As a social media marketing analyst for Simply Measured, I have had the privilege to work with a ton of data from some of the largest brands in the world. We analyze data across each of the major social networks to provide marketers with benchmarks to measure their own success, and to understand how changes on the networks impact brand tactics and audience engagement.
While it might seem like analyzing hundreds of thousands of Tweets requires a totally different approach than analyzing your own brand accounts, many of the steps you should take to guide your analysis are the same.
That’s why I would like to share some tips that will help you analyze data for your brand accounts.
Social data analysis should go well beyond regular reporting. It isn’t just about answering how much engagement you got, or which content performed best; it’s about understanding why your audience engaged and why certain types of content performed well.
Social data analysis should be about questioning and validating your tactics. Why do you post at certain times of the day? Why do you post at the frequency that you do? Are these the most effective tactics for your brand?
As social marketers, our world is changing daily. We are constantly barraged with network changes that introduce new forms of content and new ways for users to engage. With these changes, we have to question how the brands we represent can benefit from these changes, and we are often asked to justify our tactics.
For example, Twitter recently enabled vine videos and native photo uploads to appear expanded in users’ feed. A testable hypotheses would be that the new expanded content is viewed more than links and as a result, drives more engagement.
Regardless of your role, you should be asking these questions. Involve members within your marketing org, question your tactics, as well as changes on the networks. Form assumptions based on your experience and then test. This is where good analysis and data-informed decision making start.
This is my favorite part. Don’t believe me? Ask anyone here at Simply Measured. As a member of the Simply Measured team, I’m proud of our reporting, not only for the thought leadership that has gone into each reporting use case and the work that has been done to make our reports beautiful and presentation ready, but also for the fact that we enable our customers with the raw data.
This is where I geek out. Having the raw data means that you have metrics for each post already built into the report for you. From there, you can parse the data to go even further with your analysis.
For example, if you are concerned about comparing Tweet length, the number of links or hashtags in Tweets, or Tweets linking back to your brand’s domain, it’s all possible to analyze from right within our reports.
Take a look at what is already baked into our reports – they download natively to Excel. From there, you should look at your hypotheses and create any additional fields needed to run your analysis. This is where your Excel skills will come in handy. Once you’ve finished with your calculations, check your work and get a second set of eyes. Your analysis will rely on the quality of your data.
Sometimes it can be easy to think about social data too holistically. Its important to remember that on each network there are different types of posts, and forms of engagement. Different types of posts can reach different audiences, and each form of engagement can impact your brand differently.
Note: If you’re comparing engagement metrics from one network to another (i.e. Twitter to Facebook, Instagram etc), then you’re not comparing apples to apples. Engagement actions and audience behavior are different for each network.
Even when analyzing data on a single network – like Twitter – you must be mindful of maintaining a sound methodology that doesn’t skew your findings:
On Twitter, post types include @replies, Tweets, and retweets. Remember that @replies aren’t intended to drive engagement. Know that retweets drive engagement back to the original author of the Tweet rather than your brand. These are very important details.
If you’re trying to figure out whether having links or hashtags in your Tweets impacts engagement on your brand Tweets, you need to exclude @replies and retweets from your dataset. Otherwise your engagement averages will be skewed by including Tweets that aren’t designed to drive engagement.
In our Twitter study, I put this into practice in order to identify that brand Tweets (excluding retweets and @replies) with two or more links outperformed the brand average by 150%.
Calculating your brand’s average per post engagement for specific types of posts is a useful technique for analyzing content performance for different types of content.
Segmenting posts types and different forms of engagement can help you apply a sound methodology and identify correlation between post attributes and user behavior. Remember, compare apples to apples and your overall brand average.
Even when working with hundreds of thousands of Tweets, it’s still important to make sure that you are analyzing a significant sample size. Be sure not to slice your data so thin that you end up basing your findings on a data sample that’s too small.
For example, you might find that although you have 10,000 Tweets in your dataset, when you look at just Tweets with five or more hashtags, you only have a handful of Tweets to analyze. With such a small dataset, any observations you make will be inconclusive at best.
One way of expanding your sample is to extend your date range. Another is to include your competitors in your dataset.
Tracking your competitors will allow you to learn faster by gathering a significant sample of Tweets in a shorter timeframe. It will also help you explain fluctuations in engagement specific to your industry, helping you to more accurately attribute changes in engagement.
The key is to not lose sight of your sample size, this will ensure that the analysis you do is meaningful and actionable.
When pinpointing network changes, or changes to your own brand tactics. Take a look at a trended view of your key metrics. For example, you can test whether use of Twitter’s new expanded photos are driving more engagement for your brand.
To test this, you might view trended engagement before and after Twitter made the changes to the way it displays these photos. Or you can even go a step further, and measure per post engagement by day, during the same time period, for only the posts that included photos.
Since changes on the networks or to brand tactics can often be traced back to specific points in time, mapping events to trended data can often help you answer important questions.
With access to so much data, it’s important to remember that’s easy to find false correlation. When you set out with an assumption, it can be easy to find data that supports it. Make sure that you continue to test your findings.
Social media is one of the fastest changing industries in terms of metrics and data availability. What you found to be true last year, or even last month might not still hold true. Make sure that you’re not just measuring the what, but the why.
Find your answers in the data. Make sure you and your team members are asking the right questions, challenging each others tactics and testing your assumptions with the data.
If you don’t have the time or resources in-house to get at these findings, you should make that your goal in 2014. If you’re seeking to build your skills to be a more savvy data driven marketer, keep an eye out for our upcoming video series on how to get hands on with the data.
Want to get your hands on all the social data for your accounts, but aren’t already a Simply Measured customer? Request an invite to a free 14 day Trial of Simply Measured!