In the analytics world there is data, insights and the hardest of all: actionable insights.
Data could be something as simple as saying an auto brand grew its fanbase on Facebook by 6% in the previous quarter. An insight would say that 6% growth was three times higher than the industry average.
An actionable insight would say that when the auto brand capitalized on pop culture it saw more shares which correlated to surges in growth rate, therefore the takeaway for a social media manager is that keeping tabs on Internet pop culture can help boost engagement and fan growth.
But not every metric lets you create an actionable insight, does it?
When we started looking at doing some end of year snapshot reports for different industries, we knew that we wanted to give those actionable insights to readers. The ones where you can take a nugget of information and apply it to your social media strategy.
Sure, we could create beautiful infographics or set up mini sites on Twitter customer service that give you data like the Banking & Finance industry replied to over 1M tweets in 2014 or airlines posted 50,000 pieces of content on Instagram. It’s certainly interesting but is it useful? Can you do anything with that data?
Taking it a step further, we could tell you that American Airlines replied to over 231,000 tweets in 15 minutes or less per tweet in 2014, and responded to over 55% of the mentions it got. By comparison the industry average is a reply time of 127 minutes and only replies to 34% of @-mentions. That’s an insight right there, quite an interesting one too, but what action can you take on it?
2014 became the year when social media turned more into paid media. One of the biggest questions that social media managers are asking right now is how much engagement are their competitors getting from paid social vs organic social.
We knew that this had to be the crux of the reports but without any data available on what content was being promoted (the social networks don’t provide this in their APIs), how would we go about providing actionable insights?
There are programming solutions available for this purpose, but these are still being tested and refined. The only solution available to the team was to dive in to the Twitter analytics and do what humans can still do far better than machines - make inferences, connect dots, draw lines from seemingly unconnected data and analyze without knowing what patterns we’re looking for.
And that’s how we uncovered Starbucks’ paid promotion strategies on Twitter. Here's a break down on how we did it using the Unmetric platform.
1. Find the source of the tweets
Brands publish tweets to their timeline using a variety of sources. Most are recognized publishing tools, but some brands are quite happy for their employees to tweet from their iPhones.
We knew that to create an advert on Twitter, we had to go through the Twitter Ads platform. We also knew that since we track the source for each tweet, we can search for brands that have tweets from the Twitter Ads platform.
In this case, we found that Starbucks tweeted from five different sources in November, the vast majority of which were from Hootsuite - kudos to those guys! What exactly were all those tweets that Starbucks was sending out from Hootsuite? Replies!
This left 34 tweets which were brand tweets and 19 tweets which were retweets.
2. How many of those brand tweets were Twitter Ads?
We know that Starbucks published 34 brand tweets and 15 of those were published through the Twitter Ads dashboard. This means that Starbucks is putting money behind nearly 50% of the original, creative content it publishes on Twitter.
3. Identify the exact tweets that were paid for
The next step for us was to try and figure out which of the 15 tweets were paid for.
Twitter advertising lets you create something called Twitter Cards which can be targeted and published to anyone you want, but won’t show up in your Twitter timeline. Twitter Cards give you several options to pay based on lead gen, website clicks or conversions, app downloads, new followers, tweet engagements or customized for something else.
The important part here is that if a link is included, it is formatted to include this URL:
A quick browser search shows up all the times that the word ‘cards’ is mentioned in Starbucks’ tweets. In this case there’s 16 mentions and since each tweet contains the word ‘cards’ twice, we know that of the 15 tweets that Starbucks put money behind, eight were cards.
(Please don't judge me by the number of tabs I have open in my browser, I know I have a hoarding problem!)
We were now hunting for seven out of 15 tweets that had been paid for. It seemed to us that if a brand is paying to boost content, then it’ll probably have a high engagement score.
To test the theory we sorted Starbucks’ 32 brand tweets by what we call an engagement score to see if there were any tweets that got a higher engagement than normal.
It’s easier to visualize this data, which we’ve done in the chart below. The chart also highlights the eight paid tweets that use Twitter Cards which we had identified earlier.
There is a clear drop in engagement between 600 and 400. It’s also interesting to see that 75% of the known paid tweets were in the bottom 16 least engaging tweets, but we'll get to that later.
4. Make inferences to identify the remaining paid tweets
Taking a closer look at the top engaging tweets, we wanted to see if there was a reason why they were getting such high engagement. Were they resonating with followers? Did Starbucks do something extraordinary or did they just just put money behind promotional tweets?
The chart below is the result of our analysis. The two dark orange lines are the tweets we know are paid for because they contain Twitter cards. The light orange lines are the tweets that are very promotional in nature, either asking people to buy something, win something or click on a link. The gray lines are non-promotional tweets.
Just so that we're clear, here’s what a promotional tweet looks like:
And for comparison, here’s how one of Starbucks’ highly engaging, non-promotional tweets look:
Remember, we found that Starbucks had posted 15 tweets in November through the Twitter advertising platform and we were searching for seven more paid tweets. We had identified eight of those paid tweets because they were Twitter Cards and easy to find, but Twitter doesn’t tell us which tweets have been paid for so inferences have to be made.
Now, from the chart above, is too much of a coincidence that of the 13 top engaging tweets, we found seven that were very promotional in nature? We didn’t think so.
From a business objective, it would make sense for Starbucks to put money behind tweets that are designed to bring in revenue. They potentially could put money behind the non-promotional pumpkin pie tweet, but it wouldn’t necessarily give the brand a revenue boost.
5. Identify the learnings from Starbucks' paid tweets
Having felt that we had identified Starbucks’ paid tweet strategy on Twitter, we did another round of analysis on the 15 paid tweets. Why were some getting such high engagement while others were not? What exactly were these paid tweets promoting and why was Starbucks happy to put money behind it? And more importantly, what actionable insights could we take away from all this analysis.
The results of this Twitter analytics work is available in a downloadable report that we've already shared with our clients. It breaks down exactly what Starbucks did, what they were split testing, why we think it made sense for them to do it and what smaller brands who are trying out paid Twitter advertising for the first time can learn from Starbucks’ paid tweet strategy.