How operationalizing XAI enables stock management optimization?


IA de confiance

Tous les articles


Overstocks, out-of-stocks, dead stock … Retailers’ business depends on their ability to optimize their warehouse management. A lot of progress has been made in the last decade thanks to new management paradigms and control processes.

Optimize Stock Using Demand Forecasts

But stocking remains expensive, requiring infrastructure and working force. On the other hand, the market keeps asking for more reactivity, more products and more availability. Retailers must provide a first-rate level of service while reducing operational cost. Every mistake, every margin of error impacts the performance of a warehouse and improvements of the current workflows become thiner and thiner.

The recent surge in machine learning applications brings new ways to automate processes into the industry. Retailers have started to gather data from their warehouses and now want demand forecasting that is both accurate and able to capture the complexity of current market requirements while remaining easily useable by their stock managers. This is a challenge for most companies because of the multiple seasonalities and high variability in the demand at the daily, weekly, monthly and yearly scale.

Predict the Demand at the Lowest Level of Granularity

Some of the state of the art algorithms are powerful enough to model the demand despite the complexity of this task. Some are even able to predict the demand product by product, or warehouse by warehouse,. With predictions at such low level, experts are able to continuously adjust the inflow of products into each stocking points.

To model this, such algorithms need the historical records about the level of stock, paired with the relevant contextual data (meteo, sales, release of new products …). Most warehouses have access to the former and vendors can provide the latter through APIs.

Sometimes experts make mistake, sometimes Artificial Intelligence does. To make experts trust the forecast application, it’s mandatory to include explanations for every decision made by the AI. Experts have to be able to understand the reason behind a prediction and correct the AI so it keeps improving itself over time.

Second key point is the capacity to deal with the fast changing customers needs. To keep producing accurate predictions, the AI within the demand forecasting application has to update itself very frequently.

A forecast application with both the ability to update itself overtime and to take expert feedback into account will empower the stock managers and greatly improve their ability to optimize the stock-related costs.

Significant Reduction of Dead Stock

Craft ai has been designed to enable business teams to quickly get the value out of their data streams. Our AI solution was built with explainability in mind and natively supports all levels of granularity and continuous learning. It can model the behavior of a product, a product line or even a warehouse if needed.

A major french retailer challenged it, asking a quarterly forecasts for a large variety of products. We outreached their expectations and reduced by 50% the forecast error while bringing full interpretability to their experts.

The retailer was also having a hard time deploying their less performing project, with limited experience on operationalizing AI on time series. Leveraging craft ai API, our 10 000 of models are ready to be deployed in their business applications in order to optimize their supply chain ressources.

À télécharger

Vous pourriez également apprécier

IA de confiance

Garder l'humain dans la boucle avec l'XAI

Quelle place l’explicabilité (XAI) occupe-t-elle aujourd’hui en Machine Learning et en Data Science ? Le challenge de ces dix dernières années en data science, a plutôt été de trouver la bonne “recette algorithmique” pour créer des modèles de ML toujours plus puissants, plus complexes et donc de moins en moins compréhensibles.

Lire l'article


L'industrialisation de l'IA & le concept de MLOps

Le MLOps apparaît comme une nécessité pour pallier les difficultés lors du passage à l’échelle de l’IA au sein des entreprises : la reproductibilité, le versionning, l'intégration continue... C’était l’objet de l’une des conférence sur l'industrialisation de l'intelligence artificielle dans le cadre de l'Enjeu Day Industrie & Services 2022. Vous n’aviez pas pu y assister ? Retrouvez le replay.

Lire l'article

IA de confiance

A guide of the most promising XAI libraries

Using Machine Learning to solve a problem is good, understanding how it does is better. Indeed, many AI-based systems are characterized by their obscure nature. When seeking an explanation, looking to understand how a given result was produced, exposing the system’s architecture or its parameters alone is rarely enough. Explaining that a CNN recognized a dog by detailing the Neural Network weights is, to say the least, obscure. Even for models deemed as glass-box such as decision trees, a proper explanation is never obvious.

Lire l'article