Twitter has recently updated the way pictures and vine videos are displayed on the platform. Both web users and those on mobile devices can now see previews of media on their timelines and profiles without having to expand the view of a tweet.
Knowing the large positive impact that including pictures in Facebook posts has on 'likes' and other forms of engagement, we were excited about the potential for a similar effect on Twitter. An initial test that we ran confirmed our hopes: We sent out two promoted tweets linking to our blog post on tweet ideas. The two tweets were identical apart from the fact that one of them included a picture of a cute Twitter bird.
The results of this test were fantastic. Both engagement levels and click-through-rates were much higher on the tweet that included the picture. We decided to run a few more experiments with Twitter Ads to investigate the effect of picture tweets further.
We decided to set up 3 identical keyword campaigns for @Driftrock. All three broadcast two versions of a tweet that was successful for us previously - one with a picture and one without. Each campaign targetted the followers of different companies.
24 hours later, it was clear that the tweets with pictures in them were vastly outperforming those without when it came to engagement levels and clicks. For example, one picture tweet received almost 4 times the number of total engagements.
However, when we looked deeper a more complicated picture emerged. Although the number of clicks was higher on picture tweets, the vast majority of these were general clicks on the tweet rather than actual link clicks. And between 35-45% of those clicks were on the picture URL rather than the outbound link.
This ultimately negatively impacted the Cost Per Click. Paradoxically, the CPC on picture links ended up being on average 2.3 times higher than on plain tweets, despite the higher engagement levels.
You can see the full results of the campaigns in the table below.
|Campaign 1||Campaign 2||Campaign 3|
|Pic||No Pic||Pic||No Pic||Pic||No Pic|
|Cost Per Engagement||£0.26||£0.12||£0.27||£0.35||£0.23||£0.14|
|Cost Per Click||£1.34||£0.60||£2.08||£1.05||£2.04||£0.75|
We believe that the main reason the proportion of actual link clicks was so low compared to overall clicks is that most Twitter users access the platform through their mobile phones and click tweets to take a closer look at the picture.
Yet we also wanted to find out how much of the effect on engagement was due to the specific picture shown rather than the novelty of a media post in itself. For this reason, we ran another set of experiments prior to this test to determine which picture to use for these campaigns.
Reassuringly, the most relevant picture we tested was also the best performer. However, the performance of the tweets containing the remaining three pictures was terrible: Almost twice the number of people clicked on the picture link rather than the website link in those. This drove up the actual CPC to infeasible levels.
We therefore urge digital marketers who want to use pictures in their promoted tweets to start by testing several pictures and to always choose relevant media.
One of the first things we noticed when reviewing the three campaigns was that the number of impressions on the tweets without a picture was between 5-10 times lower than that on the better performing picture tweets. This made the data on them much less reliable. We therefore decided to run a final experiment. We created two campaigns, one containing just the tweet without a picture and one containing the picture tweet.
The difference in the results as compared to the campaigns that contained both tweets was astonishing.
As expected, both the engagement rate and the number of engagements on the picture campaign was much higher. The cost per engagement on both campaigns was similar. However, the CPC for the campaign without a picture was almost 40% lower!
Putting the plain tweet in its own campaign made a huge difference, as it gave it a chance to accumulate a lot of impressions.
|Cost Per Engagement||£0.38||£0.40|
|Cost Per Click||£4.50||£2.80|
When you set up a campaign that contains several tweets, Twitter will initially blast them out equally to the audience you selected. After some time, some tweets will show higher engagement levels than others. Based on these results, Twitter will then give the most successful tweets in a campaign more impressions relative to the others. If these tweets continue accumulating high levels of engagement, Twitter will refocus its targetting around the types of users that responded the best to these tweets. This is why when you look at your campaign manager, the performance of one or two tweets usually stands out.
The problem comes when this feedback loop is counter-productive to the marketer's goals. When you have one tweet that is more likely to convert versus another that engages a wider audience, Twitter will always focus on the latter, since advertisers pay equally for any type of engagement.
Ideally, marketers should create a different simultaneous campaign for each tweet they want to test to counteract this. However, this can quickly become cumbersome if you are testing more than a few tweets at a time. Unfortunately, at the moment the market lacks a powerful Twitter Ad campaign manager that would allow for this best practice solution to be implemented conveniently.
1) For organic (non-paid for) tweets, include pictures as much as possible - they clearly draw attention and engagement.
2) For inbound campaigns targetted at a wide audience, it may be worth it to include pictures in tweets. However, including pictures in tweets can derail the focus of direct response campaigns.
3) Make sure to find out which tweets will be your best performers before engaging in a big campaign. To do this, go through different stages of testing tweets that differ by just one variable at a time. One good way would be to first test the link copy, then the best picture to go with it, and finally the targetting groups for the final campaign.
3) Don't be deceived by high engagement rates on picture tweets - despite picture previews now displaying by default, users are still very likely to click on the picture link rather than the outbound link. Base your decision off of metrics such as CPC and CPA rather than vanity clicks that you pay for but don't lead to conversions.
4) Make sure to set up UTM tracking on every outbound link individually - this way you will be able to track which tweets convert better. These may not always be the most popular tweets.
Have you used pictures in your tweets? What was your experience?
Follow our blog now so we can let you know when follow-up tests go live!