Reading between the lines

Written by: David Burrows Posted: 18/11/2021

BL75_sentiment1What exactly is sentiment data? And can the insights it gives companies into the way their customers are feeling really help them grow their business?

It might sound like something out of Tomorrow’s World, but sentiment data is a pretty straightforward concept – technology designed to enable businesses, particularly those in the financial services sector, to gauge customers’ mood and intentions. 

The idea is that they can then use this data to flex the products they offer, the engagement they have with customers and the way they communicate.

However, while the concept might be straightforward, applying the data takes careful analysis and clear frameworks. So are the rewards worth all the effort?  

Adam Riddell, Director of Jersey-based Crystal Public Relations, says using digital technologies to assess customer and wider stakeholder sentiment is still an emerging discipline, but is a key area of evolution as organisations try to understand their audiences better, particularly against the backdrop of an ever-growing sea of data.

As for whether the rewards are worth the effort, he believes businesses don’t really have an option. “Sentiment is really important and new tools are opening up new avenues to enable organisations to take great strides in this area. It’s far better to try to know how your key stakeholders are going to react to something than assume or make a best guess,” he says.

“Ultimately, it will be how someone was left feeling that will impact you as an organisation the most. The old mantra is that customers don’t remember what you said or what you sold them, but they remember how you made them feel. Sentiment analysis is a fragile area and data needs to be carefully assessed.”

Data development

Many say the use of sentiment data in the investment sector is far from new. Oswald Lopes, Vice President at fintech specialist Infrasoft Technologies, agrees that sentiment data and its use in future market developments is not a new concept, but says it has taken time for the maturity and mining of the data to get more accurate. 

“There are a multitude of use cases that have already been successful in areas such as market research, customer service and lead identification, to name a few,” he says. 

“The maturity of the technology has definitely allowed for customer sentiment to be exploited based on behaviours – to enable more targeted marketing. 

“Where this has been successful and has proven useful is in the use of sentiment data in adverse media checks, to help compliance teams make informed decisions – this is proving to be invaluable.”

Riddell highlights a couple of things going on here. First, it’s the volume of data – every click, like or share on any one of the growing number of platforms is generating more and more data, and trying to capture sentiment among that data presents a new opportunity. 

Second, this new opportunity is made possible by continued advances in technology, such as AI and natural language processing. 

“Sentiment used to be tracked through social monitoring and listening techniques, using fairly clunky metrics. But the tools now are getting better at understanding sentiment in large data sets – and quickly, too,” he says.

Challenges to overcome

Lopes explains that sentiment data originates primarily from social media or the publicly available data of individuals – their ‘likes’, posts/blogs, reactions and affiliations. 

However, he warns of the need to be careful where the sentiment data being gathered is from a source that is trying to generalise over a cross-section of people – the source of such data may be biased toward a particular gender or age group.

Sentiment data can be generated from anywhere, which, as Riddell points out, is why AI tools that can be plugged into multiple platforms to get a full picture of large data sets are such an attractive proposition. 

“If you are focusing on only a select few platforms, you are immediately cutting out the noise from another channel, and that could prove costly,” he says.

BL75_sentiment2The accuracy of the technology available is still open to challenge, however, and organisations have to exercise some caution when assessing sentiment, he adds. 

“For instance, sentiment can change very quickly, triggered by a single announcement or development. Sentiment data might also be locked to a certain moment in time, so understanding that feelings can change quickly is really important. 

“There is also little in the way of standardisation in the application of sentiment data technologies – some technologies split very basically between positive, neutral and negative; others are more granular. So, maintaining some sort of consistency is important.”

Riddell adds: “Finally, sentiment is very hard to read – sarcasm, poor grammar, typos or double negatives can pose problems for tech applications. In fact, it’s estimated that when humans evaluate a text’s sentiment, they tend to agree only 80% of the time, so the accuracy of human-designed digital tools would tend to fall below that threshold.”

Given these accuracy issues, is the logical conclusion that sentiment data needs to sit alongside other research techniques, rather than as a standalone tool for customer decision-making?

“Definitely,” says Riddell. “The key to managing and understanding data is understanding the limitations of that data as well as the benefits. Qualitative data such as sentiment is really the holy grail for marketers in terms of understanding and trying to predict future behaviours, but it needs to be taken alongside and balanced by other data too. 
“Understanding how sentiment data sits within the field of behavioural science is really important, while using other traditional and qualitative research techniques can help create a bigger picture and frame the sentiment findings.”

Measuring the Covid impact 

The Covid-19 pandemic has clearly had a huge impact on how companies engage with their clients – and on consumers’ habits. So, has the turbulence of the coronavirus crisis shone a spotlight on the value of sentiment data for investors?

Lopes suggests that with the current crisis preventing the usual levels of physical interaction with individuals, the insights provided by sentiment analytics have made the value of such data immeasurable – a trend he expects to become more common moving forward.

Riddell’s take on things is that the experiences of the past 18 months have highlighted the benefits of sentiment analysis, but also its fragility.

“The pandemic has been such a complex and nuanced experience, stirring up so many different ideas and feelings, that keeping track of sentiment has really been one of the key challenges for governments in trying to implement measures that are balanced between health advice and our wellbeing as human beings,” he says.

“That’s been mirrored within the business and financial services spaces, as firms have tried to get a fuller picture of how customers or investors are feeling and how they are likely to respond in a highly charged environment.”

The journey for sentiment analysis is unlikely to stop there. Further angles may yet be explored, such as the application of sentiment analysis on employees. After all, how well do firms really know how their staff are feeling? 

“Remote working has shone a light on the struggles of people working from home – the idea that firms can better understand staff in a more hybrid working environment in the future will be appealing,” Riddell concludes. “I think we can expect to see that sort of application more and more.”

Add a Comment

  • *
  • *
  • *
  • *
  • Submit

It's easy to stay current with

Just sign up for our email updates!

Yes please! No thanks!