If 2016 had a buzzword in the tech community, it surely must have been the Internet of Things (IoT). Like the cloud a few years back, IoT has been on the tips of the tongues of CTOs across every vertical this past year, and for good reason. This past September, Gartner forecasted that 6.4 billion connected “things” could be in use by the ends of 2016, and will reach 20.8 billion by 2020.
For organizations ready to jump onto this IoT trend, the challenge lies in knowing where to start. Incorporating new sensors or extending current architecture to collect data is one piece of the puzzle. The larger, more complex piece is figuring out what to do with the data they collect with these sensors.
For many companies, this data will come at them like water from a firehose – fast, uncontrolled, and with a risk of drowning in it. As such, the combination of too many things and too much data can leave businesses in reaction mode, scrambling to figure out what happened and why, leaving no time to predict and prepare for what could happen.
Most organizations are already sitting on immense amounts of data that has either been collected through devices or via human-generated processes, sometimes for many years. Once this data is ingested, it’s stored and filed using data management. For companies wanting to incorporate sensors or other IoT technologies into their businesses, this existing data can’t be ignored or left to stagnate. Without getting a good handle on what’s already happened, it’s nearly impossible to predict what will happen, and many organizations are stuck working in hindsight.
We’re seeing a positive trend towards embracing advanced analytics, something we foresee continuing into the new year. Companies are finding ways to move forward by implementing machine learning, which collects data in real time and adjusts itself based on the data it ingests to prevent error and improve efficiency, and advanced analytics, which pull insights and trends from existing data. Combined, machine learning and advanced analytics provide foresight to stay ahead of not only issues, but competitors. Once this happens, companies can automate actions and perform visualizations to make better business decisions, drive value and solve real business problems.
Moving from data to intelligence to action is an approach that we take at New Signature, and we will continue to take into the new year. We help our customers define their business problems and identify their current state of processes using a data science-driven methodology.
This involves several steps that start with ingesting the customer’s data, and then importing it into Azure Machine Learning Studio. We then explore and visualize the data and prepare samples, generating and selecting features. From here, we create and train ML models, fine tuning as we go. We then deploy the model, publishing it as a Web service, consuming the model programmatically, and consuming the model in Excel to ensure it’s pulling the information required. At this point, we’re able to transition our customers into their ideal future state based on what’s been defined throughout the process.
Any organization thinking about IoT in 2017, needs to start small and think big, which is why we’re offering a package for our customers. This includes:
- a data set assessment;
- validating scope;
- data science process iterations, which include:
- preparing data
- exploring and visualizing data
- generating and selecting features
- creating and training ML model
- and deploying and consuming model for testing;
- and visualizing results and developing an implementation plan.
This process will get you on the right path to IoT success in a manageable way.