Posts Tagged open-data

Insight from the Internet of Things


Inference is a key insight skill – looking for patterns in raw primary data to seek new meanings and contexts for understanding. Wearables, nearables and the Internet of Things bring us new universes of fantastic primary data. Compelling user benefits can be obtained by applying layers of inference on to this primary data.

Pattern Recognition

Smappee is a great example of this approach. It is able to detect your household’s electricity usage with all its minor fluctuations as different appliances and devices are switched on and perform their duties. The smarts come with what it does with this primary usage data. It is able to recognize the distinctive pattern, or signature, that each appliance has. For example, a dishwasher’s electricity usage will vary as it goes through its washing, rinsing and drying cycles.

Identifying Outliers

Unless you’re an energy nerd, telling users how much electricity each appliance is consuming is not a compelling benefit. More valuable benefits come as you take pattern recognition to the next level, namely learning the “usual” and identifying outliers and anomalies. For example, has the iron been left on for longer than is usual? If so, alert the user to the potential issue and ideally give them a way to resolve the problem. In Smappee’s case it provides a range of smart plugs that can be controlled from the mobile app to enable individual outlets to be switched on and off.

Iterative Learning: User Feedback

Services like Smappee can continuously learn and improve via two important mechanisms; user feedback and social data aggregation. Any decent learning system needs to give users an opportunity to help it improve and enhance what it learns. For example, what if the alert about the iron still being on was unwanted as the typical household pattern of behavior is to let a mountain of ironing accumulate and then do it all in one lengthy session. If the user can easily flag that the alert was inappropriate the system should be able to adjust and learn more about what represents “usual” for the household it is monitoring.

Learning from the network

Leveraging network effects is also critical for such services to maximize their value. For example, let’s say users are given the opportunity to enter the make and models of some of the appliances identified by the service. The intelligence provided by one user can be used to help the whole service improve its appliance signature detection capabilities across its whole network of users.

Joining Data Sets

Services like Smappee don’t need to rely just on the data provided by its device sensors or from user feedback. Combining other data sets can provide additional tiers of benefits. For example, based on learning the appliances that a household uses and their corresponding energy ratings the service can make specific recommendations as to which ones the household might want to consider replacing to make the most savings on their electricity bill.

Such intelligence would clearly be enormously valuable to marketers. This is where things get more complicated from a privacy perspective, consumers are not going to want to be bombarded with offers for a new washing machine because they’ve shared the fact that they have an aging energy hog whirring away in the laundry. As always the challenge is to ensure that consumers remain in control and are always getting genuine value from the data they share.

Open Data

In many cases startups like Smappee cannot research and develop all the inference and data joining capabilities that the data they are acquiring opens up. The best strategy is almost always to open up their data and allow others to develop applications and services that use it. A complete ecosystem that relies on their primary data is enabled. Users of the service must of course remain in complete control of their data and have full opt-in and opt-out capabilities from any additional services.

From a research and insight perspective using new data sources like this and applying these layers of inference, represents a huge opportunity and we’re only just starting to explore what’s possible.