Augmented Analytics for Manufacturing: Analyzing Factory Costs
How can IT help a manufacturing company more accurately monitor, control, and predict cost associated with factory operations spread around the world? The short answer is augmented analytics and machine learning technology.
Traditionally, planning and analyzing factory costs across globally-dispersed operations has been a dark art of sorts – one that involved spreadsheets, deep knowledge of organizational silos, and tremendous amounts of legwork to gather and process all the required data. Which factory spent the most money last year? Where are the bulk of my labor costs? Simple cost planning questions such as these were difficult to answer.
A consolidated view of data
With augmented analytics, IT sets the groundwork for easier planning and greater visibility – starting with a consolidated view of data across silos and timely information augmented by machine learning – so users and managers can make more accurate, timely business planning predictions.
For example, imagine a business analyst for a steel manufacturer who needs to estimate spend over the next six months for 17 different factories across Europe.
Instead of starting the hunt for data, the analyst has access to augmented analytics – which means starting with instant visibility.
On the left, you can see all factories’ locations on a map of Europe, while the size of each circle illustrates the overall operating costs associated with each factory.
Then, at the top right, cost details for the last five years are displayed – for individual factories or in aggregate – by the various categories of spend including labor, materials, overhead, or total costs. This particular view shows that labor is the top cost contributor for the factory in Stuttgart, which is 20% above the average. A link to view more allows the analyst to drill down into the root causes for this cost overrun.
But where and how do predictions come into play? In the view below, historical costs are represented by the blue and green bars for two different factories: Stuttgart and Dusseldorf. The line at the top represents the predictive forecast generated by a machine learning algorithm.
The comparison of historical spend to the prediction helps an analyst understand how closely the real costs come to the predictive algorithm. In this case, the historical costs came closer to the prediction for the factory in Stuttgart (blue, on the left) than the factory in Dusseldorf (green, on the right). The Dusseldorf model, in other words, seems to be less accurate and needs to be investigated.
Forecasting and simulation in one place
In the example below, a business analyst can view the variance analysis for plan versus forecast for all factories in a single view (lower left). On the right, the analyst can see the value or cost drivers in a tree structure.
Let’s say the Dusseldorf factory has a particularly high overhead cost. And, there is a bigger difference (forecast to planned budget) as indicated by the red variance bar on the right of the deviation chart. The analysts should now be able to run a quick simulation that increases the planned overhead cost by 6%. The adjustment will be instantly reflected in the total planned budget for the organization at the top left side of the screen and on the variance chart – helping to speed up the planning process dramatically and with greater accuracy.