How energy producers are using Twitter analytics to track public opinion

Social media analytics is no longer just for advertising agencies and political strategists. Energy companies have also jumped on board

| follow Jesse Snyder on Twitter

March 03, 2015

Subscribe Email This Post Print This Post Bookmark and Share


At 9:56 a.m. on August 14, 2014, Cody Battershill, an oil sands supporter, tweeted a photo of four pipeline protesters. In the photo, each is wearing a full-bodied white jumpsuit and a breathing apparatus resembling the bulky outfits a hazardous materials cleanup crew might use. They stand awkwardly in front of some heavy machinery at the site of Enbridge’s Line 9 reversal project, a “Stop Line 9” sign pegged into the soil nearby. The sky behind appears overcast and gloomy. “Halloween anyone?” reads the tweet, and then accuses the protest of running counter to the public good.

“It’s like having tens of thousands – in some cases millions – of sensors out there.”

An hour later Mike Hudema, a member of Greenpeace Canada, tweeted a photo of the same pipeline protest. This time, the protesters are readily posed for the shot and the sky glows a warm yellow. The tweet reads “#line9 protests heat up against #tarsands pipeline …”

Both of those tweets, like every other tweet posted that day, were read thousands of times within only seconds of being sent. But they weren’t read only by human followers. Instead, they were analyzed by untold numbers of computer algorithms.

Algorithms are like the mathematical formulas of language. In essence, they are instructions that guide the actions of computer software. They can parse tweets by zeroing in on key words, phrases and emoticons, and ascribe an appropriate measure of emotion to them. By drawing on an enormous database of words, each correlating with a certain level of emotion, either positive or negative, algorithms can act as a kind of linguistic equation.

And those equations are getting smarter. Small teams of people at companies like IHS CERA and IBM are dedicated to the sole purpose of writing algorithms that can detect complex emotions like irony and sarcasm. And by continuously adding both data points and the context in which they exist to those algorithms, they can tell a more complete story. For example, the words “house” and “home” are synonyms, but one conveys a higher level of coziness or ownership. Apply that thinking to a database that includes phrases, abbreviations and emoticons as well as the names of previous events or noteworthy people, and the power of an algorithm increases quickly.

Entities whose success depends on public opinion, like advertising agencies or political parties, have been gauging public opinion through social networking sites for some time. That method, known broadly as social media analytics, is also now being used by oil and gas producers. This typically involves verifying the effectiveness or influence of a company’s online brand or gauging public sentiment toward proposed or existing projects, particularly oil pipelines. But proponents think it could be doing much more. Analyzing two tweets about the Line 9 pipeline reversal, for example, may only serve to show the sharp contrast between opposing sides of that debate, and provides little insight into the public’s thoughts on energy. But replicated a million times over, those voices could begin to take on a form that resembles public opinion.

“It’s like having tens of thousands – in some cases millions – of sensors out there,” says Doug Hanson, the director of a Calgary-based analytics center for IBM, a company offering social media analytics services to the energy sector. “And these are sensors that are smart and observant; they’re thinking, they’re reacting, they’re speaking, they’re collaborating, they’re discussing, all without coming back to any central control point.”

By gathering those sensors and funneling the data they collect through highly complex algorithms, companies can quickly build up enormous databases crammed full of online chatter – an expansive library stocked with all the thoughts, quibbles and noisy disagreements of Twitter’s 288 active monthly users. That data can be highly valuable to oil and gas companies.

Oil producers and infrastructure companies alike are increasingly expected to fulfill the abstract notion of “social license,” a murky and unquantified standard of the level to which industrial development works toward the public good. And while social media analytics can’t help those companies meet that standard, they might help them better understand how far down that road they’ve traveled – and how much farther they still need to go. In a sense, social analytics acts as a kind of involuntary polling mechanism, and one with far larger sample sizes than conventional surveys. With enough data, companies have the ability to tweak their messaging according to how the public is, or isn’t, receiving it.

Social analytics aren’t new, technically speaking. In the past, though, they were internally oriented and focused on things like job satisfaction. But in the last few years, as the number of social media users has exploded, companies have increasingly used analytics to determine the sentiments of everyday Canadians. Which social network ultimately allowed social media analytics to take off? “If I had to answer in a word, I’d say ‘Twitter,’ ” says Chris Hansen, the director of energy insight at IHS.

“If you’re good at mining that data, you can find out a lot about how the public feels in real time.”

– Chris Hansen

Most tweets are part of the public domain – unlike, say, many Facebook posts – and can be easily gathered and analyzed. Every day roughly 500 million tweets are fired off from various locations around the world, some of which are climate or energy related. Over the course of 2014, about 10 million tweets related to climate change were posted in North America. About 1.5 million were posted on the topic of fracking, and one million related to pipelines, Hansen says. “If you’re good at mining that data, you can find out a lot about how the public feels in real time.”

It’s not a perfect science, mind you. People can have different emotional connections to various words or phrases, and while Twitter may offer a large quantity of information, individual tweets tend to lack thoughtful consideration. Many contain no useful information whatsoever. Much of the “dialogue” between users could more accurately be described as snarky jabs or tirades broken into 140-character sections.

Moreover, Twitter doesn’t yet include a wide enough demographic to provide an accurate representation of public opinion. Most users tend to be young, educated and liberal. But younger generations are using social media in rapidly growing numbers. If a majority of citizens are eventually using social networks, it follows logically that the information they contain could come to more accurately represent the thoughts of the public as a whole. “If these analytical tools could help you understand the concerns and help to shape a constructive conversation, then I think it’s another useful tool to add to the tool kit,” says Dan McFadyen, executive fellow at the University of Calgary’s School of Public Policy and former CEO of the Energy Resources Conservation Board. “I hope that it would only be one of the tools in the tool kit, though. I think the tried-and-true method of sitting down across the table and sharing a cup of coffee allows companies to build relationships and understanding. Like polling, you need to understand what it’s telling you and use it appropriately.”

An often-cited example of social license is Enbridge’s Northern Gateway pipeline, which was approved in June 2014 with 209 conditions attached. Many of those conditions call for a more robust relationship with stakeholders along the proposed route. But without an agreed upon definition of what is explicitly required, or who gets to decide whether those requirements are met, the potential for social media analytics may be limited in the short term. The fundamental question is perhaps how government, industry and stakeholders can begin to set fair and meaningful parameters for social license in the first place.

But as analytics technologies improve, it is likely social media will become a rich pool from which to draw data on social sentiment. Finding a coherent or useful narrative from that data is another story altogether. In response to Battershill’s Line 9 tweet, a user who goes by the handle @welloiledgun replied with cheeky questions of his own: “Is it dusty? That’s why they have dust masks? I’m confused. Sandals too? If they want jobs they should gear up, we need hands.”

More posts by Jesse Snyder

Follow @AlbertaOilMag

Issue Contents