3 Ways That Top Brands Do Advanced Social Analysis…And You Can, Too
It’s easy to get sucked into the doldrums of day-in, day-out, social media analysis: “Total Engagement, Impressions…just give me the numbers so I can show my boss how well we are doing” is an understandable sentiment for any social media manager or social analyst.
5 Keys to Cross-Channel Social Analysis
For a lot of organizations just getting up and running, this level of analysis might be enough. In the big kids club, however, you rarely see the really successful brands content with this level of detail.
As a former analyst here at Simply Measured, I’ve had the privilege of working with some of the biggest and best in the industry today, implementing custom solutions for their social analytics. Without naming names, it’s safe to say that these exceptional brands think about social in a much more elegant way than most social marketers.
Here are three ways I’ve seen some of the best in the industry take their analytics to the next level, and a few easy ways you can replicate their success leveraging data from Simply Measured reports for fast results at the next tier of analytics nirvana.
1. Create Your Own Metrics
Vanity metrics can be a fast way to get a pulse on what’s happened with your social brand over a period of time. These metrics aren’t engraved in some tablet upon high, though! Someone, somewhere, decided to measure these things, and many have found these metrics useful, but it doesn’t mean this is the only recipe available to you.
Don’t be afraid to take the data available to you and create a metric suitable to what your organization is trying to achieve on social.
One of the greatest examples of this I’ve seen is a “Mother of All” metric, used to determine success for a large international brand that needed a common success language across all regions they’re active in.
Rather than be content to have multiple KPI’s with individual associated performance indicators, this brand rolled all of its KPI’s into a behemoth, benchmark-weighted metric. This enabled the brand to:
- Measure each region’s success period-over-period, based on improvement in this weighted metric
- Summarize reporting for quick communication of which regions are killing it, and which need to pick up the pace, without having to comb through each individual KPI and making weighted mental models about what’s working.
Take this work off your executive, and do the modeling before it hits their desk. They’ll love you for it.
Arbitrarily weighted or customized metrics are always hairy business, because they inherently rely on some subjectivity, but if you are flexible and willing to change your models as needed for changing business priorities, this is a great tactic.
Here’s an Example
Making sure your customized metrics are normalized is usually the way to go, especially if you plan on using the metric to benchmark different segments of your social media presence.
A simple example is something like Total Engagement growth / Audience Growth for a time period, measured across different channels. Every social media network is different and will have wildly different numbers here, but if we normalize, we come come upon something a bit more useful.
Let’s say we are active on Facebook and Twitter, and want to compare our new “Engagement / Audience Growth” metric between them. If we just used the absolute numbers from each over a time period, we would probably end up with something pretty one-sided, but if we try to normalize our values first, the resulting ratio will make much more sense in our comparison.
Here’s how we might do that in Excel:
#1: First, we calculate how much total engagement occurred in the period as a normalized percent increase. If we just subtracted end period engagement from beginning, we would have an absolute number, but it might not make sense as an absolute number compared to another network that has lower or higher typical engagement.
Instead, we do a safe percent calculation to see what the total engagement growth for our network is over the period compared to how many engagements another network might accumulate.
YES, clever reader, you are right that engagement is usually not presented in a cumulative format and you would probably need to calculate this yourself within your data, but work with me here; I’m trying to teach you things!
#2: We can do the same thing for our audience growth (this one is much more common). If you’re wondering about the “ABS()” bit, this makes the value inside of it an absolute value, so that you don’t end up with any shenanigans with any negative numbers as your numerator or denominator. Great habit to get into when doing % changes!
#3: Tada! In cell F5, you can see our “Eng % Change / Audience Growth %” score. It’s not much now, but it might be useful if we did the same thing when comparing the Facebook score to Twitter and seeing which channel sees strong correlations between Engagement and Audience (or lack thereof).
2. Segment in a Way That Makes Sense for Your Brand
One of the fun parts of social media analytics is how many different ways data can be segmented. Across region, time, account, client, so on and so on. Each of these certainly has a use case, but just like vanity metrics, don’t be afraid to approach data in a way that makes sense to your business!
One of the more successful examples I’ve seen of this is custom reporting on identified Influencer activity based on their relation to the brand and the brand’s own business categories. Doing this level of custom segmentation undoubtedly requires some manual work to get segmentation to make sense and set up, but the payoff is well worth it to get measurement that makes sense to your business, not which OS someone tweeted on.
Let’s say you’re a cookingware company, and you know that your influencers and advocates are spread in a few different directions. You’ve got your restaurant chefs who like to communicate via blogs and their Facebook fan page about using your products.
You’ve got your amateur food snappers active on image-based platforms like Instagram and Pinterest using the recipes on your websites…
A photo posted by Andy McClellan (@amc8682) on
…and maybe a whole other group whom you’ve identified as advocates for your organization’s philanthropic efforts. Each of these groups deserves their own grouping and consideration based on the wide array of ways your brand is being highlighted on social.
Here’s an Example
Welcome to lesson two! This time we’re going to look at how we can identify how many of our identified famous chefs tweeted about our brand from a list of tweets mentioning our Twitter!
First, we need the dirty work of correlating our influences to their segments. Let me just whip up some fake data real quick…
OK, great, some totally realistic data reflecting all of the influencers for our internationally-recognized cookingware brand. Now let’s look at our tweets for a period:
In column C, we need to match the names of our influencers from our original index of influencers and segments to their tweets so that we can count them. For simplicity’s sake, I’m going to move the indexed list next door and then show you how we can match these.
We’re employing the dynamic duo of INDEX/MATCH here to get our segments into column C, based on the name in column A. If you’ve used VLOOKUP before, this is the pro version, but it accomplishes the same thing (and personally I think is a LOT less confusing).
Without getting into the thick of it too much, we are telling Excel we want to return a result from column G, where it matches a specific value in column A, in its next association list in column F. It’s as hard to explain as it is to grok, but spend a few minutes checking out the image above and it should click.
When we’ve populated our C column, we can finally do a COUNTIF function with “Chef” as a parameter to get the total count of chefs who tweeted about us. We probably didn’t need a function to be able to tell us the total number is 5, but when we have sheets of thousands of Tweets, it’s definitely handy.
3. Hack Your Analysis
One of the tell-tale signs of a seasoned analyst is their ability to think laterally with the data they are given. Some of the best in the industry are able to get blood from stone by finding new and creative ways to get insights.
A great example of this is finding out whether your audience is actively engaged on your Snapchat activities, without access to direct Snapchat data. In lieu of tapping a Snapchat public API (of which there is currently none), a cheeky analyst could run a keyword analysis on mentions of “Snapchat” on high-volume platforms such as Twitter. It’s not the most reliable data, but getting a feel for whether your audience is talking about your Snapchat by also including “@ mentions” in the tweet, or just Snapchat in general, is a creative solution to a problem that many others might pass over as simply impossible.
The trick is to think about how to create your own tools when traditional ones aren’t readily available when faced with difficult data questions. This is why having access to granular data is so important to the truly sophisticated analyst; a single value is only as good as a predefined problem, and when your social media needs to contribute more than just answering “you are on social media,” having high-fidelity data becomes increasingly important.
Here’s an Example
Let’s learn to count all of the engagements from tweets that mentioned “Snapchat.” We’re going to be using SUMIFs for this one, along with a handy wildcard trick to look for any mention of “Snapchat” (the word) in a cell.
Another fantastic set of fake data:
We’re using a tried-and-true SUMIF here to get our total, but in our criteria argument, we’ve added some apostrophes, asterisks, and ampersands. This is because Excel is weird. What we’re actually doing is telling our formula to count any text in column B where at ANY point Snapchat is mentioned.
The asterisks count as “wildcards,” meaning anything can come on either side and we will still count that row. The apostrophes and ampersands are there to make Excel confusing and hard to understand (but more to help separate the wildcards from the word “Snapchat” in a way that keeps us from getting in trouble with our formula).
The answer is 152, by the way!
As more brands get practice in social media and level-up their expertise, the industry as a whole is becoming more sophisticated, and those who aren’t learning from the best and constantly working to improve their analysis will inevitably fall behind.
There’s no better way to improve your work and prove value in social than by consistently reevaluating how to better your efforts and find more insights, and seeing what those at the top of the sophistication curve are doing is one of, if not the, best way to make sure you and your organization are front-of-the-pack.