|
Gauri Vaidya |
With the emergence of the 4th Industrial Revolution, businesses around the world are striving hard to sustain themselves in highly competitive markets. The COVID-19 pandemic has made it even more challenging for industries to survive. Profit margins have traditionally been meager in the manufacturing industry and now with disruptions in the global supply chains, they are bearing additional losses on account of supply chain issues like material delays, supply-demand mismatch, and increased freight costs. In such scenarios, every penny saved becomes extremely important in sustaining the business, which is why operation efficiency is the main focus area. This is especially true for small-scale manufacturing companies which may have a very low tolerance with respect to absorbing losses. However, many of these problems, though they cannot be completely eliminated, can have their negative impact reduced substantially with the help of data analytics. Before going into data analytics, let’s first briefly understand the value chain of a manufacturing business.
A manufacturing setup broadly has the following units: Sales, Job Scheduling/Planner, Procurement, Raw Material (RM) Store/Warehouse, Quality, Production, and Finished Goods (FG) Store/Warehouse. The Figure 3 explains the process flow in a manufacturing business.
In the entire process, the occurrence of the following events will result in a financial loss to the business:
- Production idle time due to delay in RM receipt or less factory loading
- Over-stocking of inventory
- Under-stocking of inventory
- Delays in job scheduling
If the above-listed issues are minimized completely, the business will be operating at maximum efficiency. This has tradition- ally been addressed by effective Sales & Operations Planning (S & OP) in companies. According to Meredith Hart in her article A Comprehensive Overview of Sales and Operations Planning (S & OP), "Sales and Operations Planning (S & OP) is a business management process where leadership and executive teams meet to ensure each business function is aligned to balance supply and demand Hart (2021)." The main S & OP process steps are as follows:
- Forecasting
- Demand planning
- Supply planning
Figure 3: Process flow in a manufacturing business
Forecasting is the first step in the S & OP process and the most crucial. The better the forecast, the better the planning. Traditionally, forecasting has been mostly done via qualitative techniques like Market Research, Sales teams’ judgments, historic sales, etc. However, in the post-COVID world, it is required to make manufacturing units more flexible to meet the dynamic customer demands. Therefore, the S & OP process needs to also be more dynamic. This is what is going to help them to create a competitive edge. This is where data analytics comes into the picture. This is particularly useful to small businesses who generally do not have a well-established separate Sales team responsible for generating demand forecasts and it is not even a part of the business Key Performance Indicators (KPIs) to measure forecast accuracies. Moreover, if it is a push system flow, a poor demand forecast may result in high inventory cost holdings to the companies in a scenario where the demand from their customers in the Business-to-Business (B2B) scenarios is very dynamic. If the company relies completely on order backlog (i.e. start the manufacturing process only after the order is received), the customer demand will not be met in time due to internal process lead times. Therefore, demand forecasting with fair accuracy is a must in order to make a business sustainable in the market.
This can be very easily achieved through Time Series Forecasting Methods. A time series is a collection of data points over a period of time. For example, in a manufacturing unit, a time series can be terms of daily/weekly orders (also called deliveries or sales). A time series is characterized by certain parameters namely, level, trend, and seasonality. Level describes the average value of the series, the trend is the change in the series from one period to the next, and seasonality describes a short-term cyclical behavior of the series which can be observed several times within the given series. These time series components can be used in conducting a predictive analysis, to get demand forecasts. The most commonly used univariate time series forecasting methods are regression-based and smoothing forecasting models. These include using moving averages, naive and seasonal naive forecasts, exponential smoothing, and Auto-Regressive Integrated Moving Average (ARIMA), to name a few. All these techniques can be easily implemented using R or Python inbuilt libraries. The details of the codes and libraries used are not discussed in this article, however, there is a lot of material available on the internet that can be easily modified and reused with little knowledge of Python / R programming. Data analytics and visualization software like Tableau, also have an option of visualizing the future trends in time series plots.
To evaluate which forecasting model works the best for a business, various models’ outputs can be compared using metrics like Mean Absolute Percentage Error (MAPE). It is important to note here that a demand forecasting model which works the best for a certain part, may not be suitable for other parts, as their demand patterns might be different. Therefore, validation through accuracy measurement metrics like MAPE is crucial. More detailed information on time series analysis and forecasting methods can be found in the book “Data Mining for Business Analytics_ Concepts, Techniques, and Applications in R”, by Galit Shmueli, Peter C. Bruce, Inbal Yahav, Nitin R. Patel, Kenneth C. Lichtendahl Jr. Shmueli et al. (2017) and "How to Choose the Right Forecasting Technique," an article by John C. Chambers, Satinder K. Mullick, and Donald D. Smith Chambers et al. (2023). Shanthababu Pandian’s Time Series Analysis and Forecasting: Data-Driven Insights article on the website Analyticsvidhya Pandian (2023), can also be referred to for Python codes for developing forecasting models on real-time data.
Inventory Modeling and Supply Planning
Once the demand forecasting model is finalized for all the parts/products sold by the company, the forecasted demands can now be used as input for supply planning. Supply planning is controlled through the purchase requisitions (PRs) placed for raw material (RM) ordering. PRs are generated when the RM inventory goes below the reorder point. Inventory is controlled through Min, Max, and Reorder points. Therefore, to ensure smooth material supply for production, as well as to ensure that there is no overstocking of material, setting accurate inventory control points is crucial. For more information on inventory planning, Peter L. King’s article on Understanding safety stock and mastering its equations can be referred to King (2011).
The control parameters for inventory management can be derived through the following inputs:
- Procurement Lead Time (L months)
- Demand forecasted for L months
- Standard deviation of the demand forecasted for L months
- Mean demand for L months
- A factor z associated with service level (the expected probability of not hitting a stock-out during the next replenishment cycle)
The standard formula for calculating inventory is:
(Mean Demand) ∗ (Lead time L) +z∗(std dev of demand) ∗ √LeadTimeL
The first term in the above equation represents the minimum inventory required for running operations and the second term is the safety stock. Together, they will provide the target inventory level or the reorder point, i.e. if the stock goes below this point, an RM order needs to be placed. Now, the forecasted demand is obtained at the Finished Goods(FG) level, however the same can be used for getting RM requirements as well. The process flow for ordering RM can be seen in Figure 4.
This target inventory level is based on the customer demand of the next L months’ forecast. It will be updated every month based on changes in customer demand patterns. Thus, with this methodology, customer deliveries can be on time and even the occasional spikes in demand can be catered through the safety stock. Moreover, as demand and supply are matched, the inventory holding costs will also be optimized.
Thus, small manufacturing companies can benefit largely from the application of some of the easy predictive modeling techniques in Sales and Operations planning.
Figure 4: Process flow for ordering raw material
References
Chambers, J.C., Mullick, S.K., Smith, D.D., 2023. How to choose the right forecasting technique. Harvard Business Review URL: https://hbr.org/1971/07/how-to-choose-the-right-forecasting-technique.
Hart, M., 2021. A Comprehensive Overview of Sales and Operations Planning (SOP). Technical Report. HubSpot.
King, P.L., 2011. Crack the code. understanding safety stock and mastering its equations. The code - MIT - Massachusetts Institute of Technology URL: https://web.mit.edu/2.810/www/files/readings/King_SafetyStock.pdf.
Pandian, S., 2023. Time series analysis and forecasting: Data-driven insights (updated 2023). Analytics Vidhya URL: https://www.analyticsvidhya.com/blog/2021/10/a-comprehensive-guide-to-time-series-analysis/.
Shmueli, G., Bruce, P.C., Yahav, I., Patel, N.R., Lichtendahl, K.C., 2017. Data Mining for Business Analytics: Concepts, Techniques, and Applications in R. 1st ed., Wiley Publishing.