On a recent trip, I met a local manufacturer who specializes in building and supplying kites. The kite season, which lasts a couple of weeks, drives the local industry and the success of the enterprise depends on good weather during those weeks. This manufacturer prepares for these crucial weeks by first stocking raw materials and then constructing large quantities of different sizes and color combinations along with accessories required for kite flying. When speaking with the team, they were able to discuss their general business plan and what was remarkable about this small business owner was his ability to be hyper agile when adapting to business challenges and changing trends and reflect them in the company’s products. In a microcosm, this individual is the archetype of what machine learning solutions can provide to manufacturers at scale including the ability to:
- Adapt quickly to customer demands.
- Navigate supply chain issues around sourcing and stocking.
- React to market drivers to price competitively.
The Evolving Nature of Manufacturing Business Goals
Pandemic behavior has transformed consumer buying patterns and changed the nature of business operations in many industries. Post-pandemic purchasing trends continue to be different from what they were pre-pandemic and economic uncertainties have compounded the ability of businesses to confidently make plans based on past behavior. To manage the complexity of the ecosystem, manufacturers must invest in ways to react quickly if things are not going according to plan. Unlike the kite maker they likely have hundreds of product portfolios and yearlong manufacturing cycles. They will also have complex distribution strategies along with complex people and operational infrastructure. Three unifying goals for businesses looking to build successful digital infrastructure can be summarized as follows:
- Improve profitability of business operations including focus on sustainable practices.
- Retain and improve the efficiency of the workforce.
- Grow customers and build a robust product portfolio to address future market demand.
Leveraging artificial intelligence and machine learning (AI/ML) strategies to augment existing business strategies can help businesses meet these goals. These technologies leverage the single largest untapped competitive advantage i.e. decision history stored as data to provide extra agility and competitiveness.
As with the kite maker, the ability to react to customer demands is the difference between long term success and failure.
Introducing AI/ML Decision Engines
AI is different from ML even though they are used interchangeably. While artificial intelligence is a broad discipline including applications like robotics, sensors, and vision systems to help manufacturing processes, machine learning is a key unifying ability because it leverages technology, software, data and a business process to aid and augment the ability of businesses to adapt.
There are various ML techniques, but the following are key components of a decision engine:
- Digital business platform that supports big data pipeline, orchestration, and analytics
- Machine learning platform supporting multiple models and maintenance.
- Composable architecture that supports integration to different business systems.
Leveraging a decision engine allows companies to take advantage of ML because it allows for the business users to answer the following questions: what [to do], when [to perform the action], how much [quantity or relevant unit of measure] and the why [the action is recommended].
The kite maker's decision engine is based on understanding the current market preferences, competitor pricing and stock as well locations where to get the optimum foot traffic. Each product has an elastic price based on customer interest and order size. While that works for stand alone locations AI/ML Decision engines can allow similar “personalized” experience with more scale and automation and without the requirement of as much critical and highly specialized business knowledge.
The decision engines help manufacturers to understand what to build and sell and how to upsell at a basic level.
The Practice of Demand Forecasting
The concept of demand forecasting can loosely be applied to various supply chain opportunities because it leverages past information to predict future demand. Demand in this case would be manufactured product or as labor or cash flow depending on the perspective needed. The advantage of ML-based forecasting is that it can leverage attributes like seasonality, product and customer information to provide a superior forecast. It can be set up to be reactive when needed to allow for supply chain disruptions.
An additional advantage is that demand forecasting can incorporate external factors to improve the predictions. The ability to automatically correlate several factors enables demand forecasting models to provide adaptability to new product introductions or expansion plans to new locations and markets. Additional advanced applications of a demand forecast would be to help suggest inventory movement and allocation strategies based on targeted business goals, such as maximizing profit and revenue.
Recommender Systems
Other popular applications of ML can be loosely classified as recommender systems. The concept of understanding customer patterns or product relationships to help improve the customer ordering experience or even to help improve product portfolio planning for future product improvements is a very common and easy application. The best example of this is popular e-commerce portals like Amazon but this is also very applicable to smaller manufacturers, distributors, and retailers. Recommender systems can also be applied in learnings and trainings of workforces to help improve productivity and provide career path guidance to your workforce.
In the case of the kite maker, their business strategy involved “bundling” kites and string as well as providing discounts on multiple kites/colors/sizes to leave the buyer satisfied at their individual order.
Aiding Today’s Businesses
The concept of data-driven decision making allows various solutions to coordinate together to provide business outcomes and augment manual decision-making processes. Leveraging a decision engine and these solutions can help any manufacturer achieve several percent improvements in accurately planning for the demand and supply parts of the business.
Ability to scale to allow kite maker business model to larger, more complex businesses by mimicking the agility and personalization and emphasis on customer experience is an easy quick win for other manufacturers to apply AI/ML to common supply chain challenges.