One of the most complicated and time-consuming components of predictive analytics is managing secondary Asset Operational and Maintenance data to get it in a format that is amenable to predictive analytics. Predikto deals with secondary data exclusively. By secondary, we mean that the data comes from our clients’ EAM / CMMS / Operational Indicators data storage and we have no control over how the data was collected and organized historically. ERP, EAM, CMMS, and other operational databases are typically not able to export raw data in a format required for the sorts of analytics that we do. Database designers don’t think about such complex analytics when designing a storage protocol; why would they? Their job is about data entry and collection, not about subsequent analyses that are oftentimes very complicated.
There is a considerable amount of custom computer programming that we have to develop in order to make our client’s data usable, which subsequently has to be done for each different asset (unless we’re dealing with identical machines that capture identical sensor data). In the past, we’ve experienced variations in how the same data is sent to us. Clients to try massage the data to make it easier to interpret (e.g., building in headers, or re-naming fields), but this complicates things for Predikto. It is critical that we get the exact same data every time in an absolutely raw and unadulterated way to ensure the complex data transformation and statistical models work every time as designed.