Big data mining for product feedback

Darren Baguley

Wednesday 6 April 2016

People talking about what they do and don’t like about a product on social media, internet forums and the comment pages of reviews has become a valuable source of commercial insights.

When ‘big data’ first became the new, new thing, it was initially about the standardised types of information that all organisations hold, such as inventory, sales, accounting and financial records. Many organisations now have the technology, people, processes and procedures in place to analyse these types of data, while the early adopters are taking it to the next level – unstructured data. While this can be unstructured data within the organisation, those at the cutting edge are using social media, review sites, call-centre information and user forums to extract data on their products.

That supposition is, however, somewhat tendentious. Some people believe the hype surrounding big data can obstruct its meaning, purpose and value for many organisations. These same people argue that many companies are gathering and storing big data, but they aren’t really extracting any meaningful insights from it. To shedsome light on how big data is being used for product feedback, three experts in the field share their views.

Anthony Volpe, Chief Corporate Analytics Officer at Lenovo

Speaking at an SAS event in September 2015, Lenovo’s Chief Corporate Analytics Officer, Anthony Volpe, talked about how his team was mining text to detect product quality issues early and evaluate products against its competitors months ahead of previous practice.

For Volpe, the most important thing was to be specific when seeking insight. General gushing praise along the lines of ‘I love my new Lenovo X-1 Carbon’ doesn’t tell the team a lot. “When we do text mining, it’s not about whether people like us or not. It’s mining for specific insights. If it’s not for anything and everything, so for specific things, text mining is a much more manageable problem,” he said to CIO magazine’s Rebecca Merrett.

Under Volpe’s leadership, Lenovo built a Lenovo Early Detection (LED) system which applies a statistical process control (SPC) model to data in order to detect potential problems. It mostly mines data from sources such as user forums, call centres and social media. Again, specificity is the key; the team doesn’t just search for ‘Lenovo keyboards’ he told CIO’s Rebecca Merrett, they will look for a particular model of keyboard.

John Riccio, partner, digital services and experience centre leader at PwC

John Riccio thinks many organisations seem to be focused on reporting and analysing performance rather than gaining insight from the data. “There is a lot of hype around it [big data] and a lot of urgency around doing something with data. If you break it down we think the big challenge – especially at the big end of town – is that people feel, ‘I need to collect data, store it and do something with it.’

“Doing big data properly is a combination of art and science. How can we use data to understand the current context and future intent of our customers? Once organisations start making that shift – and it’s a profound shift to make – data needs to be real time and needs to be now. A lot of organisations have set up analytical departments that store a lot of data rather than focusing on the value piece which is insight. Taking that approach may get the wrong outcomes.”

Dan Banyard, managing director at Edentify

“Big data is not really the thing, it’s getting it all together and then it’s understanding the information – they’re the tricky bits. Getting data together is always a challenge but many organisations are quite good at it now and have systems in place for understanding structured data in areas such as sales, inventory and financial. They can pull it all together to look at it holistically and understand it.

“The unstructured data is different, and is also the latest trend. Many companies don’t have the ability to understand it because it’s in natural language which may be telling you all sorts of different things. Computers are terrible at analysing intonations in language, sarcasm, false positives, double meanings – it can be really difficult. For example, the film was terrible but I loved it – it was so bad it was good. This sort of thing ties computers up into knots.”

This piece originally appeared on ThinkFWD