How Twitter’s Trending Feature Can Be Applied to Every Area of Business

Giving organizations real-time insight into customer sentiment

Companies can use similar algorithms to identify unusual spikes in customer activity.
Getty Images, Twitter

In 2013 Twitter launched its Trending feature — a list of algorithmically-determined popular hashtags. Facebook rapidly followed suit, and since then the trending feature has become a staple of the social media experience.

But trending algorithms have use cases that extend far beyond learning about Beyonce’s twins — from detecting earthquakes and security threats, to identifying real-time trends in customer support queries. Here’s how to apply trending algorithms to every area of business:

Have a global idea of what people are talking about

In 2015, the U.S. Geological Survey began using Twitter’s API to detect earthquakes in parts of the world where there aren’t earthquake sensors. Similarly, finance firms can use sentiment analysis to forecast market movement based on trending topics and sentiments around these topics. Twitter trending can even be used to predict cyber attacks based on news events.

The concept can also be implemented beyond Twitter’s borders: companies can use similar algorithms to identify unusual spikes in customer activity, surges in certain topics coming through customer service, or to identify trending geographic locations for sales. Trending topics can give organizations real-time insight into customer sentiment, allowing them to act in moments that matter most. Whether it’s preparing against a cyber attack, collecting information from remote parts of the world, responding to a problem with a product, or pivoting PR strategy to respond to a news article.

Calculate velocity—and then act on it

Of course, implementing a trending algorithm is easier said than done. But Twitter provides a strong model to emulate.

Twitter’s algorithm polls tweets for repeated hashtags, and determines whether or not they are trending based on volume per time period. In other words, “trending” is based on sharp spikes rather than gradual sustained growth. A one-day spike in the word “covfefe” will be deemed trending, whereas a gradual one-month increase in #MAGA is just basic news.

Developing this algorithm depends on two things: normalizing data, and then determining positive and negative differences. For instance, if you typically have a high volume of customer service requests regarding payment issues, then the baseline for this topic will be fairly high; it would be inaccurate to have it always be trending, as it’s not an unusual volume per time period.

Establish different levels of prioritization through a “trending” algorithm — such as classifying unusual, emergency, or negative topics. On most days, trending issues will be of interest, but not emergencies. Negative trends in customer service, for instance, could be reflective of an updated FAQ section, or the implementation of a ticket deflection chatbot. And while this wouldn’t be cause for immediate action, it provides valuable information. Emergency trends, on the other hand, would show as extremely sharp deviations from the baseline, and would require immediate action.

Identify live trends

Twitter trending has extremely relevant use cases for customer service managers, who can use the same type of algorithm to determine real-time global increases in customer service topics. Currently, other than by simple ticket classification, trends are difficult to spot: individual agents may only see the same problem twice in a day because of individual capacity, but if that same problem is cropping up in every ticket, that’s a trend.

The implications of this go beyond its real-time nature. It gives businesses insight into problems that may have arisen for customers outside of the product. For instance, we all remember the iOS bug that turned “I” into “A [?]” — say that bug prevented users from performing certain actions in your app. While you may not have detected a problem with your product, the trending feature would allow you to see that people were having problems stemming from the iOS bug.

Other, less tangible issues can also be detected via a trending algorithm. Say, for instance, that a viral makeup YouTube video instructs viewers to ask a beauty company via chat a specific question to get something for free. Whatever the wording of the query may be, it triggers a chatbot response that offers the customer compensation. This may not set off alarm bells for any particular agent — but across all agents — it is a trend worth responding to. By using a trending monitor, though, the algorithm can pick up on this repeated ad hoc phrase and report an unusual trend.

Managers can then implement an emergency procedure for responding to a trending query to avoid further detriment to the company.

Listen to the chatter

Consumers today are spread out—they communicate sporadically on different channels, of which social media is just one. And they use information from one medium to inform how they act on another. Businesses need to listen to this chatter, both on and off of social media—after all, a trending topic in your customer service could be the result of a trending Instagram post. A spike in one word in a certain region could indicate a natural disaster. Patterns of text could predict the outcome of an election. Listening to this chatter can give businesses of all kinds valuable information about their customers, and the world in which their customers live.