Over the past 18 months, companies in the APAC region, like many areas across the globe, have faced volatility and disruption across their supply chains. For example, rising commodity and energy prices are causing global transportation disruptions and delays. As a result, many companies are left trying to quantify the effects that supply chain disruptions will have on their business and react in ways that will minimize the financial impact.
One of the ways companies can minimize the impact of supply chain disruptions is by implementing comprehensive AI/ML planning solutions. One of the most obvious benefits of AI/ML technology is financial. Taking on digitization initiatives not only can increase sales and productivity, but can streamline the cost of goods sold.
Another significant benefit is uncovering how much data and information actually runs through your business. An AI/ML planning solution enables demand drivers to be layered: the first layer includes sales history, the second layer is internal promotions and internal prices, a third layer is external data like macroeconomics, weather, or seasonality, and a fourth data set is competitor data. When combined, this information transforms historical data into a much more robust set of data points that can ultimately improve demand forecasting capabilities.
“Implementing an AI/ML solution is a discovery journey… and that learning is circular,” says Simon Joiner, Senior Product Manager at o9 Solutions. “The more you learn, the more you’re able to improve, the more you will want to do to further the journey in order to grab more data and gain more learning.”
To make the most of an AI/ML solution, here are three points for companies in APAC, and across the world, to keep in mind:
What data is required to reap the benefits of AI/ML
Demand forecasting is driven by consumer activity and over the past two years, COVID sales patterns have dramatically impacted consumer activity. Therefore, sales history alone no longer provides what’s needed for accurate demand planning in an environment that currently is dealing with volatile demand and supply shocks. AI/ML platforms can connect to external sources including weather, geography, logistics, promotions, price, seasonality, etc., to better forecast demand.
There are two key types of data that need to be collected and they are master data, which relates to structures like addresses or standard naming conventions and fact data, which equates to the information that flows through master data. Master data tends to remain static while fact data will change frequently. Consider an address as master data and the orders that you placed for delivery to that address as fact data.
This data can be collected through numerous routes, some are freely available from institutional sources, some can be purchased through specialist providers and others can be obtained through web scraping. This final approach, using web scraping or intelligent extraction to capture data from competitor promotions to product reviews can provide cutting edge data intelligence and transformative insight into your forecasts.
What skills are required to successfully leverage AI/ML solutions
In order to fully leverage vast data and turn it into insights that inform the demand planning process, system features like data collection, harmonization, as well as employees with strong data analysis and evaluation skills are a must. To ensure successful data evaluation, companies should have a data scientist and a business analyst working together to gather insights. “If you can have resources that are both data science and business then that’s perfect,” says Simon. “But typically you need people who are algorithmic experts and business experts and they have to work together to get the best blend of data analysis and business benefit.”
Another necessary skill is evaluation because planners need to react to the data that the AI/ML system is providing them by incorporating data visibility, understanding variable importance, and having the ability to run scenarios based on the AI/ML output. Many traditional planning solutions don’t have the capability to load, calculate or display external driver data and having resources that understand data usage is crucial. Additionally R/Python skills are also beneficial because open source solutions are a growing trend. “Black-box” forecasting systems can only evolve in ways understood by the solution provider but open-source systems can be adjusted whenever a planning breakthrough is developed.
The challenges of transforming the planning cycle and how to facilitate change
The volume of data that is becoming available for businesses to use is extensive and continually increasing. The biggest change over planning deployments from the past is “the sheer processing power that is available,” says Geoffery Thomas, founder of XAct Solutions. “It’s the big data phenomena and the ability to process and most importantly, to manipulate large amounts of data in shorter time-frames.” The challenge of vast datasets is resolved by moving away from on-premise ERP systems to managing big data with cloud computing solutions.
Ideally, planners will want to shift their planning cycle from a monthly to a weekly/daily cycle but there are challenges in doing so. One significant obstacle is trusting the results of AI/ML solutions. “There will no doubt be resistance to change amongst planners, who will often still back themselves against best-fit algorithms in many categories,” Geoffrey says. “If the implementation of machine learning forecasting is well scoped, and its potential value is understood by chains at the outset, then this can be overcome.”
AI/ML will enable planners to interpret demand signals with much greater speed and granularity and this agility is, for many APAC businesses, just as important as accuracy improvement. Organizations can begin to transform their planning cycle by first understanding their current maturity phase. Once that’s determined, a company can create a plan to progress on its demand planning journey. Companies also need to understand how they will use data and systems, which includes market intelligence, knowledge models, and AI/ML-based analytics, to enable them to improve their planning maturity level. Knowledge models will need to be built and incorporated into calendars and industry makeup, but these structures will allow planners to finetune the structure of data and use it in business operations and in creating more accurate forecasts based on real-time data.
The most important factor is to have a key business sponsor who works with the business to come up with very specific, tangible use cases or business cases that the leadership team can understand and approve. “You have to link it back to the money and not just to the data—that’s where the success will come from,” says Rahul Teotia, Executive General Manager & Group Chief Supply Chain Officer at Pact Group Holdings Ltd. “So I think if you look at a project or a lifeline of an AI/ML project, 65% of the time needs to go up-front in getting people on board, once they’re on board implementation can be quick.”
