- March 22, 2020
- Posted by: marciadwilliams
- Categories: Digital Transformation, Logistics, Supply Chain, Warehousing
You got the data, now what? Data Analysis, Predictive Analytics, and AI in Supply Chain
What are you doing to have a significant impact on the financials of your organization? What are the actions that high performers are taking? There is an area of computer science that is disrupting industries. The results are impressive.
According to a recent McKinsey report, companies that have adopted this are enjoying digital cash flows of up to 20% greater than those who have not. In financial services, the gap is 30% while in high tech, the difference broadens up to 80%. These percentages are indicative of its benefits to operations and profitability. Do you know what this is?
The embracement of AI leads to significant results. Indeed, the same publication indicates that the most frequent success cases are in marketing and sales and in supply chains. AI adoption in supply chains implies the use of real-time data for demand sensing, forecasting, and planning; as well as scheduling and optimization.
AI in Supply Chain
We hear about AI every day. What does it entail? What’s the difference with the traditional data analysis?
Consider demand planning in supply chain, as an example. This is a critical function. With regard to its relevance, APICS states that “proper demand management facilitates the planning and use of resources for positive and profitable results and may involve marketing programs designed to increase or reduce demand in a relatively short time”.
From this, there are two key takeaways:
1. By having this right, you optimize inventory and thus reduce cost while delighting customers with on-time deliveries every time.
2. Supply chain is interconnected with the other functions within the organization. Examples are marketing and sales and finance.
There are different maturity levels when analyzing data; from a limited use of historic information throughout statistical models to machine learning.
The first phases in data analysis consist of understanding historic data. In this context, marketing and sales teams look at historic sales, annual volumes, SKU’s, customers, and make adjustments based on current market information or on projections. These projections consider a number of assumptions. Along these lines, the finance team analyzes historic data to assess the performance or profitability by product or by product group. When the team identifies gaps, a strategy or roadmap is developed to get back on track.
In more advanced data analysis phases, teams make use of statistics and statistical models to understand relationships among the different variables and forecast demand. A well-known model is regression analysis. The greater the model explains variation through the R-squared and the fewer the variables to do so, the better.
Machine learning goes even further than statistical models. It builds algorithms based on a huge amount of variables that includes historical data but also social media, the weather, customer reviews, among many other sources. Then it compares these algorithms against reality and makes adjustments. Each time, the algorithms get closer and closer to reality. This is the reason behind its name. With the multiple iterations, the machine “learns” a little bit more each time by increasing its accuracy.
So where do you start?
Reporting is a good starting point. You probably have different sources of information in your company such as ERP, CRM, Logistics, and Warehousing. Each of the systems has capabilities and talks to each other. If they are not interconnected, your vision would be limited. Precisely, the system interconnection makes process flows visible.
Transport Management – TMS
A thorough reporting considers the multiple sources, interconnections, and provides reporting for each of the functions. For instance, in transport management there are the following:
· Freight management
· Carrier Rating
· Route and Mode
· Carrier Optimization
· Customs and trade
Warehousing – WMS
If you take a look at warehousing, you find the following:
· Inventory tracking
· Shipping and receiving
· Labor management
With respect to the ERP solution, you can find – among others – the following:
· Order Management
· Inventory Management
Concerning CRM, you note the following, among other data:
· Conversion rates
As you can see, the systems in place do not operate in silos. For instance, inbound and outbound logistics – cross-docking, palletization, scheduling, among others – requires that the transportation (TMS) and warehouse (WMS) systems work together.
Leading-edge technology focuses on the combination of these different sources of data to provide cross-functional insights, going beyond each individual function. At the same time, it meets the specific needs and requirements of each individual function.
This is not merely gathering the data from the different functions; other dimensions are added. It is a highly enriching process where the sum of their individual parts is something greater. This is called metadata.
So what is metadata? It comes from a combination of sources. It is new information, a new dimension that is generated. It is where automation becomes real. You are probably thinking that this is great but what is it in practice?
A real-life example is a system with strong capabilities to generate freight invoices through automation. Invoices that show what is paid by each concept. This automation can be extended to warehousing.
Learning from the Past with Historical Data Analysis
These reporting capabilities provide new insights based on historical data. It is like traveling into the past to see what happened. The reports refer to a certain period of time. Examples are last month, last quarter, last year, yesterday, you name it. What is important is the analysis that you can derive from your report.
Let’s say that one of your key metrics is on-time deliveries. For the last couple of months, you have been implementing a new process to improve such a metric. How do you know if the new process has had the desired outcome?
Through reporting, you define the baseline level for on-time deliveries or your starting point. Then you measure again after the process is implemented. Did it work? You can see the value of reporting standards for learning and for continuous improvement by analyzing the past.
Anticipating the Future with Predictive Analytics
With strong reporting capabilities, you can travel to the future. You have a preview of what’s coming. This means that you can anticipate and predict through the use of data science and analytics. This opens the doors for opportunities by taking action.
Remember the example of improving on-time deliveries? With predictive analytics, you can determine the time that the goods in a customer’s order are going to be delivered. Predictive analytics drive processes. What is fundamental is the action to take out of it. Execution is what really matters.
For example, if you see a traffic jam through the data source, the pickup can be rescheduled accordingly. It is all about taking action. Another data science application is to optimize inventory that comprises having the right product in the warehouse; defining better pick profiles; and building a more efficient way of picking up the goods.
Going further and beyond. Best practices. Trends.
A recent article by Gartner about the trends in Supply Chain includes AI – comprehensive of machine learning – advanced analytics, and automation. Some companies have started to explore these technologies. The rollout of these technologies will not magically fix your supply chain issues. Sorry for saying this. To make it happen you need a solid process foundation. It implies a business transformation with many stakeholders involved. You may have great reports built, but again, the execution makes the difference. A transformation is challenging, but the rewards are worth it.