IoT data from connected products not only provides business value in real-time but also offline, as historical data is analysed to understand usage and detect and predict anomalies. At Waylay, we already provide an ETL service and a BI service to help our customers take advantage of historical IoT data. But in order to accelerate time to value even further, we have now built an end-to-end time series analytics software module to complement our real-time data orchestration offering.
And since automation is in our company's DNA, we designed it as an integral part of our automation platform, in order to enable customers to immediately put analytical models to operational use, for their specific IoT business case. Analytics and automation working together is a powerful differentiator when it comes to increasing the return on IoT investments.
What is time series analytics (TSA) for IoT?
Time series analytics (TSA) for IoT represents a set of methods used to analyse historical IoT sensor data stored as sequences of time-stamped values, in order to extract meaningful statistics from it. It makes use of mathematical algorithms to predict future events based on past data.
Once your IoT install base starts streaming data and you store it into a time-series database, it quickly adds up to high enough volumes that it becomes statistically relevant. By looking at the data describing the normal functioning of an IoT device, trends and patterns can be identified, that can later serve to identify anomalous behaviour that is outside of the predicted range.
What prevents companies from seeing business value from time series analytics today?
In order to see benefits from analysing time series data, there are a number of steps you need to take, that start with access to the data.
- You have to first have the infrastructure in place for data collection, storage and visualisation.
- As any other advanced mathematical challenge brought about by big data, time series analytics falls within the realm of data science. Building an in-house data science team or finding an anomaly-as-a-service offering on the market would be the next step towards reaching your goal.
- The in-house team or external provider can now begin their work of understanding your data, building and testing anomaly detection and prediction algorithms.
- Once the algorithms have proved sufficiently robust against the data, you would now need to create an interface that integrates with the back-end of your TSA service. A process very prone to errors as this is usually done by completely different teams using a different tool set/ language.
- When this connection is up and running, you can finally start thinking about operationalising it or putting it in production, and creating business rules to call the TSA service for anomaly detection, threshold crossings etc.
- You may find it useful to have the insights or notifications delivered via your business channels towards your most relevant teams. This means integrating with your ticketing system, your ERP or mailing service or third party providers or partners.
All of these steps are time consuming and significantly slow down the process of seeing business value from analysing offline IoT data. At each of these steps things can go wrong and potential blockers can drag the process even further.
The Waylay Time Series Analytics accelerates time to value by enabling users to operationalise AI algorithms
The Waylay Time Series Analytics module was built to reduce the time it takes to operationalise time series analytics to mere minutes. Business analysts don't need advanced data science and statistics knowledge to use it and can easily put mathematical algorithms to work inside the IoT business workflow.
Within the Waylay TSA Designer, users choose the IoT devices for which they want to analyse data, explore the data, choose the type of algorithm to work with or let the system auto-magically choose it for them, test the algorithm against the data, and save it as a sensor block. All this is done in minutes thanks to most of the data science in the backgound being automated by the Waylay TSA module.
Pictured above: The Waylay TSA Designer configuration panel
Once the TSA model is saved as a sensor block, the user can now move to the Waylay Rule Designer and include the newly created sensor in the automation rules.
Pictured above: The Waylay Rule Designer, including the TSA model as a sensor input for an IoT automated workflow
Read this use case description of using Waylay TSA to optimise production yield of a fish farm with a step-by-step video and text guide to how to create and deploy an anomaly detection model.
For a detailed description of how the Waylay TSA Module works, have a look at our technical documentation page.
Analytical models are a means to an end, automation keeps the focus on the business results
At Waylay, we believe in automation as the key technology to drive business value and growth from IoT investments. Our view on automation extends beyond restrictive definitions that apply to only parts of a technology stack. For us, automation enables uninterrupted and unassisted interactions between all parts, with the central goal of taking business-relevant automated actions across sytems.
To this end, the end results of performing Time Series Analytics on historical IoT data, namely the trained models, are a valuable input into the automation system. Within the Waylay automation plaform the analytical models sit side-by-side with the real-time event streams, the meta-data models, data from back-end IT systems or cloud apps. Users have a significant advantage by being able to easily train and configure analytical models and then immediately put them into production from the same integrated platform.