Setting out the best strategy to implement data analytics solutions in an industrial environment can be a roller-coaster ride. Where should you focus to succeed?
Setting out the best strategy to implement data analytics solutions in an industrial environment can be a roller-coaster ride. In theory, the key to solving many of the challenges that your business is facing is simple: Data – the more, the better. The constant drive for operational efficiency, equipment optimization, availability, and supply chain, just to name a few of these challenges. All have the potential of being solved through application of modern data science techniques.
WHERE’S THE DATA?
Sounds easy. Then you hit a few bumps. Data sits in siloed repositories scattered throughout the organization with no logical context or connection. Other essential (and immense) sources of operational data are stranded in remote locations. The data sources are a complete Pandora’s box of challenges. For example, how do you handle handwritten maintenance logs where engineers write in a dozen languages using different explanations for the same failures?
One avenue is to take a data consolidation approach. Thanks to the incredible capabilities of the cloud, we have access to huge amounts of cheap storage. This gives you the ability to create a corporate data lake. Gartner defines data lakes as “enterprise-wide data management platforms for analysing disparate sources of data in its native format”. The advantage of a lake is having all Information Technology (IT) and Operational Technology (OT) data stored in a central location, providing one main version of truth for all of your data.
DROWNING IN DATA
This is indeed a noble approach which, unfortunately, can lead to unhappiness. The process of data classification, cleanliness, accuracy, and structure is an immense task. Often it’s a multi-year project, but focusing on the data platform ignores one valuable ingredient - what are you actually trying to solve?
Many organizations avoid the expense and disappointment of this approach by:
- Identifying the business problems they wish to solve in the beginning of their journey, and
- identifying the minimum amount of data required and where it exists.
Yes, we all need a data strategy and yes, over time, a data lake will evolve and it needs to be planned for. However, unless you start with a maniacal focus on the business case you run the risk of being unable to demonstrate business success early on.
LEAD WITH THE END IN SIGHT
The solution is to lead with identifying the top business issues you’re faced with. Then approach them from the viewpoint of:
- Is the solution technically feasible?
- Is the data required for the solution available?
- What’s the impact in terms of margin, revenue, risk, or other key business metrics?
- What’s the cost?
What’s great about this approach is that you can agree on a prioritized set of agile projects –” infrastructure on demand”– delivering business value in a matter of months, rather than waiting until you have your complete data “house” in order. Inevitably, the business units will tire with waiting and multiple uncontrolled skunkworks projects will ensue. By focusing on the business problem you also get business buy-in and avoid frustration with large investments that fail to deliver real benefit before business patience runs out.
The Industrial Internet of things (IIoT) is offering immense insights and gains as the IT and OT worlds collide. There are several articles written about the potential benefits. But to make these benefits real in a timeframe the business wants, we must first focus on which use cases we need to solve.