Moritz Krol
18 Sep 2021
Most online shoppers can relate to the frustration of reading “Available only in 12 weeks” when buying products online. Up to 58% of frustrated shoppers are lost customers due to product unavailability in in-store sales and 65% in online sales. This can be a reason to take a closer look at the root causes of out-of-stock situations and, therefore, increase sales. That’s what we did together with the leading online store for e-mobility charging stations in Germany.
The following 7 steps of algorithmic improvements and integrated collaboration on the numi platform helped to improve customer service levels, reduce canceled orders and, thus, increase sales significantly.
The starting point of our implementation was our quick supply chain check-up. We analyzed customer orders, inventory data, and purchase orders for the last 5 years. Current planning parameters and the customer's KPIs were benchmarked against our algorithms.
Process Mining Algorithms helped us to identify process bottlenecks and actual process lead times in the purchasing and order management process. Especially decreasing supplier performances due to demand peaks (thanks to the e-mobility funding by the government) resulted in many service level hits.
The customer used basic forecasting and inventory planning algorithms. Therefore we benchmarked current forecast accuracies and inventory turns against our Machine Learning and Operations Research algorithms and calculated an algorithm value-add. The simulation over 12 planning periods showed a significant increase in forecast accuracy and inventory turns.
After this quick, initial proof of value and pain point analysis, we connected to the ERP system and started to create improvement recommendations with numi Predict, Supply, Inventory & OnTime on a daily basis to create sustainable customer value for their urgent problems.
The integration of external data like e-mobility funding by the government or open customer orders and stock-out information is fed into our Machine Learning forecasting models and integrated into a weekly forecasting process. This helped to get more than 60% forecast accuracy on a single SKU level across all warehouse locations.
Further, the weekly forecast is enriched by our Demand Sensing engine to give the user proactive alerts. Reduced stock-outs, especially for New Product Introductions in a very fast and volatile market environment was the result.
Instead of relying on the assumption of normal distributions, we use a best-fit selection of suitable inventory optimization algorithms. This means that we used probabilistic forecasts where forecasts can be applied, fit actual distributions and use a Machine Learning classification logic for very expensive and slow-moving parts for the decision to stock or not to stock.
All inventory optimization results are provided as replenishment, rebalancing, expedite, or sell-off recommendations to the user so that informed decisions can be made and pushed back into the ERP system.
One major issue we identified together with the customer during our supply chain check-up was late inbound deliveries. Therefore, we have added a proactive approach to our process mining algorithms for detecting actual lead times and lead time bottlenecks. We predict an expected lead time on a single SKU and supplier level.
numi uses these regularly updated lead time forecasts for replenishment recommendations to avoid stock-outs. This way we reduce supply variability to keep safety stock levels and thus inventory levels small.
Another beneficial impact of knowing the expected lead time is that incoming customer orders can be matched against available inventory and up-to-date inbound notices. Order confirmations can be automated and more importantly updated with the latest information to avoid negative surprises to the end customer.
A further positive side effect is up-to-date “available at” dates, which are directly pushed back into their webshop.
numi also provides aggregated reports on several hierarchies and dimensions. Therefore, we defined all relevant KPIs with the customer and implemented these customizations based on their needs. Since numi predicts everything, all KPIs are calculated for the next 24 weeks to make proactive decisions on the revenue at risk, working capital investments, warehouse and logistics capacities, or inbound expediting decisions.
Besides all these algorithmic improvements the major success factor to increase sales of this e-commerce customer is integrated operations and collaboration. The whole supply chain team, from sales to inventory, to master-data, to purchasing experts is working on a single SKU level on daily supply chain issues and their impact on customer service level.
This horizontal integration of teams is enabled by connecting supply chain data for each single SKU, enrich with all relevant information like late inbound shipments, forecast accuracy, inventory values, and demand at risk. It's presented to the user on one single screen. numi takes also care of prioritized recommendations, addressed to the right user persona.