Social Media Commercial Opportunities: How Outlier Detection Can Make Social Media Work For You

Social media provides the platform for users to have digital interactions with each other, shrinking the globe by offering almost instantaneous sharing of media and ideas. It also provides forums for consumers to have these types of interactions with brands as well, and the data gathered from these interactions can be extremely valuable as they provide real-time data on consumer perception of a company or its products.

Outliers in that data, in turn, can signify drastic events – including ones in the real world – which require an immediate response in order to stop losing money, or make even more.

A little bird told me…

Let’s say you’re monitoring mentions of your brand on social media (like the number of times per hour that your brand’s Twitter hashtag is used) and suddenly, you notice a large, unexpected spike in this metric. Often, this can be a great thing because it could indicate there’s new demand for your products, or a particular geographic location or demographic is interested in one of your products.

If either of those is the case, your company could wisely react by raising prices on the particular items in demand, stocking more merchandise in anticipation of a surge of new orders, or by reaching out to customers with a large social media following or to other key influencers. This is one example of how identifying outliers in data gathered from social media can directly lead to increased revenue.

That surge in mentions, however, could also indicate a problem. What if dissatisfied customers are actually posting stories of receiving horrible customer service from you, or complaining about problems with your product? In that case, an entirely different set of actions is required: publicly and proactively reaching out to the dissatisfied customers on social media, offering refunds, product exchanges or credits, and so on. In this case, you need to act fast to stop and repair the damage to your brand before you lose droves of potential customers (and thus their business) due to bad reputation.

Sometimes, however, a surge in seemingly positive social media mentions can actually be a bad thing. If your online store starts giving customers very large discounts because of a pricing glitch, those who benefited from the mistake are often very enthusiastic about sharing their discovery. Those posts often go viral as more users are notified of the huge discount, take advantage of it, and share their exploits, further feeding the cycle. There are websites dedicated to sharing newly discovered glitches, with pages describing how social media platforms like Instagram can be used to find out about price glitches. Likewise, people used <href=”#.kqkOddr9Q7″>Reddit and Twitter to spread the news about the glitch in Bloomingdales’ loyalty point system which erroneously increased the buying power of those points by two orders of magnitude.

The necessity of a large-scale solution

Determining whether a given outlier in social media data is good or bad requires context, and that means monitoring more metrics. As the list of metrics grows, manually monitoring all of them in real time quickly become impractical. This is why automated real-time outlier identification is necessary for extracting actionable insights out of social media data.

One key advantage to automated outlier detection, is the ability to scale. Social media is not one platform, but several: Facebook, LinkedIn, Twitter, Instagram, Reddit, Pinterest, etc., and each platform provides its own set of data for analytics. On Twitter, in addition to the statistics of the hashtags mentioned in tweets, there’s also the number of times a particular tweet has been retweeted (shared), the number of comments on the tweet, and the number of likes. Since each platform offers many different metrics, finding outliers in the data from all the major platforms requires a system which can handle multiple metrics and still accurately find outliers.

That’s quite a tall order, but modern solutions built on multiple layers of machine learning are available, and they can help companies mine social media for the vital insights they need to maintain their brand’s reputation, spot new growth opportunities and outmaneuver their competition.