of their tweets preceded by the @
sign. A similar trend was evident when
examining accounts with greater than a million, vs less than a million
followers.
o This seems reasonable, given the difficulty
of actively engaging with a large number of followers. After all, a twitter account holder only has
so much time and attention. The more
time devoted to conversation, the less time that can be devoted to more the
more general content, that builds audiences.
·
Highly followed accounts tend to have a lot of tweets
that contain links.
o
For accounts with 1000 or more followers, about
45% of the tweets contained links. For
accounts with less than 1000 followers, only about 12% of tweets contained
links.
o
This also
seems reasonable. An account that
appeals to a large following probably can’t be too personal – there are just
too many people. But it can be
informative, interesting, educational or authoritative on some subject. That is what can generate followers (unless
you are a celebrity, in which case the minutiae of your life may hold wide
appeal).
·
More tweets are better for building an audience.
o
When he looked at the number of tweets per day,
vs the number of followers that the account had, the number of followers peaked
at about 22 tweets per day. It didn’t
fall off very quickly from that point.
So, the takeaway is, it is hard to over-tweet.
o
This also seems reasonable, at least within
certain parameters. You need to repeat
messages, to have them heard in a crowded room.
But, you have to avoid coming across as a spammer – it alienates people
and Twitter doesn’t like it. Rich
content also helps – if the content is good, repetition is probably tolerated
better.
·
Accounts with a higher percentage of
self-referential tweets have fewer followers than accounts that don’t talk
about themselves too much.
o
This also seems reasonable. You had better be an intrinsically
interesting person, if you expect a lot of people to be hanging on your every
thought or action. Again, what works for
celebrities doesn’t work so well for the rest of us.
·
Accounts with a higher percentage of negative
sentiment have fewer followers than accounts that are more positve.
o
Again, this seems intuitively obvious. Most people don’t care for a lot of
negativity (though a smattering of it is ok).
·
Accounts with a picture/bio/profile get more
followers (250 on average vs 25).
o
Again, this seems intuitively obvious. It is harder to trust people who seem to be
holding back. Of course, not including a
picture/bio/profile might also be a sign that the account holder isn’t all that
social, anyway.
Regarding Retweets
·
Tweets with links get more retweets than those
without links.
o That seems reasonable, as retweets are often
about sharing content.
·
People retweet tweets with links, even though
they don’t necessarily click on the link themselves. So, a catchy headline is important.
o Interesting, and rather counter-intuitive.
·
Asking for retweets is effective.
o Calls to action work.
·
Tweets made in the late afternoon (3-5 Eastern
Time) get a higher percentage of retweets, than other times of the day.
o People are tired at the end of the day, so
retweeting is a nice way to stay engaged without working too hard.
·
Retweeting peaks on Friday, though tweeting in
general peaks earlier in the week.
o Similar to above.
·
Tweets with novel or unusual content get more
retweets. This was determined via text
analysis (the percentage of words in a tweet that were not common words).
o People like to be the first to share new
information.
Regarding Click-through Rates (CTR, or the tendency to click on the link or tweet, rather than just read it)
·
Longer tweets have higher CTR.
o Longer tweets are more likely to have
something “clickable” to look at.
·
Tweeting at widely spaced intervals through the
day leads to higher CTRs.
o Quick bursts of tweets are more likely to have
at least some of them missed (maybe people are busy clicking on the previous
tweet?).
·
Tweets with more action words (i.e. verbs and
adverbs) have higher CTRs.
o Similar to the “call to action” phenomenon?
·
Tweets on weekends had relatively high CTRs.
o People have more time to click through and
read content on the weekend.
I should note that some reviewers objected to referring
to these findings as “science”, since they are correlational in nature, and
therefore don’t really speak to causation.
That’s true – findings based on observational data frequently have that knock
against them. But, given the nature of
the phenomenon, it is difficult to set up experiments, so it seems to me that observational
data is the best one can hope for.
Better that, than theorizing uninformed by data.
-------------------------------------------------------------------------------------------------------------
And since calls to action work, I should include one in the blog, for one of our books. Since spring is approaching, and I know people interested in Twitter are great lovers of novelty and new experiences, here's a plug for a travel book, about a hiking trail on the west coast of Canada:
The
hiking journal "A Walk on the Juan de Fuca Trail" is available on Amazon
for 99 cents. Here is a summary.
The
Juan De Fuca Marine is considered by many to be one of Canada’s finest hiking
trails. It hugs the southwestern shore of Vancouver Island, between Jordan
River and Port Renfrew for a distance of about 48 kilometres. Like its
(perhaps) more famous neighbouring hiking trail just to the north, The West
Coast Trail, it features both beach and forest hiking along a rugged coastline.
The hiking is a nice test of one’s fitness, the views are spectacular, the
wildlife (marine and forest) is plentiful and the people are friendly. What
more could one ask for?
What
follows is a journal of a five day trip, taken in early September of 2002. It
is about 13,000 words in length (60 to 90 minutes reading), and contains
numerous photographs of the trail. There are also sections containing a brief
history of the trail, geology, flora and fauna, and associated information.
U.S. Amazon http://www.amazon.com/gp/product/B013VKEXV2
U.K. Amazon http://www.amazon.co.uk/gp/product/B013VKEXV2
Amazon Germany http://www.amazon.de/gp/product/B013VKEXV2
Amazon Japan http://www.amazon.co.jp/gp/product/B013VKEXV2
Amazon Canada http://www.amazon.ca/gp/product/B013VKEXV2
Amazon Australia http://www.amazon.com.au/gp/product/B013VKEXV2
Amazon India http://www.amazon.in/gp/product/B013VKEXV2
No comments:
Post a Comment