"I like my brand new car".
"Driving my car is like driving a tank".
"Like it or not, you need a car when you live in the country".
If you are a car manufacturer, knowing how many people like (or don't like) your new model is pretty useful. How many people like it? Is that better or worse than the previous model? Do they dislike the colour? Handling? Do they like the price? Any human can tell the difference between the above three statements - trying to get a machine to do this is not quite as easy.
In the time before social media, consumer focus groups were the medium through which this qualitative data was collated and fed back to product and R&D teams. Marketers had to venture out and discover these opinions. Now, the opinions are out there; marketers just need to find a way to tap into them. A series of focus groups may gather a richness of data from, perhaps, dozens of consumers. But with about one billion tweets every week, and growing, how about a richness of data from thousands? How about from hundreds of thousands? As consumers become increasingly socially connected and vocal about brands and products, this type of insight, at this scale, can be immensely valuable.
While the technology behind sentiment analysis has been around for some time, appetite for using sentiment analysis in making strategic and operational decisions – perhaps in deciding which products to launch – is gaining traction. Advanced sentiment analytics models can now score sentiment to a level of accuracy approximately the same as humans (about 70-80% by the way – even humans disagree...).
However, scoring sentiment accurately about a business or brand is not enough. To be actionable to a business, models must distil the expressions of sentiment down into more detailed dimensions. Which specific product is a customer talking about? Is the customer annoyed at the price or the colour? How do consumers feel about your corporate responsibility? Do customers like you? The combination of structuring this 'noise' into discrete dimensions, scored accurately, is where the insight starts to become really powerful. Yes... start. Take a business dataset – revenue data from different product lines perhaps – and compare it to sentiment data from the same product lines. You may find something quite compelling there.
I may sound like an evangelist, but by no means is sentiment analytics a panacea for businesses, nor is it easy to do (don't be deceived by free online tools). Further, there may be challenges for certain industries which are seen as commoditised, where volume of opinion about specific brands is low. Or there may be internal challenges in convincing sceptical senior stakeholders that social media matters (it does) or that advanced models can live up to said claims (they can).
That said, it is important to understand the difference between a 'Like' and a like... and now we do.
Click here to review SentsCheck - an exciting, new digital sentiment analytics solution and quarterly index for the UK retail banking industry, launched recently by Deloitte.
Jayson Robinson
Jayson has written a number of previous posts including:
Discriminalytics,
Live in the Cloud and
Cloudforce '12 wrap-up. Jayson is in the Customer Analytics
team in Consulting and has worked across industries, using
business data to maximise value within the business and with their
customers.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.