How to Adapt Your Supply Chain Analysis When Forecasts are No Longer Accurate

Supply chain management is never stagnant. Effective supply chains continually leverage analytics to fine-tune the process:

• Efficiently disperse inventory from centralized production to customers, with narrow lead times;
• Reduce the risk of supply chain loss and obsolescence, especially with perishable goods; and
• Efficiently scale operations, i.e. how many warehouse employees, trucks, and drivers are needed to distribute inventory during recovery toward “normal” operations?

Even the most tightly run logistics programs now experience an enormous slack in the supply chain, due to an abrupt failure in forecasting.

Shutdown orders and consumer behavioral changes in response to COVID-19 shattered market norms overnight. Companies are feeling an inescapable correlation—the more Actuals deviate from Forecasts, the more [negative] impact.

What is certain, however, is that operations still reliant on models, methods, and assumptions established prior to COVID-19 can no longer effectively address inventory management, labor planning and retention, and how to maintain service level with customer needs, all of which relentlessly evolve at the pace of a volatile market.

Throw your AI/Machine Learning Models out the window
Consider any machine learning or AI models trained on historical data to be obsolete. Your Data Scientists are not able to predict local ordinance changes, shelter in place orders, closures and re-openings; and they cannot predict the impacts of protests and curfews on timing and volume of demand at even a macro level— much less at the hyper local precision required to usefully inform your Logistics. Further, your product volume and SKU Mix is rebalanced, as access to distribution channels arbitrarily flex with new outbreaks and regulatory response.

Get back to the basics
To overcome the challenges of failed forecasts, we recommend you return to the basics with a singular focus and rudimentary (and timely) metrics. While machine learning models have proven (for now) ineffective at predicting what is going to happen tomorrow, what happened yesterday has now become the best predictor of what will take place today—and the conditions experienced today are the best predictor of what will happen tomorrow, conceptually speaking. Monitor velocity and acceleration of market closures or reopening with recent data of various lookback periods in view of each other. Benchmark distribution sites across the company to identify where to spend management efforts.

Know your productivity metrics!
Understand at a local granularity what is “optimal” performance, and monitor what your assets (people and equipment, capacity) are achieving now versus the prior year (or an otherwise cyclically appropriate for comparison). If you know the most efficient expectation of a Picker in terms of widget pallets per hour, and the ideal relationship between drop size and number of stops on a route, you can bring a highly methodical approach to determining new driver, truck, and warehouse personnel levels.

Eliminate blind spots in inventory levels and flows through the distribution sites
In the absence of forecasting or downstream data from point of sale, we can use, for instance, a 10-day rolling daily sales rate to retailers as a gross approximation to patch into existing forward looking analyses. Refine this then for immediate term decisions by analyzing finalized orders just ahead of delivery day. Collaborate with your sales team to establish say a 48-hour lead time on fulfillment from time of order placement, buying time for intervention on Planning & Service Level priorities.

Be an advocate for your customer and maintain Service Levels
With so much uncertainty around the pandemic, it’s a complicated time for your customers. It’s likely that their business model and/or that of critical supporting supply chain partners is not running at full capacity. That said, with a strong data program around customer-centric metrics, there is an opportunity to outperform competitors during recovery. Companies that step up as a good and reliable partner for their accounts during a challenging era will not be soon forgotten.

Understand your customer’s evolving needs
If you cannot confidently derive the nuanced needs of your customers via existing data, then simply ask them.  Survey your customers and clients to see how your approach can evolve to meet their needs. Add this data point to your analysis for a more holistic customer assessment.

Pursue end-to-end visibility
Integration is key, as pieces of Logistics information viewed in their individual silos can’t tell you the story of the whole. Knitting together the data flows from the myriad of systems which support warehouse and delivery operations is the ultimate challenge. Success with this type of project requires a team with the right mix of skill sets, driven by a shared vision, and equipped with the requisite tools and technology. End-to-end visibility starts with a current data strategy. Big budget integration projects not supported by a strong data strategy will fail.

The analysis shouldn’t stop at your supply chain
Recovering from a recession requires a holistic view of your company’s data, not just the supply chain. Two other key focus areas that should be factored into your decision-making include the customer and your people. Without truly understanding your customer, your business can’t survive; and without understanding your workforce, your company won’t retain the very people who serve your customer.

Talk With a Data Analytics Expert