Mining M2M big data: dig deep to find the gold
Cisco Visual Networking Index estimates that by 2014, the average M2M module will generate nearly half a gigabit of data a month. Some machines generate many times that.
The question is, what do you do with all this data? Is it simply a vast trove of useless bytes, or can you mine it for insights to help your business grow?
M2M data can be defined as a subset of the vast amounts of data generated by billions of connected sensor-wielding devices. It includes machine sensors sharing various breeds of diagnostic, location or maintenance data, web clicks, or even solution-specific information such as stock or stock condition levels.
Some of this M2M data is generated by consumer devices like the Fitbit or Withings scales. And then there are vast swathes of industrial data from smart factory equipment, servers, intelligent agricultural systems, power grid and meters, RFID tags and industrial equipment. It’s your office vending machine asking for a service, the movement of baggage through an airport, or the power grid switching to new sources of electricity to cope with spikes in demand.
M2M data might include location, time, or date. It may include temperature. It could even change in response to extrinsic events: for example, population movement, traffic flow or an electrical storm, though such relationships are unlikely to be articulated or recognised without comparative analysis.
And this data accumulates fast. The sensors inside a Virgin Atlantic Boeing 787 generate half a terabyte of data per flight. The airline stores this information in the cloud while it works with big data vendors to develop useful analysis of this data.
"As you move to a big data world you can start to see the trends in that data. You can move towards predicting what will happen with the plane so that you can do maintenance before a problem occurs, or look at where the efficiencies are and find out how to fly the plane differently to get better fuel efficiency,” says Virgin Atlantic IT director David Bulman.
United Parcel Service (UPS) uses data generated from across its fleet to improve business efficiency. This includes engine performance, fuel efficiency, driver performance, travel and driver-inputted data. Use of this data has enabled UPS to save 39 million gallons of fuel since 2001, and helped the company develop new services for customers. One service developed in response to this is MyChoice which lets customers adjust delivery time and destination at any time using a smartphone.
The key is to comprehend large and diverse data sets. The LA police department experimented with a mathematical model that was previously used to predict aftershocks during earthquakes. They feed the model with crime data taken from the last 80 years. Based on recent crime locations they found that this enabled them to predict hot spots in which crimes may take place before they did take place. They arranged local police patrols in response to this information, and saw a 12% reduction in property crime and a 26% reduction in burglary as a result.
Another example of combining data sets is UK retailer Tesco which combines 100TB of M2M, retail, and customer loyalty data with third party sources such as weather reports inside its big data store, crunches and munches, then has insights which help plan what to stock in its stores, saving £6 million each year in wasted stock.
Now the whole point of M2M is to understand the location, status, environment or actions and to perform an action, such as trigger stock refresh or recommend a visit to a garage. That's the primary goal. The power of M2M big data will be unlocked when enterprises can mine further and identify patterns among the seeming chaos, the data that's left behind or was thought to be irrelevant.