Lenskart - Stock Replenishment System

The case study is related to Stock Replenishment System for 500+ stores via a central warehouse for display stock.

There is a central warehouse that has 10,000 SKUs of eyeglasses (products), spread across a total of 10 lacs quantity worth Rs 40 Cr. There are 500+ stores each with 1000 eyeglasses. How would we replenish all these stores?

How the model works:

• Eyeglasses shipped from the central warehouse are used as display stock at the store

• When a customer comes to the store they try out these glasses and then select an SKU

• Basis of their selection order is taken

• Order flows into the warehouse management system from where it is fulfilled to the customer’s home

Questions to be answered: • How to decide what to send to which store

• When and how much replenishment should be done

• Should this be pushed by central warehouse, or ordered via stores

• For managing this what will be the mockup of system which will be designed

Deliverables: • Mockups of system to be used for managing the model

• Model of what parameters are important and to be considered (on basis of assumptions)

---

Assumptions-

1. Display stock cannot be sold to customers.

2. Products delivered are not damaged and there is no shrinkage in the whole process

3. Inventory left across stores is not redistributed

4. Historic sale data has been collected and a model to forecast the sale is already available.

SKUs

1. Empty frames with box & cleaning cloth

2. Stock lens with various prescription combinations

3. cleaning liquid, optical tools

There is one central warehouse with a stock of 10 lacs in total comprising 10,000 SKUs. There are 500+ stores with 1000 SKUs.

Lets divide the spectacle frame product into -

1. New models - 20% (assumption)

2. Old models - 80 % (assumption)

Sales Forecast

We may only get forecasts for the old models, based on historical demand. For new models, we might have to experiment by sending minimum quantities equally across store

1. Qn - New models quantity sent to each store (Product Launch)

pls note new product sales forecast doesn't necessarily have to be fixed min quantities. we can hypothesise

1. Qo1 - Old models quantity demand for store1 per day

Lead time - Time taken for SKUs to reach the store since the restocking request is placed. This differs for each of the 500+ stores, because of the distance from the warehouse is different for each store.

L1 - Lead time for store1 in days

Period of replenishment (Options)

1. Week (might be too frequent if no sales) (more cost)

2. 15 days (moderate cost)

3. 30 days (might be too late if there are a lot of sales) (Least cost)

4. As and when required (this way frequency and cost is optimised )

Recommendation

The following is a schema for data and product movement across systems and locations

Once a customer places an order, the sale is recorded in the database of the central inventory management system, for order fulfilment. It is then the onus of the store to fulfil the order from their inventory. I.e, customise the spectacle for the prescription if any and deliver it to the address mentioned by the customer in the order.

Since the inventory are not low ticket items and conducting refills periodically might be ineffective for 500+ stores.

I would suggest the Fixed-Order Quantity Model for restocking the inventory.

Stock replenishment order is initiated by the store once the stock balance reaches Reorder point(R)

R = L1*Qo1 (Lead time for store 1 x Demand for store 1) + Safety Stock

Safety stock = z*𝞼

( z - Number of standard deviations for a specified service probability

for 95% probability of not stocking out z= 1.64

𝞼-Standard deviation in Lead time )

Since the demand for a store is known, the stock quantity sent while replenishment should be

for store1, Qo1 + Qn (In case, New product launch)

Additionally I would recommend the following processes -

Picking process

While loading the truck from the warehouse, the SKUs are packed in a container with a QR code - that specifies the pick container ID. All of this data is recorded in the transaction table.

Receiving and stocking process

​Transaction Table Product ID Product Description User ID Quantity adjusted Pick Container ID Date Store ID

​Product Table Product Id Product Description Purchase Cost Retail Price Vendor ID

​Stock table / Inventory Table Store ID Quantity Product ID

Further Improvements -

1. SKUs can be categorized in different buckets. Each bucket will have a base trend/ pattern/ history upon which forecasting can be fine-tuned for individual SKUs

2. Imaging systems equipped with AI/ML algorithms can be explored for stock monitoring