Stock Management
Background
I was reached out by a small restaurant from Barcelona with serious issues with their stock management.
This restaurant main dish is the Argentine "empanada". At that time, the current stock management strategy consisted in granting a fixed amout of precooked units of each product according to the ammount of sales of each one historically.
From time to time, they faced scenarios where they were unable to comply with this minimun stock strategy, but they also ended up wasting empanadas because of expiration.
Since the bussiness owners had no background on gastronomy, they were looking to state a data-driven strategy to manage their stock.
Problem Statement
By conducting an interview with the managers, we noticed they were not taking into account their expected sales to manage their stock.
The responsible of managing the stock has stated an optimal minimum and maximum amount of empanadas to be preserved in their refrigerators. Therefore, the strategy consists in just to order more empanadas every time the minimum stock is reached.
This approach is based on the reorder poing formula (ROP)
ROP = (Average daily output x lead time) + safety stock
It was indeed a good approach, but it had one big downsides: how to measure the average daily output. This KPI was not optimal in this case because of two distorsive events:
The wasted empanadas due to expiration. This waste was included in the calculation. But we can asume some of this waste could be avoided by improving the stock management. So we may want to exclude the waste caused by overstocking from the calculations.
The sells lost because of out of stock. The restaurant was losing sells every time they runned out of empanadas, and this was not considered in the calculations.
On top of this, the managers observed that sells were quite volatile and the historical average was not describing accurately the behaviour of the sells.
The stock management strategy is not aligned with the sells.
Data Exploration
This project conducted a brief analysis on historical sales and provided a simple sales forecast still to be tested and validated by the restaurant.
Since the only historical data we had was the restaurant daily closing, from which we'll only take date and number of empanadas sold. I collected the data from the cash register and shared some findings with the managers.
My first conclusion was that the sells were not that volatile, because they seemed to be following some trends based on the day of the week, or the day of the month.
This was not new for the restaurant managers. They were already aware that Fridays and Saturdays were "hot days" for them, and they also believe that they tend to receive more customers after the salary payment days.