By now, it is a shared understanding that we are on the cusp of an industrial revolution led by AI, data and digitization. Industry 4.0 is a revolution that can pay off handsomely: A 2016 study by McKinsey revealed that the data-driven supply chain could gain up to 6 percent in additional revenue.
Yet many enterprises have not necessarily gotten to a level where they have all the tools and capabilities in place to access all needed data. So, the question is this: should companies first work to perfect their data-gathering capabilities, then get started on the advanced analytics?
Not really. In this post, we’ll cover some ways to get started with analytics and overcome seeing data perfection as a roadblock.
Let use cases help you prioritize which data to focus on
One approach is to consider the ways in which advanced analytics could add value to your supply chain. From there, develop a 1-2 year roadmap on how your enterprise would go about implementing them.
For example, your enterprise may want to focus on use cases related to understanding historical demand to plan for future inventory and capacity requirements. These require transactional data. Fortunately, most of this data comes from the ERP systems - which your enterprise is most likely using to some extent. Data from these systems, though not perfect, is generally of acceptable quality.
Let’s take this example a step further. Understanding transactional data may set you up for use cases focused on optimizing supply planning, transportation and demand forecasting. Worried that this data is only available only for a subset of regions, product groups or BUs? Just start there and add subsets as you go. The solution quality, in such a case, will not be optimal from day one, but it will be an important first step towards getting there.
Use artificial intelligence to fill in the gaps
There have been several advances in AI, especially related to machine learning, that can help fill crucial data gaps with representative data.
Let’s consider product demand, for example. It’s important to have continuous data points with no missing data to capture trends, seasonality and causality. The larger and more frequent data gaps there are, the larger the potential reduction of algorithmic accuracy.
One way to tackle this is by using the exponential weighted moving average based on neighborhood values. The nearest values to the data gap have the highest weights and vice versa. Weights can also be fine-tuned over successive future periods through feedback loops driven by comparison predicted to actual outcomes.
This is just one example of how machine learning can help your organization move forward with analytics when facing imperfect (or nonexistent) data. Even marginally improved predictions are better than none at all.
Make assumptions where necessary
This is a simple, but often an ignored and potentially effective workaround. Remember that your organization is staffed with experts in their fields. Reach out for their feedback and dig deep into what they say. You’ll find this may be enough to make reasonable assumptions around missing capacities, expected downtimes, missed demand and more. This alone could help your organization move forward without waiting for the sophisticated data feeds.
Formulate a data roadmap
Of course, your enterprise’s move toward improved data analytics should include steps toward implementing more advanced tools. The use case roadmap we already discussed can drive the deciding of which technology and systems should be put in place for collecting data on an ongoing basis.
Let’s consider a consumer products manufacturer that wants to capture information on the location and conditions of its transportation equipment. If the company was functionally siloed, it could use IOT to capture efficiency and identify improvements to reduce transportation costs only. However, if the company had stated “supply network optimization” as part of the monthly S&OP process and had implemented the tools to formulate optimized supply plans, it could benefit from the same technology to further improve the overall end-to-end supply chain performance.
Making the most of your data
You don’t need to wait for the perfect data to get started on implementing advanced analytics at your enterprise. This may sound like a paradox, but you are already sitting on data that can be used effectively. Instead, the challenge lies in developing the right analytics strategy that fits your needs and circumstances.
Given the promise that properly implemented advanced analytics hold, it’s a much better trade-off to get started with the available data and build-on as you go, as opposed to waiting for perfect data to become available and then start!