In the current age of information, organizations of all kinds are increasingly collecting and analyzing large amounts of data on everything from production information to client buying behavior and preferences. These data sets are big. Very, very big. And that’s the scientific reason we call them “big data.” More technically, big data is any high-volume, high-velocity, and high-variety informational asset.
Data analytics is the process of obtaining meaningful information from that data through cost-effective and innovative information processing. It is the process through which we gain deeper understanding of our businesses, and in the manufacturing world, we call this manufacturing intelligence. Here are five steps to implementing a manufacturing intelligence strategy in your business.
You cannot analyze what you don't collect. Develop and implement a strategy to collect and store data from your plant floor and production management systems. Ideally, all this data is kept in one central location where various analytics and visualization tools can access cross-sectional information about your business. Automated data collection is always preferable to manual entry. In a manufacturing environment, data historians from both plant and building automation systems will be valuable sources of plant floor data.
Ask the right questions
In order to analyze any data, you first need to understand what information you want to know. Answering the who, what, when, when, where, why, and how questions is a great basis for this. For example, do you want to know what happened to a certain product lot or do you want to predict a future outcome? The answers to these questions will help determine what type of analytics algorithm is used and perhaps what visualization tools are best suited to display the answer.
Understand data type limitations
After you know what question you are asking of the data, you need to understand the nature of the data you will be analyzing. On a high level, data falls into two classifications: Qualitative data, which includes text, images, audio, video, and opinions; and quantitative data, which includes numerical data and quantifiable variables. The type of data being analyzed helps determine or limit which analytics tools will be useful in answering the question.
Build an analytics tool
Once you know the nature of your data set and what answers you are looking for, an appropriate analytics technique or machine learning algorithm can be applied. While there are some “off the shelf” tools that have built in tools for this (Microsoft Azure Machine Learning Studio, Google Prediction API, Amazon Machine Learning, Tableau), odds are you will need a programmer for this step.
Whether you use one of these tools or fully customize with programming languages such as Python or R, the process is effectively the same in the end. The system will import and analyze the data and give you an answer to your question. The first run (or man runs) is typically off the mark, so you will modify your model and try again until you get reasonable results. This can be a time-consuming process, but once the model is built, you can feed new data in as it becomes available. This is the true power of data analytics and where you will ultimately find long-term value.
Visualize the answer
The output of an analytics tool can sometimes be cryptic until viewed in the right way. Model outputs will most likely be more than a line graph or bar chart can manage. Visualization methods like heat maps, dendrograms, treemaps, and a host of other tools will quickly turn your answer into actionable information. Tools like D3.js are valuable when building your own models from scratch. Software like Tableau and Microsoft Power BI are both analytics and visualization tools in one. Either way, consider using dashboards instead of reports to deliver real-time information straight to desktops and mobile devices.
Strategic, effective application of data analytics will have a tangible, positive effect on your business operations. Outcomes such as better understanding of client needs, increased process automation, recognizing patterns and trends, classification of data, discovering hidden correlations between variables, and predicting the future of certain assets all work to improve and maintain revenue and reduce costs. Big data analytics are becoming increasingly accessible to mid-range companies, and the time is definitely now to implement a data strategy as part of your continuous improvement initiatives.
For a deeper dive, check out our white paper on developing a digitalization strategy.