Manufacturers need to take a pragmatic approach to the Internet of Things, working back from concrete business problems that they genuinely need to solve, explains Pentaho’s Davy Nys.
Well into the 20th century, coal miners took canary birds down into the mine tunnels with them to detect poisonous carbon monoxide levels. The poisonous gas would kill the canary before killing the miners, raising an ‘alarm’ to exit the tunnels immediately.
Fortunately for canary welfare, today’s miners now use sophisticated infrared and catalytic ‘heat of combustion’ sensors to detect poisonous and combustible gases.
Hook these sensors up to the internet and now a safety officer in a remote location can monitor and even predict potential hazards before they have a chance to escalate.
Welcome to the Internet of Things’ (IoT’) ‘killer app’ – predictive operations and maintenance.
Although this application is IoT’s sweet spot today, various stratospheric market forecasts are based on there being enough applications to create an economy nearly twice the size of Germany’s 2014 GDP ($3.9tn).
In fact by 2020, IDC predicts that the global IoT market will hit $7.1tn and for IoT data to account for 10% of all available data.
As you would expect, increasing numbers of companies want to claim their share of this IoT bounty, but many are unsure of where to begin and how to scope projects.
Searching for articles online there seems to be no limit to potential IoT applications out there, I’m concerned that we’re all going a bit ‘sensor-happy’.
I recently read that dairy farmers are putting sensors in cows to “detect oestrus (when a cow is amorous) and when they have increased milk yield.” This is all well and good, but remember: just because you can read sensor data doesn’t guarantee success.
I urge companies to take a pragmatic approach to IoT.
Rather than start with a huge, elaborate IoT concept, work back from a concrete business problem that you genuinely need to solve.
The dairy farm I mentioned might, for example, examine its supply chain and discover that its main concern is not milk yield, but spoilage and waste.
In that case it would make more sense for it to deploy sensors to help keep inventory fresh during transit and storage than using them to help raise milk production.
Adopting a pragmatic approach that focuses on one concrete business problem at a time gives you the space to explore and solve some of the unique technical issues associated with IoT data analytics.
Five common examples:
- Data volume and “noise”: IoT often involves finding a needle in a haystack to act upon. However depending on the application, ‘things’ can generate huge data volumes. The challenge here is to filter out the noise and find those genuinely important ‘needles’. This is why many companies are layering on ‘stream analytics’ and ‘process analytics’. The first provides real-time information from streams of data such as clickstreams, logs and metering data; and the second provides more actionable information from these streams. Another way to cut down on ‘noise’ is to take regular ‘snapshots’ or ‘windows’ (time intervals) of machine data, rather than unleashing the full torrent of real time data.
- Urgency and latency: There are different levels of urgency and latency requirements with IoT. This is important to understand because people expect to interact with the ‘real world’ in ‘real time’ so many events call for zero latency. Still, other IoT information may not be needed ‘just in time’ – such as the data regularly gathered to continue to refine and improve the predictive model itself. That could potentially be collected and processed a few times a day, for instance. Real time and batch processing architectures both have a place in the IoT.
- Variety of data due to missing standards: Although industries are actively working to address this, a major IoT concern today is the lack of standards. Crucially, there are limited standards governing the types of data generated from different sensors. This means that similar devices from different manufacturers can use completely different data formats and generate data at different frequencies. To cope with the massive variety of data types today, your analytics platform needs to be open and device-independent.
- Data blending: Companies want to collect data from ‘things’ and blend it with relevant relational data. Of course it’s essential that companies can trust the data on which they’re making critical decisions. This ties into…
- Data lineage: This is essentially the recording and authentication of data’s ancestry and veracity. This is particularly important to maintain data ‘health’ and provide an auditable chain of custody for the data.
Manufacturers should be looking for IoT analytics solutions that can help address all of these issues.
Most will need tools that can blend, enrich and refine any data source into secure, analytic data sets and offer different levels of analysis for end users and applications.
Collaboration between IT and subject matter experts will be more important than ever. You won’t just need people who understand data, but expert analysts who understands data in the context of specific devices or sensors.
Whereas any analyst can understand data in the context of simple business KPIs, only a veterinarian would be able to explain what a cow’s temperature swings mean in relation to fertility and milk yield.
And finally, just because I advise companies to take IoT in stages, I’m not at all suggesting that you shouldn’t be ambitious.
As part of Hitachi Data Systems, Pentaho has set its sights for IoT extremely high – to build ‘Social Innovation’ systems designed to leave the world a better place for the next generation of people…and of course, the humble canary!
Do you feel overwhelmed by Business Intelligence (BI)?
BI is taking a pivotal place in a company’s success, but is evolving at a rapid pace.
As industry moves further towards service-based thinking, Industry 4.0 and increased volumes of big data, make sure you aren’t left dazed and confused while your rivals take the initiative.
Connect BI blends together a unique combination of in-depth case studies and peer-to-peer networking with individuals who are also procuring, implementing and exploring BI systems.
Connect BI will provide you with the opportunity to further your understanding of BI, giving you the upper hand with unrestricted exposure to both peers and the vendor community.
The must attend event to explore potential business partners, building much sought after connections. With Connect BI, you can condense six months works into just one day.
14 June, 2016: London venue – TBC