Q&A: Getting the Greatest Value from Social Media Analytics (Part 1 of 2)

Predictive analytics can be used on social media data to better understand and react to customer behavior. Combining it with in-house data can yield deeper insights.

Although some companies are starting to use the powerful technology called predictive analytics on social media data, few are maximizing its value. Analyzing a blend of social media and data stored in-house is where the richest insights can be found, explains Information Builders' Dan Grady in this interview, the first of two parts.

Grady, the social media analytics and enterprise search sales manager at Information Builders, has worked on social media analytics, search-based business intelligence, mobile applications, predictive analytics, and dashboard design in his 15 years at the company. He blogs about social media, business intelligence and more at

Grady recently spoke, along with Fern Halper, TDWI's research director of advanced analytics, at the TDWI Webinar Social Media Analytics -- Getting Beyond Tracking the Buzz.

"Social media data by itself is very good at exposing either problems or opportunities," he says, "but how you respond to those problems or opportunities is usually going to involve using other data to support the follow-up activities."

Some companies are already using analytics on social media data, Grady explains. Going a step further and analyzing a blend of social media data and stored data from elsewhere in the company -- customer, sales, or financial information, for example -- can yield deeper insights and better guidance on what steps to take, but it's seldom done.

We asked Grady to discuss the untapped potential of predictive analytics and social media.

BI This Week: Let's start with a definition. When we say "social media analytics," what are we really talking about?

Dan Grady: Social media analytics is the analysis of the interactions and engagements that are taking place on social media sites such as Facebook, Twitter, and all of these places that we visit and contribute to.

In the technology space -- and Fern Halper addressed this really well at the beginning [of her recent TDWI Webinar with Information Builders] -- there are two different approaches to social media analytics. First, there are social listening posts out there that are primarily focused on managing the conversations your organization is having with everyone in the social media world. ()

There's another set of technologies that is focused on the analysis of those conversations and the engagement that is going on there. Here at Information Builders we're very much focused not on the back-and-forth dialog but on analyzing the conversations -- what's going on out there and how that impacts your organization.

The place where most organizations -- and many vendors -- are trying to get is the ability to bring social data together with other business data out there. It's a very immature space. Earlier this year, Forrester put out a report [saying that] to be in a leadership position this year, vendors had to clearly articulate how they were going to bring social media data together with other business data. As a vendor, if you didn't have a road map for doing that, you were knocked out of Forrester's leadership positions.

In terms of technology, where is the industry right how regarding social media analytics? You said it's a very immature space.

Yes, and a very dynamic space. ... From a technology perspective, if we focus on the analytic pieces, you have plenty of people focused on tracking what I call the "engagement level" of data -- if, when, and how much activity is taking place on all these different sites. That's a commodity -- you can get that from the platforms or sites themselves. If you look at your Facebook account or at Twitter, that information is readily available. You can track engagement over time and so forth.

However, engagement-level data is not overly insightful from an analysis point of view. From an organizational standpoint, it's nice to know, but it doesn't really help you get better as a company. That's where the textual data comes in. From our standpoint, we try to get our customers to focus on the textual data -- that's where the real value is going to be, we think.

As companies and technologies start to mature, they start to focus in on analyzing the conversations themselves. That's when they start looking at things such as sentiment analytics on text, or word frequency analysis, or entity extraction, or dynamic categorization of the text itself. Those technologies are constantly evolving. There are some companies that have been doing it for some time, but text analytics is still an immature space. Language is a complex thing, and it's tough to figure out.

That's where the speculation, and the hesitation, comes from in many organizations. How much faith and how much trust do you put in the analysis of the text itself? Things like sentiment are very difficult to track and understand and probably will never be 100 percent accurate.

Speaking of that hesitation, where are companies themselves in terms of analyzing social media? (You mentioned three levels of maturity in your TDWI Webinar.)

Right. It depends somewhat on the industry you're in -- consumer products and retail companies have to be doing social media analytics. It's ingrained in what they do, and for that reason they tend to be at a fairly mature level.

However, if you're just looking at the engagement-level data we just discussed, that's what we call the lowest level. ... Then one step up from engagement is what we call the listening level. Most organizations that we see at Information Builders are either at the first level (engagement), or the second level (listening).

I haven't seen too many companies that are actively merging social data with other corporate data -- that's the third level. If they are doing it, it's an extremely manual process.

When you start getting down to smaller and midsize companies, very few of them are doing any real analytics. They're mostly just tracking the engagement-level data. They are definitely not at level two, the listening level. They are tracking the number of fans they have on their Facebook page, and they probably have someone who responds to a question or query, but they're not looking at things like, "Hey, we had a three percent negativity on this day versus a seven percent negativity the day before. What did we do that generated that change in perception of our brand?"

You gave some great examples in the Webinar of social media analytics in use. Can you share some of those here?

In terms of use cases, there are a few that come up fairly open. At Information Builders, we tend to focus on those that involve more than just social media data because what we tend to find is that social data by itself is very good at exposing either problems or opportunities, but how you respond to those problems or opportunities is usually going to involve other data to support the follow-up activities.

The primary example you'll see is campaign analysis, with a digital marketing campaign. If you're doing engagement-level analysis, you'd be looking at what is happening in the campaign during a set time period. Maybe social traffic went up across your different sites, but you're also going to want to see, over the long run, whether transactions or revenue or sentiment analysis went up as well over that time. Social media analysis will track part of that, but revenue or transactions, for example -- those are areas where you need to pull in other data from other systems.

Another example that we see quite often is brand crisis analysis. One good example is a cruise ship that suddenly stops moving. You then notice that over a 10-day period, there's a gentleman who makes hundreds of posts on social media, the majority of them negative. That's a lot of sudden negativity about your brand, all coming from one person. You need to respond as quickly as possible. The data that can help you to understand who this person is and how to respond probably isn't on social media. Is he on the ship, for example? You need to look at the manifest system to find out. Has he ever been a customer before? You need to look at your CRM system. Is there a travel agent he used?

In any case, you're going to need some data you have elsewhere in order to respond in a way that turns this person from being a brand enemy into a brand champion. That's an example we use for customers to help them understand how social media analytics could be used effectively by merging other types of data.

Another big area of opportunity for social media analytics is competitive analysis or peer-to-peer analysis. Very rarely in the past have you had access to your competitors' data -- but with social media, now you do. I can go out and track our brand and our company, just as I can go out a track all of our competitors and the conversations around their brands.

An example I like to use is comparing McDonalds, Wendy's, and Burger King. Let's say I'm Wendy's, and I see a spike of activity on McDonalds' Facebook page. I'm going to drill in and try to understand what caused that spike. Are they running a new marketing campaign? Did the McRib sandwich just get re-launched and are people excited about it, or are they in a brand crisis situation? Did someone just find a shoe in a cheeseburger and now there's all this activity on the McDonald's page about a shoe in cheeseburgers? Wendy's can immediately start running a campaign -- hey, no shoes in our cheeseburgers.

One thing we try to get people to understand is [that] you need benchmarks. You need to compare numbers over time, or versus your peers, to understand if 7 percent or 13 percent negativity or whatever is good or bad.

TDWI Membership

Get immediate access to training discounts, video library, BI Teams, Skills, Budget Report, and more

Individual, Student, & Team memberships available.