French firm Michelin has around 110,000 employees across its global operations, and generates most of its 20 billion revenues through the manufacture and sale of around one million tyres each year.
To gain more insight into its operations, the company has made its first steps in a big data analytics project that will process data from a variety of sources and provide feedback to business users.
One example is the ability to monitor defects in products at its manufacturing plants through Microstrategy dashboards, with three pilot projects currently underway in North America and plans to extend this to 67 facilities across the world.
"Each time we have defects on the tyres during the manufacturing process it is stored with a specific defect code, then we can analyse per product, per machine, per operator, how many defects we have in the plant. We know that this kind of dashboard can be updated every fifteen minutes so it is near real-time, so it is ready to steer activity in the plant," said Sebastien Douaillat, BI enterprise architect at Michelin.
"This is one of our first steps to use big data internally in the plants. Our machines are generating a lot of data, a lot of logs, so this is one of the use cases we will have in the near future."
To provide insight into the data generated, Michelin is using Microstrategy analytics tools. "The department is only providing the meta data and the database that is collecting a lot of data from the plant, and then the users are allowed to create their own dashboard," he explained.
Speaking at Microstrategy, Douaillat said that the company is dealing with petabytes of data across its business, with volumes expected to increase as the company brings in information from wider sources to analyse. For example, the company plans to fit its tyres with sensors, and is buying in telematics data from insurers and car manufacturers.
"We are starting to bring in a lot of this data that can provide us with information about the way people are using our tyres," he said.
Social media analysis
Another area where Microstrategy is being used is to support big data plans is around monitoring of social media feeds. As well as producing tyres, Michelin is also known for its travel and restaurant guides, which, alongside its main product set, generate discussion online. For example, there have been 75,000 tweets about the company in the past five months.
In order to assess the company's online reputation and plan marketing campaigns, Michelin decided to use Microstrategy tools to gain better insight into the way it is perceived by customers social networks.
"This is a big data use case because we need to capture, analyse and store a lot of data. There is a lot of information, it is being refreshed often, and there are a lot of different sources," he said.
The company created a prototype in less than two months with open source tools, using Microstrategy to build dashboards.
"We chose to start with Twitter as it is relatively easy to get data from its free API. It is a little bit more complex for Facebook," he said.
The system relies on MongoDB as a NoSQL database to store the tweet information extracted from the Twitter API. The data is then processed in Hadoop using the Hive programming language, with Impala to create queries.
The tool allow the company to view information such as time and data of tweet, geo-location, number of tweets, what industry influencers are discussing, and monitor positive or negative keywords being used in relation to the company.
"For example, this can show that a certain time is better for sending a particular tweet as part of a marketing campaign."
There are now plans to incorporate other feeds, such as Facebook, and integrate the R programming language into Microstrategy to create more robust algorithms to enable customer sentiment analysis.
"We were looking at whether words were positive or negative, but if you are doing only that and you see Michelin near a word that is good or bad, it does not mean the whole sentence is good or bad," he said.
"We are starting to look at sentiment analysis, and are trying to build a more advanced algorithm. For that we need to use R, which is a very complex language, and then we will see the results in Microstrategy."