IoT is driving huge data growth. Enterprise architects need to think about implementing an IoT analytics service for cloud applications in order to generate new business value.
The growth of the Internet of Things is leading to a significant rise in the amount of data processed and stored by the cloud applications behind IoT applications. Enterprise architects need to be proactive in instrumenting the APIs behind these applications in order to maintain a competitive edge, said Mike Gualtieri, principal analyst at Forrester Research Inc., at the WSO2Con conference in San Francisco. This will make it easier to implement IoT analytics processes that improve customer experience and deliver new business value.
"Our analysis shows that companies are only analyzing about 12% of the data they have," Gualtieri said. "Data is all moving fast, but for some reason, most analytics is done later. This is a function of this BI descriptive mentality. I am not saying you don't need that. But now you need to look forward to predictive, streaming and prescriptive analytics as well."
The bulk of business intelligence (BI) applications have focused on tools for aggregating and analyzing information in conjunction with data warehouses. This approach can help steer business at a strategic level but does little to improve business operations at a tactical level.
Think like a venture capitalist
Perishable insights can have exponentially more value than after-the-fact traditional historical analytics.
Mike Gualtieriprincipal analyst, Forrester Research
"IoT analytics is no longer something that happens outside of the application architectures," Gualtieri said. "It becomes part of your application infrastructure. There is a need for an advanced analysis engine that is part of your applications infrastructure and platform"
The use of proactive analytics tools is relatively immature and many organizations don't know how they are going to derive value from them. Consequently, enterprise architects should think about how to flexibly try out new ways of linking together these proactive analytics tools to test out new business processes. "Enterprises need to think more like venture capitalists that invest in a lot of small companies or applications," Gualtieri said. "This is a different way of thinking for companies that want to know the [return on investment] before investing."
Use it or lose it
Traditional BI applications can be bolted on using data warehousing techniques after the fact. But proactive analytics tools need to be woven into the applications and their APIs to be effective. IoT devices like heart rate sensors, activity trackers and automotive sensors are generating enormous amounts of data that can be woven into new applications.
For example, Spotify recently implemented a feature to match songs to a user's run cadence. Other developers are looking at applications that can leverage tire slippage data from cars to alert other drivers to hazardous road conditions.
One of the challenges with leveraging real-time data is that it has a shelf life of value. A service for sending coupons to consumers based on GPS data is far less valuable after they have left the area. Similarly, a fraud detection algorithm does little good after the fact. These kinds of use cases require real-time streaming analytics capable of triggering business processes within tightly defined windows.
"Perishable insights can have exponentially more value than after-the-fact traditional historical analytics," Gualtieri said. "Most companies are not capturing perishable insights. The problem with these insights is that most companies struggle to make these insights actionable. Mobile and IoT are changing that. You have to build analytics into your application."
Building a proactive analytics tool chain
Predictive analytics are tools and techniques that use data to find models. These models can anticipate outcomes with a significant probability of accuracy. They don't predict the future as much as identify probabilities. A good predictive model can recommend products and improve customer messaging.
Streaming analytics is the real-time component used to capture perishable insights. It is a large category of techniques that includes complex event processing for filtering, aggregating, enriching and analyzing a high throughput of data. At a practical level, real time means business time which varies by use case. Financial trading might require microseconds. Fraud detection would be fine with milliseconds, while a recommendation engine could do well with half a second of latency.
Prescriptive analytics are tools and techniques for triggering business rules and business processes in response to analytics data. These allow enterprises to implement new business processes that respond to data in real time.
Staging the analytics services
The new WSO2 IoT platform includes a Data Analytics Server that includes both BI and real-time components that can be baked into IoT applications. It includes a complex event processing module that runs in two megabytes of code that can be pushed out to IoT gateways and mobile phones, or run in the cloud. This allows enterprise architects to explore new application architectures for IoT analytics. For example, a home monitoring app might convert a real-time stream of power line signals into discrete events relating to appliance usage.
In other cases it makes more sense to push data into the cloud for processing. Dmitry Sotnikov, vice president of cloud products at WSO2, said, "It makes sense to analyze data in the cloud for multiple reasons. The edge is more expensive to maintain. The IoT device CPUs like the Arduino are not capable of much processing. Also, you want to aggregate the data so you can analyze it across customers."
