The problem of measuring sentiment on social

The use of social media has rocketed over the last few years. According to a report by Statista, in 2020, almost half (49%) of the global population uses social media. Furthermore, other figures show TikTok alone gains eight new users every second.

However, vastly increasing numbers of social posts means more information to analyse,monitor and measure. In fact, experts suggest that data creation and replication will soon outgrow storage capacity.

Not only will data storage become an issue, but data analysis is already becoming increasingly difficult and burdensome, as highlighted in the Making Sense of Mass Mentions whitepaper. PR professionals and marketers need an effective and scalable solution.

The problem for PR

Making matters worse, we now live in the age of ‘cancel culture’ and boycotts from a passionate and confident generation, where missing one negative opinion could lead to thousands. In 2021, 80% of people would boycott a brand based on widespread, negative information they see online, regardless of the facts. Reputations are on the line, and brands are under more scrutiny than ever.

In these circumstances, quickly and effectively identifying a change in tone and sentiment can be the difference between a win and a crisis. Deloitte found one in four people moved away from brands that they believed acted in self-interest, meaning PR teams have to work even harder to monitor and manage brand sentiment online and also judge their responses carefully.

Authenticity to a brand’s established values is the golden ticket for building trust. Butwhat about issues that are truly new to a brand and where there is no established ‘position’. For example, the conflict in Ukraine, how much do a given brand’s audience care and what do they expect the brand to do, if anything? The speed at which user-generated content is being created is proving challenging for teams to keep up with, where every second counts if a response is going to be effective.

Traditional media monitoring is not enough

Traditional methods of media monitoring and social media listening aren’t fit for purpose. The standard use of machine learning in such services is the use of ‘sentiment analysis’. These algorithms analyse the words people use to share their opinions to determine if they’re positive, negative or neutral, but these models don’t understand context. They can’t tell whether people are using positive language to describe negative outcomes, or vice versa. We all know sentiment analysis has its flaws yet, without any alternatives, the industry has become desensitised to its failings.

Context is king. Take our Ukrainian example. Almost every opinion shared on the topic will score negatively for sentiment, so how is a PR team supposed to analyse the huge volume of content to determine how any specific brand should act and whether that action, or inaction, has made any impact on the brand’s reputation? With simple sentiment analysis PR teams are unable to determine the real intent behind tweets, posts and conversations, leaving PR teams with a mountain of data and no effective way to analyse it.

Stance analysis is one solution

New AI models for automatically identifying the specific trending narratives around brands enable millions of comments to be analysed in seconds. These narratives can then be measured for their sentiment within context, enabling much more accurate and actionable insights for PR teams.

The latest research in natural language processing has made these ‘stance’ models more sophisticated, enabling brands to understand the true intent behind audience conversations.

Stance models can identify the subject of an opinion and the lexical cues that indicate a positive or negative stance towards any given topic. Analysing stance over sentiment means PR teams can now identify specific influencers that are ‘for’ or ‘against’ a brand or topic as well as whether the number of influencers and stance within a topic is increasing or decreasing.

By using this intelligence to identify trends as soon as they begin to develop, PR teams can assess the risk of content, driving efficiencies and increasing the quality of understanding conversations, perceptions, and intent.

They can get deeper and more complex insight into audience opinion to pre-empt stories before they potentially threaten brand reputation, and they can do it quicker than manual analysis or relying solely on sentiment.

When each and every opinion counts

With user-generated content shared via social media accelerating, brands are under incessant scrutiny when it comes to their take on societal and political issues, especially in light of recent global events. Without the appropriate monitoring tools in place, brands are at risk of sleepwalking into a catastrophe.

Opinions and their context for brands are more important than ever before for PRs, whether it’s analysing a campaign or gauging an audience’s opinion on a new product launch. This calls for a new solution, as relying on sentiment analysis and manual monitoring is no longer a viable option for PR professionals and marketers.

Stance analysis can help brands by using enriched data points that go beyond what traditional sentiment analysis has to offer. Harnessing the power of artificial intelligence can help PR teams save their clients time and money, reduce risk and create opportunities.

Mass mentions can now be monitored intelligently at scale, making it imperative for brands and agencies looking to deliver faster and more decisive action. Leveraging the power of AI will put PR teams and their agencies back on the front foot.

Written by Antony Cousins, CEO at media monitoring specialist Factmata