Big brand CMOs are growing weary of ambiguous and non-actionable insights from their current social media listening tools. Here is a common scenario heard throughout marketing departments today:
One of their trusted social media gurus slacks the CMO: “Latest campaign just earned an 80% positive share of voice. Boom!”
CMO responds curiously: “Yeah? Is that positive sentiment coming from our male or female audience? Outside of our brand, what else are they interested in?”
Response: “Can’t say. But our listening tool shows Brooklyn Pet and Groom positively tweeted ‘That’s what she said’ with a smiley face emoji.”
Positive sentiment? Questionable.
Male or female? Neither. Brooklyn Pet and Groom is a store.
Other interests? Not a chance.
Sound familiar?
Most legacy listening software just doesn’t cut it any more. With a reputation for being poor at best and useless at its worst, the only polite word to describe it is “limited,” and today, limited may as well mean useless. You can’t afford to waste your time or money on garbage. So when you’re in the market for a comprehensive, new generation analytics platform, how can you be sure that what you’re buying is worth it?
First, let’s take a look at the reasons why making this purchase is so difficult.
When you’re looking for a new generation analytics platform, to help determine its worth, you want to look at four things: accuracy, speed and cost, underpinning, and insights.
1. Accuracy
Depending on who you talk to, reported accuracy for legacy systems is 65-75% for 3-way (positive, negative, neutral) sentiment classification. This accuracy is not good, but it’s the best most legacy analytics systems can offer. While the results are passing at best, when you’re making an investment, you should be looking to score an “A” at the very least.
Why the near-failing grade? (Let me caveat this by saying that sentiment prediction is difficult and humans disagree on many sentiment labels.) One word: Upkeep. More often than not, the maintenance of the algorithms lag, meaning systems can’t incorporate new methodologies and data. For example, they don’t keep up with slang. Some claim 90%+, but this is usually based on a measurement that is designed to favor their system methodology.
2. Speed & Cost
What’s holding an older system model back from performing better? Simple. The speed and cost of maintenance. Increases in performance requires someone to continually upkeep and monitor the system and its data. This is a large engineering and data science challenge, not to mention prohibitively expensive and incredibly slow.
3. Underpinning
Older systems put a greater burden on the user. You get out what you put into them. While elementary boolean keyword search-based systems can be used to identify posts with brand or topic related references, such approaches are less adaptable because they lack a scalable foundation. Simply put, they don’t take to change very well. They require extensive human care and configuration, therefore it is even harder to incorporate new predictions for emotions, intent and personality types let alone other important factors like age, gender, country or region.
4. Insights
The data quantified by older systems rarely leads to actionable insights. In the best case, what passes for “analytics” consists of simple summations of social media data (e.g., number of tweets, number of followers, metadata rollups). Typically, measures of success are based on “soft,” non-actionable social media metrics like brand awareness or reach. These rarely tell us about the actual user behind the conversation.
So how do next generation analytics tools resolve these issues?
When there’s a new generation of analytics tools that are more effective at generating actionable insights from the vast ocean of social media (the endless flow of Tweets, Facebook wall comments, blog posts and customer reviews), trying to retrofit a basic monitoring analytics system to do deep insights is a fool’s errand.
An example
Say you’re sitting in a busy restaurant in the middle of a lunch time rush. Over the din you are told that before you can order, you need to be able to identify who’s eating what, where they’re from and what they do for a living by listening to the simultaneous conversations being had throughout the restaurant. This task seems impossible; it’s so loud that you’re having trouble concentrating on your own conversation, let alone anyone else’s.
That restaurant is the ocean of social media and those conversations are the data you need to record to craft leads. If you’re going to have a seat at the table, you have to be able to collect. Luckily, the new platforms have you covered.
With the new text and social graph analytics solutions on your side, this task isn’t as nearly as difficult as you think. First off, next generation platforms like People Pattern scale human intelligence to better understand “who” is having the conversations. This is conversation monitoring (which the most basic platforms do) and helps you determine the “who,” basic information for age, gender, region, and country.
When you are able to determine the “who” through conversation monitoring, you are able to move onto the next step, which for us is a new era of analytics called “audience insights”. This step applies new technologies and approaches to get to richer, more meaningful insights.
So, why is this important?
By learning to mine the “who” through machine learning techniques and applying them to new data sets and workflows, we can refine searches and segment the audience based on the indicators from social media content such as geography, gender, age, brand preferences and lifestyle interests. In fact, we can even match social media conversations to customer segments that a client has established from surveys or customer data. In other words, this greatly increases your odds of hitting a bull’s eye.
This idea isn’t new; advanced analytics platforms have been making it possible to generate actionable insights from social data for some time now. However, audience insights platforms discovered that when you characterize individuals through the content they share and their graph connections, you move beyond the content and into the realm of the “who,” which allows you to target your audience beyond the instance of the keyword they mentioned, and based on demographic and psychographic attributes.
And if you’re still wondering why?
According to Gleanster research, most brands conduct only a basic level of segmentation and data filtering around their listening efforts. However, 73% of marketers surveyed have yet to segment based on audience behavior, demographic and psychographic attributes. These marketers are missing opportunities and new analytics platforms like People Pattern will allow you to capture them.
Once you create your people-based dataset from unstructured social data using a next generation analytics platform, companies can create a single view of their customers through the integration of other sources of structured data.
If you feel like your current sentiment tool is “limited” and want to up your game, please feel free to email me for more information about our audience insights platform, or give us a shout on Twitter @PeoplePattern.
Recent Comments