The most memorable stories from the California Gold Rush focus on the lucky souls who struck it rich. Perhaps just as interesting were those who supplied the tools that the miners needed to be successful -- the picks, shovels, and pans. Some of these suppliers were amazingly innovative and entrepreneurial: Levi Strauss (along with partner Jacob Davis who patented the copper rivet reinforcement in the heavy-duty pants) famously supplied denim jeans. More than 150 years later, Strauss’ company is still successfully supplying jeans, and more, to people around the world.
We’re now in a new gold rush of sorts and the treasure is the ability to deliver more value from industrial data. In everyday terms, this means, for example, predicting when equipment will need servicing (optimizing the revenue/cost model) instead of waiting for equipment to fail and then disrupting production to make repairs. Or creating algorithms that tell you how to configure a machine to be more efficient, to generate more power, or maybe to shut it down because it is costing more to operate than it is saving.
These aspirations are gold for the owner-operators, but we’re still early days. We need better tools. Peter Norvig, Director of Research at Google, is often quoted as saying “We don’t have better algorithms. We just have more data.”
Asset operators have been collecting data for years via their historians. But trust and context are critical to understanding that data. Take the following random data set:
What does it mean when my reading is 0? Does it mean I have an intermittent sensor problem? That the asset is offline for service at that time? Or, within the context of upstream or downstream processes, is this an entirely expected result and I don’t need to worry about?
Now let’s extrapolate these questions and problems from 35 data points to 100 billion or more with new points added every few seconds. Panning for gold just got a lot more complicated.
According to McKinsey, less than 5% of all industrial data is reviewed and analyzed. It is a trade-off of the specific industry and expertise required to consume and gain insight. The same people who today manage the asset are the most qualified to consume, comprehend, and optimize. The smartest data scientist cannot make up for the lack of hardware insight and access to trusted, aggregated data.
There are companies out there using data scientists and trying to move beyond aggregation and build better algorithms. Compute technology today is certainly better at processing huge volumes of data. And the algorithms required for modeling both assets and processes will improve over time.
Underlying all of this insight is the expectation that the data is combined, managed and shaped to answer questions being asked by people and applications. In the simple example above, contextualizing the time series data with service history, whether data, and process information would all shape the answer. This context creates a truly robust asset twin that fully reflects the reality of the situation.
But none of those applications can work without being able to reliably shape and trust all the data out there.
5% won’t cut it anymore.