How machine learning is revolutionizing stock management

For those who work in stock management, the dilemma of determining inventory levels is all too familiar.

Order too much and you risk bleeding your profits through holding and acquisition costs. Order too little and you jeopardize your sales and, worse, undermine the trust your customers have invested in you.

Having to regularly make decisions about inventory is a headache every manager would rather do without. But while it’s so far proved unavoidable, at least we’ve made it manageable.

Traditionally, we’ve used stock management systems to maximize profit while minimizing loss.

The classic stock management system is the Newsvendor model.

The newsvendor model takes the predicament of a newspaper vendor, whose stock will have become worthless by the end of the day. Putting the mathematics aside, it aims to calculate the optimal inventory levels for a perishable product in a volatile market.

This model applies beyond the broadsheets to any industry where stock values fluctuate in the blink of an eye. The fashion industry is a good example, as demand might increase or drop off at the drop of a hat, several times within a season.

 
lauren-roberts-470110-unsplash.jpg
 

The Newsvendor model isn’t the only stock management system

Another stock management solution is to strive for the Target Stock Level, or TSL. The TSL is the maximum level of stock you can order where you have a fixed, regular supply order. But as surplus stock results in squandered profits, its limitations are clear.

Such limitations are no wonder when you consider which factors these stock management systems use to determine how much inventory to order. The answer, in short, is historical data.

This is all well and good, but it neglects those all-important exogenous factors like market trends, market competition, and (important for fashion, but notoriously tricky during a British summer), the weather.

Traditional inventory management solutions use formulae to arrive at static numbers. But is this really the way forward?

Relying on a formula to respond to demand is not just inflexible, it's generalizing.  

These traditional, formula-based inventory management solutions are company-centric, not taking into account the customer’s perspective as expressed through such things as web and social presence and rival markets.

 
roman-mager-59976-unsplash.jpg
 

Stock management formulae calculate generalized probabilities; machine-learning software generates precise predictions. Tweet This

For each inventory item, the machine learning software calculates optimal stock levels for maximum profits.

The software factors in the potential costs of stock outs, unsold inventory, and clearance discounts. And as well as analysing endogenous factors (like past sales and executive decision-making), it considers exogenous factors in making its inventory replenishment forecast.

The more data the machine learning software absorbs, the more accurate its forecasts become.

 
nasa-43569-unsplash.jpg
 

To put our money where our mouth is, Evo’s machine learning replenishing software doubled an Italian fashion company’s efficiency from 43 to 86 percent in three months.

But this is just the tip of the iceberg. And as the difference between actual and expected demand is where most profits are earned or lost, the question you now have to ask is whether going it alone is a risk you can afford.

 

Our next newsletter will explore more about how Evo is transforming the fashion industry. Subscribe here!


About the author

alexander_meddings_evo_960.png

Alexander Meddings is Evo's content expert on artificial intelligence, machine learning, and related topics.

He is an experienced journalist who covers branding, social media, marketing, and technology, with degrees from the University of Exeter and University of Oxford.