Warehouse Management

The previous post discussed parameters that influence stocking policy in warehouse pick face. The stocking policy should answer questions such as:

• What items should be stocked in the pick face?

• How much items should be stocked in the pick face?

• Where should a specific item be stocked in the pick face?

• How do you replenish items to pick face?

To answer these questions we need to simplify our approach. One way to do that is to use profiling i.e. grouping items and locators using Pareto Principle. The typical ABC analysis is an example of Pareto Principle. The idea is that certain items or locators can be grouped together in order to simplify analysis.

Locator Profiling

Profiling locators in pick face implies grouping locators into categories based on pick convenience. Pick convenience is a measure of the distance a pick operator travels to perform  a pick and the pick complexity required to perform a pick. Based on these factors, locators can be grouped into categories. A sample grouping could look like:

If you prefer a more graphical look, here is how the locator can be categorized:

Item Profiling

In order to perform this grouping, we need to analyze item demand patterns expected from the pick face. If the pick face in your warehouse is designed to store only Eaches, you absolutely need to remove full case or pallet picks from this analysis. We are also not concerned with the pick quantity either as we are assuming that the amount of time to pick is not not that dependent on pick quantity. Once the pick frequency is available, its possible to rank the items in descending order of pick frequency. Ideally you will see a small percentage of items constitute a large proportion of pick task.

Update: The following graphs shows this type of "Long Tail" distribution of pick activity. The picks for each item on the X-Axis are plotted against absolute number of picks on Y-Axis.

In this example, 80% of picks are coming from 20% or less items.

If the demand pattern in your warehouse is somewhat uniform across items, you will need a large proportion of items to cover a smaller percent of pick tasks.

Fat Tail Pick Distribution

In this example it takes 40% of items to cover 60% of tasks.

If you have a "Fat Tail" pick distribution, having dedicated locations in the pick face may not be that efficient.

Slotting

Next step is to use this information to slot items based on locator information coming from locator profiles. So for example, if there are 5 "gold" locators, you need to identify the top 5 items that constitute the bulk of pick activity, then identify the remaining items to slot the "silver" locators and so on.

By profiling locators and items, we have significantly simplified the problem. Instead of slotting hundreds or thousands of items into a few locators, we now have to slot a few item categories into a very limited number of locator categories. An obvious solution would appear something like this:

If a given slot could not be assigned to an item due to capacity or other constraints, that slot can be assigned to another item with a lower pick frequency. Not all items may get slotted based on number of locators available in pick face.

Fixed or Floating locators

What if the demand for an item or a set of items is highly volatile? Does it makes sense to assign a "gold" locator to an item on a lean day? Possibly not. Sure you can reslot an item during lean times. However reslotting is costly and could get complex especially when there are not enough empty locators.  A possible solution may be floating locators. In such a situation, the locator in the pick face is dynamically determined based on demand for that item on a particular day or even within a pick wave. Sometimes when demand is very volatile and non-uniform across items, its also possible to have a floating locator for an item with a "pick to zero" policy i.e. determine replenishment needs across all orders within a wave, replenish the total demand to pick face and draw down the locator to empty when the wave is completely picked.

Replenishment Policy

Since items that are stocked in pick face are usually fast movers, its not a good idea to have a stock out in the pick face. Therefore the locator should carry a minimum level of stock during the time it takes to replenish. Thus if it takes 1 day to replenish pick face, the average demand for a day should be the minimum quantity of stock at the pick face. The maximum quantity is usually determined by the locator capacity. To fully utilize pick face capacity, you may want to maximize the cubic volume of the pick face.

To reduce material handling during replenishments, its a good idea to specify replenishment lot size as the UOM that is stored in reserve area e.g. if pick face is replenished from reserve area storing master cases, specify replenishment lot size as the quantity of Eaches in a master case. This would ensure replenishments are done in integer quantity of master cases.

Clearly there are number of other factors that go into making an optimum slotting decision. Factors such as order correlation (slot items usually ordered together in the same aisle), crushability (slot crushable items at the end of pick travel path), item likeness attributes (slot like items apart from each other), etc. are important. However this simple approach should be a good starting point.

In the next post, I will cover how Oracle WMS can help in formulating and implementing a stocking policy for your pick face.