Building an insight-driven business
Having huge amounts of information is one thing. Harnessing data to create new business models is another. Alun Jones, Data Scientist at Konecranes, talks about utilizing machine learning, understanding security issues and improving operational efficiency as steps in building an insight-driven business.
Technology evolves rather quickly, while products inevitably do not. There is a challenge to figuring out how to make certain that older pieces of equipment talk to newer ones.
The Internet of Things (IoT) is relatively new. On the technical side, a lack of standards and protocols persists mainly because of its nature as a converging market place made up of competing players who believe that their individual technologies should be the standard.
Then there is what can be viewed as vendor discrepancy, or where a vendor adds something to a standard because they see it as the way forward. There may be little cause for concern about compatibility when using a manufacturer’s proprietary equipment. Keeping up with updates, however, can prove to be a challenge when combining equipment from different manufacturers. Cultural issues surround the ownership and usage of data, which are internal to organizations.
Utilizing machine learning
Machine learning is really where the benefit of having huge amounts of data comes from. What machine learning does is that it can very quickly go through a very large data set and pick up salient points which you then have to verify in real terms. Using algorithms that learn from data in an iterative manner, machine learning allows computers to detect hidden insights without being explicitly programmed where to search. Machine learning methods drive much of modern data analysis across fields such as engineering and the sciences, and commercial applications.
“Machine learning methods drive much of modern data analysis across fields such as engineering and the sciences, and commercial applications.”
Machine learning will tell you that there is a correlation between this and that, which then implies that if this changes in the future then something else is going to happen. I believe that this has to be verified, always. We have machine learning at a very high level that then picks out the relevant patterns in the data which we can then validate against other sources as well.
While machine learning is already an incredibly powerful tool that can solve difficult classification problems, certain points must still be kept in mind. Increasingly, the role of human input in these automated business processes will involve overseeing and tweaking the machine learning algorithms. After all, algorithms are only as smart as the intelligence put into them. This is where the potential for misunderstanding can arise.
Machine learning depends on small “errors” being made by the machine and has to be done and redone over and over. A data-driven hypothesis is first derived before it is tested against new data. When the machine hypothesis is found to be incorrect, the machine then refines the algorithm or hypothesis to suit both the new and old data. The process is an ongoing one.
Don’t worry about security
First, don’t be overly anxious about security. Assume security is going to work because there are professionals who will sort matters out, so don’t allow that be a barrier to doing something.
Second, start small. Pilot – and test and test and test. It would also be wise to learn from other people’s experiences. It can become a bit difficult if you pin your hopes on a form of technology and invest in it, only for it not to work. Try and be as flexible as you possibly can. A particular type of technology may be widely used today, but it might change rapidly. As big data introduces a new level of integration complexity, integration technologies then require a common platform that supports data quality and profiling.
Improving operational efficiency
The next thing, which is the crucial part, is harnessing all this data and coming up with new business models. For instance, the steel industry in the UK is going through a very bad patch. Steel plants have had to close in England, enabling former employees to sell their expertise within the industry. The knowledge that these individuals have, as well as the data they have access to, can be used to improve the steelmaking process and optimize operations.
Once operational efficiency has been attained, the next move would be to think about what else all this data allows you to do.
Another way that data can potentially improve the service-based economy is by making manufacturing more flexible and more small-scale. There are opportunities for small-scale manufacturing using Industry 4.0 or IoT in general. Rather than having to build 10,000 of something, for example, you can build 10. Or you can then have a tailored experience.
Generating value from Big Data, and using it as a key component in a business’s growth strategy, is a matter of connecting data to insights to action in a quick, repeatable way. There are more options we haven’t even thought about yet.
Alun Jones works as a Data Scientist at Konecranes
Image credit: ImageFlow / Shutterstock.com