twitter experiments post header
Insight

What we learned from four weeks of experimenting on Twitter

Estimated read time: 8 minutes

Does the way you use Twitter day-to-day have much of an effect on things like the amount of impressions, engagements and link clicks you generate? Only one way to find out; test!

Those of you paying close attention to our Twitter feed might have noticed that we’ve approached the platform a little differently these past few weeks; that’s because we were running four weeks of experiments, all for the purposes of this very article.

So, sit back and relax as we take you through what we did and what it showed us.

Our baseline

Our Twitter account went live in early January, shortly before our site launched. Since then, we’ve tweeted every day – mainly about our content – and have clawed our way past 250 followers.

All of our social media has been organic and our growth has been slow and steady.

To give us our baselines for this experiment, we pulled the analytics for all our tweets up to 6 April and crunched the data. Below is how an average tweet for us performed in that period:

  • 576 impressions
  • 14 engagements
  • 2.1% engagement rate
  • 0.5 retweets
  • 0.2 replies
  • 2 likes
  • 2 URL clicks
  • 12,886 impressions per week (average total for all tweets sent in a week)

With a baseline in place, we then decided on how we’d shake things up each week.

Our experiments

For week one (7-14 April), we made sure there was either an image or a gif on EVERY tweet we posted, because we all know visuals are the real deal, right??

For week two (14-21 April), we put all our tweets out in clusters at the start of the day, at lunchtime and at the end of the day, to test if more people see our stuff during their commute/break.

For week three (21-28 April), we only tweeted about one thing per day – namely, whatever new article we’d posted that day. We posted about it once, three times a day to test if quality over quantity gave us anything different.

Finally, for week four (28 April – 4 May), we just tweeted with furious fingers. We posted three times an hour between 9am and 5.30pm, mixing up our own content, cool stuff we found online and retweets with comments (so we could get some analytics on them).

The results: week one

Here’s how things looked when we added a visual to every tweet, with an emoji to indicate whether each stat was above or below our baseline. Each tweet this week averaged…

  • 261 impressions ⬇
  • 10 engagements ⬇
  • 3.42% engagement rate ⬆
  • 0.5 retweets ↔️
  • 0.1 replies ⬇
  • 2 likes ↔️
  • 4 URL clicks ⬆
  • 5,472 total impressions for the week ⬇

It’s probably no surprise that the engagement rate was up for this week – after all, that’s what you expect when you add a visual, isn’t it. interesting that we averaged more link clicks this week too – as a publisher, that’s kind of what we’re after.

This week’s tweets required quite a bit of work to create and/or source the visuals for each message, but when you see a significant increase in our average engagement rate AND a doubling of our link clicks, you’d have to say it’s worth it.

The results: week two

How did clustering our tweets fair for us? Let’s have a look – each tweet averaged…

  • 154 impressions ⬇
  • 2 engagements ⬇
  • 0.21% engagement rate ⬇
  • 0.3 retweets ⬇
  • 0 replies ⬇
  • 0.6 likes ⬇
  • 0.5 URL clicks ⬇
  • 6,448 total impressions for the week ⬇

Wow, look at all those downward arrows! We won’t be repeating this in a hurry!

In a way, it’s good that this week bombed; tweeting in this way felt like a really disconnected way of doing it. Sure, it took next to no time to ‘do Twitter’ each day, but it felt like we gave off an impression of not really caring…and trust us friends, we care.

Admittedly, it’s somewhat weird that this approach earned us more impressions than the previous week. But, judging by all of the other metrics, those extra eyes counted for nothing, as they clearly kept on scrolling.

The results: week three

So, how about tweeting just three times a day, and only about a single thing? Quality over quantity? Here’s how those tweets averaged…

  • 260 impressions ⬇
  • 7 engagements ⬇
  • 2.5% engagement rate ⬆
  • 0.6 retweets ⬆
  • 0.2 replies ↔️
  • 1.5 likes ⬇
  • 2 URL clicks ↔️
  • 3,386 total impressions for the week ⬇

A bit of a strange one to judge here; the small increase in engagement rate is interesting, but pretty much everything else is pointing down (we’re choosing to ignore the 0.1 increase for retweets because, well, how do you do 0.1 of a retweet?).

Again, this method of tweeting felt very hands-off, very blazé. It didn’t really feel like us.

The results: week four

Finally, what would happened when we tweeted as much as we could while still actually getting our work done? Here’s how our furious fingers averaged…

  • 132 impressions ⬇
  • 2 engagements ⬇
  • 1.44% engagement rate ⬇
  • 0.2 retweets ⬇
  • 0 replies ⬇
  • 0.5 likes ⬇
  • 0.5 URL clicks ⬇
  • 15,327 total impressions for the week ⬆

Right, it’s obviously no surprise that our total impressions are way up for the week – that was to be expected purely because of how often we were posting. But look at all of those other stats. All significantly down – we were clearly posting too much and simply posting loads did not help get more people onto these pages and reading our delicious articles – in fact, it almost killed our traffic.

If you want the truth…it’s a really good thing that this method was so fruitless. Why? because it was damned exhausting and damned stressful, that’s why!

Final thoughts

While this little experiment hasn’t left us with a clear idea of what we absolutely should do on Twitter, it has clearly shown us what doesn’t work for our account.

The biggest thing we’ve learned though is to know who your audience is and then, well, go with your gut. It’s no surprise that the second, third and fourth weeks didn’t yield amazing results because it didn’t feel like the right way for us to tweet while we were doing it.

If anything, it proved that we were on the right path before this experiment – around five to eight posts per day, mixing up our own content (new and from the archive) with some shares of other interesting stuff.

We’ll probably throw in more images from now on – especially for things we’d really like to drive traffic to – but we certainly won’t be posting multiple times an hour.

So, with that done, it’s probably time to get back to it!

 

Related Posts