Implementing and beginning to utilize data-driven analytics to operate and manage your supply chain has gotten easier than when a few hardy pioneers first “test drove” early installations. Like always, inconsistent data is the biggest culprit in obtaining good analysis. We are going to search the World for you and find what else can cause issues.

At a high level, manual data input and the data interpretation process can be troublesome. As users learn and ramp up to speed, they will easily discover any issues with the validity of all algorithms.

Lets divide this into three areas and start with

 

(1) DATA QUALITY
When the ERP transaction system is the only feed to the analytics system, “Garbage In, Garbage Out” is easier to locate. But as analytics has improved, it now relies on multiple sources. We may need to train users of other systems and stress the importance of what they are doing. We should test validity wherever possible. Other workaround actions include embedding calculations in the system and using estimates to replace missing or inaccurate data.

(2) FILTER FOR DATA OUTLIERS
Let's first simply define Data “Outliers”: Small groups of data that are exceptional when compared with the rest of the data. A lot of analytics methods can be quite sensitive to outliers. They can add a lot of misleading confusion approaching chaos to the analysis.It should be a user responsibility, otherwise it is difficult to diagnose.

It could be anything causing the outliers: for example, products misclassified in posting to categories. This shows the importance of a data cleaning step in the process. Otherwise, you might have to “reverse engineer” to find the outliers and this could take time. Those personnel assigned to “software maintenance” need to be familiar with algorithms development as well as the data structure and content requirements.

(3) THE ANALYTICS USERS NOT BEING FAMILIAR WITH STATISTICS INTERPRETATION
Analytics use statistics, but even with the same dataset, you can get different results if you do not understand the calculation. Looks like a user training issue. Things are not as simple as they look. Some ideas to help users: Collecting data shows how the company works. Missing or inaccurate data can be estimated by matching to the historical database. Again, understand the calculations behind the reports. Benchmark results against others in the staff.

Finally users, you have a powerful tool if used correctly. Follow the process and data-driven analytics will amaze you with the results.