How To Coax Social Media Insights From Google Analytics
For most of us, analytics software behaves like idiot savants: they can surprise us with amazing feats of calculation, but totally miss the social cues that the rest of us take for granted. So, it’s no wonder that analytics packages struggle to tell us how we are doing socially.
As a follow-on to my column last month on measuring social media, I’d like to show you how I got Google Analytics to tell me how I was doing socially.
Ask Very Specific Questions
Google Analytics (GA) is good at counting things. You can’t ask GA how your social network feels about you; it doesn’t get the concept of feeling. You can ask it to count how many people click on links in your social media posts.
Last month, I asked some very specific questions in the form of a report:
- Which social networks were delivering new subscribers to my email list?
- Which social networks were giving me the best conversion rate?
If I could answer these questions, I could decide where best to spend my digital time and digital dollar online.
The Social Experiment
This summer, we set about trying to measure the direct impact of social media on my practice’s bottom line. We wanted answers to the above questions.
The content-oriented strategy that we put in place went pretty far in telling me what we wanted to know. Unfortunately, our results don’t rise to the standard of statistical significance.
It is more important that you see how we were able to aggregate social media results easily and repeatably, drawing a bright line between social network activity and bottom-line conversions.
My goal was to get new subscribers to my email newsletter. Education is the heartbeat of my practice.
My blog, The Conversion Scientist was to host my social landing pages, and I present an opportunity to sign up for the email on each page.
Each page of the blog contains the subscribe conversion beacon.
Using a Google Analytics goal, I instructed GA to count a conversion as a visit to the “welcome” page delivered after this form has been completed and the visitor has confirmed their subscription by email.
URL shorteners are a great way to track clicks on status updates and tweets that may get passed around the social Web. My preferred choices are Bit.ly and Budurl.com because of the excellent analytics and reports they offer.
However, I needed to further instrument my social links so that I could sort them in Google Analytics. GA’s link tagging features provided the answer. I discuss this in a bit more detail in my previous column. You can download the link-tagging spreadsheet I use to generate tagged URLs.
We can then use GA’s “advanced segments” feature to isolate traffic that is coming from specific URLs. When I tag my URLs, I put the service that delivers the message containing the link in the “source” field.
In GA, I can create a segment that identifies all traffic where the source is one of the social networks I’m targeting—in this case Twitter, Facebook, LinkedIn or Flickr.
Simply running a report on referring sites won’t work. I’m not interested in all traffic from social networks. I want to isolate only those visitors that are coming because I exposed them to my content.
Once defined, I run almost any GA report against just the traffic that is visiting because of my social promotions.
My promotional social media visits vs. all visits to my site.
So, Google Analytics knows what a conversion is, it knows which of my traffic I’m trying to isolate for analysis. Now all that is left is to build the report. You may remember the magic report that I mocked up from my previous column. We’ll use GA’s custom reporting to generate that for us.
The first thing I wanted to see was how many conversions I was getting from each social network, which is tagged as a “source.” So, for each source—Facebook, LinkedIn, Twitter—I want to see how many visitors arrived, and how many new subscribers I got. GA will also calculate my conversion rate.
Custom report tells me visits, new subscribers and my conversion rate for each Source.
I created this report as well as to show me which campaigns did the best job of generating subscriptions and which of my posts and pages generated email subscriptions for me.
During the three-month experiment, we generated the following predictive results:
- We completed a total of 274 updates, tweets, and posts as part of the experiment, and does not include any conversational social media I engaged in.
- Spredfast reports that I had a “reach” of 1.7 million individuals, most of them from Twitter.
- Spredfast tracked 810 bit.ly clicks in total. This includes all posts, not just those supporting this experiment.
- We increase my Twitter followers by 24%, Facebook friends and fans by 16% and LinkedIn connections by 8%. Overall, I added 17% more friends to my social graph.
We also tracked the following definitive results:
- Google Analytics tracked 257 visits from links sent out through social networks for this experiment. Assuming 1.7 million impressions, that is a click-through rate of 0.02%. This is not an unusual CTR in social media marketing, but we thought it would be higher for a curated social graph like ours.
- Social media visitors viewed about the same number of pages, but spent 26% less time on the site per visit.
And the final count of new email subscribers from all of this activity?
Five. This represents a 1.97% conversion rate for the experiment traffic.
With so few conversions, it is impossible to draw any real conclusions from the experiment. We added a great deal more subscribers through search, email and referrals, and we can compare these to our social program thanks to this effort.
Here’s the report we were looking for. Despite the results, it is a powerful tool to analyze social media conversions.
The report linking social media network to conversions.
We were disappointed with these results, but the results aren’t the point. What is important is that we’ve identified a way to reliably measure the conversion rate of our social media efforts in a way that keeps us from having to aggregate data from several places. You may see very different results from your social graph, and I’d like to know if you try this technique.
The reasons for this low conversion rate don’t necessarily lie with social media as a channel. The causes may include:
A poor social media landing page. We may be doing a poor job of converting visitors on our social media landing pages. More contrast to draw the eye to our offer will be one of the first things we test next.
Low trust in the social networks. Our social graph may not trust us since almost everything we post is content-oriented and we aren’t doing enough conversation-oriented updates.
A technical error. None of these numbers is “accurate.” Analytics is an estimation that only correlates to reality. It is possible that we are way under-reporting the conversions in this case.
Not counting visits from through other means. We know that URLs get shortened when others share our posts and tweets. This may have stripped out our tracking code in some cases. Our social media postings might also have created indirect visits. Did our efforts increase search traffic? We did not see such a correlation. Nor did we see a correlation to traffic from referring sites.
In general, 1.7 million impressions resulting in five conversions has left us realizing that we don’t have a social graph large enough to support the focused level of investment in social media right now. I suspect we haven’t generated the trust level to get noticed in the stream as well.
We need a little conversation with our content.
We don’t intend to abandon social media, but we’ll be reconnecting our Ping.fm service to provide more automated updates, and will be doing less cultured network-specific messaging. For now. We are also spending more of our efforts on email, video, audio and joint ventures this fall.
I know this will generate some good discussion. Please join the dialogue in the comments section below.
Analytics news and expert advice every Thursday.