in 2011, sentiment analysis for Twitter was just beginning to phase into mainstream business conciousness.

So, when London-based DCM Capital announced a hedge fund based on Twitter sentiment it was a bold move that ultimately didn't gain the exposure the company were looking for, and the project closed after a few short months.

The company has continued to refine its algorithms however, and recently spoke to Bloomberg about their latest project: A spread-betting platform based entirely on social media sentiment...

There's a huge range of tools out there that claim to be able to tell you if you're hot or not with tweeters of course, from free tools like Klout and Kred, up to large paid-for systems designed for enterprise usage, but so far I've yet to see a single one that can accurately track human emotions, let alone in real time.

Generally speaking, human interactions are just too complicated, and machines will always struggle to track variables like sarcasm (and we're talking about the internet here, so snark is par for the course).

Despite these problems, accurate analysis, especially tied to provable financial ROI, is the holy grail for many marketers. More and more we're hearing that businesses want to use social to directly affect the bottom line (We do as well, we are a business after all). 

In the world of finance analysis moves at incredible speed, often faster than human minds can deal with. Trades are made based on miniscule changes mapped automatically, with automated bidding making split-second decisions alongside human operators.

One of the clear attractions of social media to the financial world is its immidiacy. We're constantly told that social is real-time. Twitter is a live data stream, which makes it perfect for customer response. Get your customer satisfaction levels up and you're stock will rise. At least, that's the theory.

Unfortunately when it comes to extracting truly valuable insight, the real power lies with those who have the time and inclination to look at social as a data store as well.

Sentiment often isn't a key indicator in financial performance if viewed in real-time.

Customers often blow up on Twitter over small transgressions. Once the problem is solved, then very few of those complaining would ever take the time to thank the community manager who helped them out.

If we have a problem we tend to make it public, so people will often say "Hey @CompanyX, the product I ordered hasn't shown up!". Nice and visible. Once @CompanyX fixes this though, if customers bother to reply then it will likely be "@CompanyX thanks, that's great" Rather than "Hey everyone, CompanyX do a great job". This is par for the course, getting genuine evanglism is a lot harder than getting a positive response

Likewise, social sentiment isn't the be-all and end-all here.

Social responses only cover a small preportion of customers with the inclination to get in touch or voice their displeasure on social platforms. Many of us simply don't take the time to hunt out a Facebook page to really complain.

I only tend to turn to Twitter when I get fed up with being stuck in the call line (and usually I get an almost immidiate response. Twitter is the que-jumpers medium of choice). This means that by the time a complaint actually hits Twitter, I'm getting pretty exhausted with the regular channels, so I'm already riled up and ready to rumble.

If I moan about you on Twitter, you can be pretty sure I've already tried to moan somewhere else, so social complaint levels are often inherently skewed. 

The key for businesses on Twitter is not to rely on "Positive/negative/indifferent" as an indicator, but rather to use a variety of tools across different channels to judge sentiment over time. Likewise don't just wait for customers to tell you; go out and ask them. And always question your data and consider underlying bias carefully. 

IF DCM can really accurately map stock prices through social they'll have something very special indeed on their hands, and it will be interesting to see how this plays out in the coming months, but untilthen it'll be a brave investor indeed who puts their money where their Retweets are. 

Do you think sentiment analysis is accurate enough to base real transactions on? I'd be fascinated to hear from any of you who use sentiment systems in the comments. 

Matt Owen

Published 14 January, 2013 by Matt Owen

Matt Owen is a marketing consultant based in London. He was previously Head of Social at Econsultancy and currently runs Atomise Marketing. Opinions expressed are author's own.

204 more posts from this author

You might be interested in

Comments (2)


Eoghan O'Neill, Social Listening Analyst at Ipsos MORI

Great article Matt. Like many others, I suspect, I was excited by the initial approach that DCM took, and the academic paper behind it, without having robust enough scientific knowledge to be able to evaluate it.

Which makes this post VERY interesting for anyone in a similar position to me:

I think that one of the tricky things about sentiment is that it can be either very general or very specific, depending on the granularity of timescale (etc) that you're looking at. Social media sentiment, when it's looking at (say) a brand, as presumably any individual stock-based approach would be, can be centred around news stories, or individual experiences, or overall "general" sentiment towards a brand or parent company. Indeed news stories could be around share price itself, so there's the danger of a self-fulfilling prophecy.

The conclusion of the original Bollen paper was interesting enough - a strong correlation between social media sentiment (of the entire social media universe) and overall stock market performance (with a lag of a couple of days) but that's a far cry from saying that on much shorter timescales for individual brands. Not to mention all the well-trodden challenges around automated sentiment systems. Still, fair play to DCM for carrying on trying - instinctively it feels plausible that there could be a link (and for anyone who cracks it, there's potential £££ of almost mythical proportions!)

Anyhow, very interesting thoughts Matt.

over 5 years ago



I've seen the DCM Sentiment indicator mentioned elsewhere online. I'm a bit skeptical of such sentiment based approaches to be honest. Though I have heard of hedge funds doing such sentiment analysis to take positions on companies etc. There seems to be times when twitter sentiment data will not align with general sentiment, for one twitter is not a representative sample being overloaded with a young audience who may have particular negative sentiment towards certain brands etc. Some minor use of sentiment may make sense though.

over 5 years ago

Save or Cancel

Enjoying this article?

Get more just like this, delivered to your inbox.

Keep up to date with the latest analysis, inspiration and learning from the Econsultancy blog with our free Digital Pulse newsletter. You will receive a hand-picked digest of the latest and greatest articles, as well as snippets of new market data, best practice guides and trends research.