Prediction as a Service

One of the promises of the Internet of Things (IoT) is that it allows things to speak. Or even more precisely enables products and services that listen to things. Product maintenance is a great example of a such service. Imagine a lighting system on an airport. There's probably 50 thousand lights installed and statistically around 10 of them may be failing every day. Now having a service that precisely pinpoints the 10 that have just failed (or a single one as it fails) is clearly a value.

But this is passive, reactive, poll-type device monitoring. Ping it periodically and once it stops responding, assume it needs to be checked, as it has probably failed.

But actually if things can speak, we can shift from passive and reactive to proactive monitoring. A thing may be able to indicate that it will fail. If it is a lamp, it may report it has worked for a given number of hours at full power, so the performance has significantly degraded. If it is a power supply, it may report it has experienced input voltage spikes that may affect its lifetime. If it is a pump, it may report increased level of vibrations, indicating a bearing is about to fail.

The platform that listens to all those reports may be quite sophisticated. Usually it has way more processing power and storage and machine learning capabilities, so it can make sense of all the data it captures. In most cases, looking for anomalies, because the real value is in unexpected data. Expected data you may usually discard.

It all wraps nicely into prediction as a service. People maintaining the lighting system on an airport will appreciate predictive notifications about a light that may fail. People maintaining the water heating system will appreciate predictive notifications about a pump that may fail. The beauty of this approach is that multiple predictive analytics platforms / services can compete, offering different service levels for different price points. More competition means better products and services, so even more value to customers.

Prediction as a service is definitely one of the most promising areas IoT is going to bring us.