Replenishment - New Thoughts on an Old Problem

Silhouette of business people on skyscrapers

A replenishment system's purpose is to place the right amount of inventory at the right place at the right time. It's a basic problem yet requires complex execution. Over the years, product explosion, location additions and contractions, omnichannel shopping, and fulfillment option expansion have made replenishment challenging. And while there has been technological assistance, additional automation, and process refinement, they've all been building on the same premise of adjusting service levels, order points, up-to points, and mix/max guardrails. It's time for a new approach. 

The word "right" is obscure. What is the "right" amount of the inventory at the "right" place at the "right" time? Does "right" mean to maximize the sales volume? Does it mean customers can walk into any location and buy the item they want? Or does "right" mean that you have all items in stock when the store opens? Or is it only to ensure the customer can get what he wants or something similar, even if it isn't in the store? There is no correct answer to these questions. 

But there are also the trade-offs of "right." What costs are acceptable to achieve "right?" How much inventory are you willing to carry? How much can the space fill? What incentives are vendors offering you to take that space? 

And then there are the realities of making "right" happen. How long does it take to deliver to each location? How reliable is the supplier, and what happens when demand shifts dramatically? 

Defining "right" is hard, and getting "right" right is even harder. 

In response to this dilemma, replenishment systems have concentrated on optimizing service levels and using rules and guardrails such as presentation minimums, lead times, and case pack size. But service levels have generally been the focal point. Yet this can be problematic. 

Let's set the stage. Suppose you had an item you always wanted available; you'd set the service level to 98%, which translates that you wish a 98% probability that you will always be in stock over the replenishment cycle. It's not saying that you will fulfill 98% of customer orders.  

Here is another way to look at the service level, as presented by Peter L. King, CSCP, author of Lean for the Process. At a 98% service level, expect that 50% of the time, not all cycle stock will be depleted. For 48% of cycles, the safety stock will suffice. But in 2% of replenishment cycles, expect a stockout.  

The service level definition is confusing, but it determines the inventory amount. Mathematically, there is a hyperbolic relationship between service level and inventory, as seen in the graph below.   

Inventory level compared to service level graph

What does this have to do with anything? As you can see, every 1% improvement in service level becomes exponentially more expensive, and the difference between 95% and 97% is significantly more than 85% and 87%. But retailers want high service levels to maintain customer loyalty. And so often, when service levels are set to meet these goals, high inventory and additional costs are the outcomes. 

But why wouldn't people then pull back the service level to 95% or 90%? Well, in some cases, they do. But when and what is that trigger? What is the "right" inventory level to maximize profitability across the enterprise, region, or store? And shouldn't that service level change over time to optimize the inventory and profitability?   

These are the questions we ponder. How do you connect financials to inventory? Or, conversely, how can you connect your replenishment decisions to financials? Rather than optimizing service levels and keeping your inventory levels in check, you focus on maximizing profit with the same amount of inventory investment. Strategically, you would want to designate a higher preference for some items. But imagine the flexibility you gain as your service levels, order points, and order-up-to points adjust to meet your financial goals.  

There are other benefits to this approach too. When retailers tighten their belts, they can lower inventory investment while maintaining profitability. If they want to invest in a location, they could increase their inventory levels to see if they can increase profitability.  

Much of this approach requires forecasting demand and predicting outcomes. After all, it requires data crunching, analysis, and AI and machine learning. But herein may be a concern for retailers. Wouldn't this approach be restricted if a good percentage of a retailer's assortment uses min/max settings for replenishment? 

Simply, yes. But let's back up some. Even before this approach, min/max replenishment settings were never ideal. Generally, they are used for slow-moving items where forecast predictability is questionable. But today, forecasting enhancements have increased the forecast accuracy of slow-moving items. "Dynamic aggregation" is one method we at antuit.ai use to scour the data and determine the right level to forecast.  

Additionally, our demand calculations include non-traditional leading indicators such as weather, pollen, or local events generating more responsive and accurate forecasts. It opens the door to items and categories that people once considered unforecastable - this isn't your dad's forecasting tool. 

Driving retail decisions through financial metrics is not a foreign concept, as markdown optimization solutions have applied this convention for years. Rather than using predefined discount levels at specific times (25% at 12 weeks, 50% at 16 weeks), markdown optimization solutions use sell-through and margin goals to determine when and how deep to discount. The same financial decision logic also has found its way into allocation decisions, where markdown costs and lost sales are evaluated as part of the inventory allocation decision. Even size profile optimization solutions consider financial benefits and pitfalls as customers push for more size inclusion.  

Most traditional technology systems automated an existing process, but as technology advanced and external market conditions changed, new systems brought a fresh approach to solving an old problem. Due to AI advancements, new data sources, and rapidly shifting market replenishment systems are ready to take that next leap.