Portfolio

Replenish

What should we send each store this week, and by which size?

A size level demand planner for five stores. The engine proposes a number for every size in every store, so I only read the rows where it is wrong. One saved version becomes the week.

You can't stock every size, and you can't afford to miss the sale.

Sales are less predictable than they seem.

Eighteen sizes. Sales never land in a clean curve. Online a shopper picks 12.5. In the store the same shopper buys 12.

Many shoppers insist they are a size that is not theirs, and walk when the shelf will not bend.

The tail is rare, and it is not never.

<1.6% chance of selling sizes 14–16. If a style sells 30 times in a month, there is still about a 50% chance one of those three sizes sells.

ExpectedOne month, actual

Not too much, not too little.

The stockroom is finite.

Equal allocation

20models

Demand-weighted

37models

The customer walks.

40%of an intended sale is what the retailer loses when the item is not on the shelf.

What a shopper does when the item is missing

Source. Thomas W. Gruen and Daniel Corsten, “A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry”, Grocery Manufacturers Association, Food Marketing Institute and National Association of Chain Drug Stores, 2007, page 10: “the cost to retailers is 40 percent of intended sales.” Response shares are Figure 4. nacds.org/pdfs/membership/out_of_stock.pdf

I broke the app down into three parts.

Strategy Lab Each category gets a demand curve Tiers set the allocation rules The Grid Review size-level proposals Retier when the gap says so The Review Review the summary Approve week-over-week changes Send the week to the warehouse

Step 1: Each category is assigned a sales curve

Algorithm defines the curve - Judgement refines it.

The demand curve

Store, 365 days E-com, 365 days The demand curve

Store weight0 / 100

Smoothing0

Step 2. The curve is turned into tiers - which become the allocation rules

A planner assigns it. The engine only applies it.

See the tiers in action

Walk through the week in the planning grid.

The engine proposes a target on-hand for every size. The planner decides.

Send the week for review.

The planner goes through the granular details. The reviewer receives the information they need to approve orders.

approved by the reviewer, 09:12

    What else it does.

    Sales analytics

    Category and size performance behind the plan, on a longer window than the week.

    Sep Aug
    Men’s Sneakers26.4%
    Women’s Sneakers24.3%
    Men’s Dress20.9%

    Returning to the warehouse

    Sizes a store will not sell, called back to the warehouse.

    STORE 77 rows 214 units WAREHOUSE back on the shelf
    Rows to clear77
    Units to come back214
    Held on tier watch2

    Transfers between stores

    One store is long in a size another store is short in, so the units move sideways.

    LAKESIDE long, size 10.5 6 units MIDTOWN short, size 10.5
    Store to store moves12
    Units rebalanced63
    Stores involved5