Lenovo is using machine learning to analyse unstructured data from YouTube and Instagram

Lenovo is aiming to build better products by analysing customers' views

The world's biggest PC maker Lenovo is analysing unstructured data from social channels including YouTube and Instagram to help the firm build products with customers' feedback in mind.

Mohammed Chaara, director of the Customer Insight Center of Excellence, Strategy & Analytics at Lenovo, explained that four years ago the firm had been an engineering and product-focused company.

"Whatever engineers and the test and labs team said was the ruling idea in terms of products and strategies, but under the leadership of our CEO [Yuan Yuanqing] we transformed our organisation to become more customer-centric," Chaara told the media at SAS's headquarters in North Carolina this week.

He said that the firm wanted to be able to bring together customer feedback, whether it be through traditional methods such as a call centre, online forums or surveys, or through social media, including reviews on YouTube.

"I've developed an app that allows you to capture the variety of feedback that is expressed by the customer, then structure it, mine it, get context out of it and provide that information back to the engineering team," Chaara explained.

The application has been in production for the past two years and Chaara believes it has yielded success.

Indeed, Chaara claimed that the company used unstructured data when it first built its Yoga device - a mix between a tablet and a laptop.

"We were in a situation where the media said it was the end of the PC and the tablet was going to take over, but our hypothesis was that this wasn't the case. But we needed to figure out what capabilities and experiences people wanted out of mobile devices, tablets and PCs," he said.

According to Chaara, the firm broke down the criteria into four form factors that were common in each of these devices, then it dug into its unstructured data to try to identify the common denominator for all of them. "That is what this product has; all of these in one," he said.

But how does the company analyse the data from pictures or videos?

"The video channel is definitely difficult [to analyse] and images are as well - we also wanted to look at YouTube and Instagram so if someone posts a picture of a Lenovo desktop or a laptop that is broken or doing something, we want to read that," he said.

Chaara said Lenovo used a traditional method which involved a team of people mining and reviewing all this data.

"They're creating a training data set for us so that we can use machine learning and deep learning to go and automatically identify the contents that come out of this," he said.

"One of the rules in our organisation is that I take requirements from my business owners and I translate those into analytical and data solutions. When I get to a point where SAS, my partner, cannot help me anymore and there's nothing in the marketplace, I take my requirements to the R&D team to develop solutions for it - so the example [of YouTube] is what the R&D team is involved in with - how to read video and get the contents out of it," he added.

SAS CEO Jim Goodnight told Computing that the analytics firm is already working on analysing the contents of pictures, and suggested that videos would come soon after. He and SAS CIO Keith Collins explained that while analysing these social feeds were important, machine learning for analytics - what Lenovo's Chaara referred to himself - is of far greater significance.