Experiment Design

Hearing about all the hype around the Twitter Ads platform, we decided to give it a test run. We have used Google and Facebook ads in the past with varied results. Twitter could be that right tool to give some social exposure and boost initial social scores, hence providing a snowballing effect… or at least this is what we are banking on.

Ok, here is our setup:

Day 1

Day one is coming to a close. But first things first. Since we have just started using the Twitter Ads platform, we will share our first impressions. We have used Facebook Ads in the past and found some similarities with what Twitter has to offer. The social advertising business model is definitely viable and is a sound competitor to the Google approach for several reasons:

  1. It reaches out to very relevant audiences and creates a lasting tail effect. Those who “like” or “follow” you, indicate that they are open to hear from you again and become your fan base.  For that reason, social advertising, in our opinion, provides a better long-term return.
  2. Social ad platforms are generally very robust with the exception of LinkedIn
  3. Both Twitter and Facebook provide a multitude of relevant and very useful metrics.

We found the Twitter Ads platform very intuitive. It took us less than 5 minutes to set up a campaign. The analytics interface comes with tons of useful features, demographic and geographic data.


Tweeter Ads analytics deserves a special mention. It offers lots of useful information. You get a detailed breakdown of all activity in the Tweet Activity tab. The Follower tab offers a detailed demographic breakdown of your followers, but also provides information about their interests and gender. Twitter Cards offer a visual break down of the entire campaign and the ROE. You can also link your website and track clicks to it.

Between the hours of 5PM to 9PM, we have received 10.4k impressions, 51 engagements and spent $9. Out of these 51 engagements, there were 47 clicks, 1 re-tweets, and 3 follows. Our engagement rate clocked at 0.49%. We ran 4 tweets and they have had approximately same engagement rates.

It’s notable that @MarcGuberti followers generated most engagement, followed by @blogboy2 and @GuyKawasaki.

Approximately, 65% of our audience was male, 40% from the US and 60% from Canada, which is in line with higher social engagement rates in Canada.

Day 2

There is one very important nuance with Twitter Ads. The Clicks engagement type is the clicks on your tweets, not on the inbound links in them. In our opinion, Twitter does not make it clear enough. It is very difficult to find a definitive guidance on the value of the Click engagement type on Twitter. It’s a measure of how many times someone actually clicked on the tweet itself, NOT the link in it. As such, we find this engagement somewhat useless given that a tweet’s content is fully visible regardless weather you click on it or not, and a minimal additional engagement benefit is obtained by clicking on it. Within the Tweet Activity tab, one can find the actual number of link clicks for each tweet, which is very disappointing – only 3 clicks, which is in-line with what Google Analytics reports. The same information can be found in the Websites tab. Our conclusion is that the Click engagement type is confusing and not very useful to pay for. But it also eats up most of you budget, 13-18 cents at a time. For examples, today Twitter analytics is reporting  87 clicks but we actually can see only 3 inbound link clicks in Google Analytics! The average bounce rate for these clicks is 100%.


There is a also a tracking problem, where the inbound clicks are not being picked up by Google Analytics. According to Hendry Lee, “Lack of Referrer header for traffic from other Twitter clients. Desktop and phone-based Twitter clients are popular. According to TweetStats, web accounts for about 51% of the whole Twitter apps, so where do the stats for the 49% go?. When someone clicks on your link, it will not carry the Referrer header. Your log file and web stats are more likely to track this as direct traffic.”

Hendry suggests using the Tweetburner URL shortener service which allows tracking. We tried to use the Tweetburner link, which caused Twitter to pop an error message refusing to post it. We used the Google URLs shortener and Twitter accepted the link.

Day 3 – Final Results

The third day of our campaign is coming to an end and as the dust settles, somewhat disappointing results are emerging.

We have spent $41 in three days, or $4 below our planned budget and what do we have to show for it:

  1. 21,071 impressions, which don’t have much value in our opinion as a performance indicator, however it shows that our tweets are being displayed. Great!
  2. 151 clicks on our tweets, which, as we explained before, is a metric of a dubious value, especially on Twitter where one can see the entire tweet without necessarily clicking on it. We don’t see much value to this engagement type and consider it somewhat misleading.
  3. 2 re-tweets, this is useful
  4. 6 follows, this is useful
  5. 10 clicks on our inbound links, this is great, but the bounce rate is 80%!

Our average engagement rate is .75%

The average cost of engagement is $0.26 according to Twitter, but in our opinion the number is not correctly calculated as it factors in tweet clicks.

Here is our calculation:

2 re-tweets + 6 follows + 10 inbound clicks = 18

$41 / 18 = $2.27

This is quite comparable to Google and Facebook PPC costs. It’s important to note that this part of our experiment was completed over the weekend. The weekends have statistically lower engagement rates than the weekdays. The next leg of our experiment may show different results, but it would be interesting to see the average.

In conclusion, our impression is that Tweeter does not have that much engagement depth. The Clicks engagement type eats up your budget very quickly providing little tangible benefit. Meanwhile, real engagement such as re-tweets and follows is very difficult and expensive to get. We will be putting our advertising dollar in Facebook and Google.


Comments are closed.