Building intelligent
systems for retail

Trusted by leading Retailers & Brands

Current performance

0 %
Forecast accuracy
0 %
Overprediction
0 %
Underprediction

WAIR performance

0 %
Forecast accuracy
0 %
Overprediction
0 %
Underprediction

You can rebuild your whole business with accurate forecasting

WAIR takes retail inventory management to a higher level with deep Learning demand forecasting. Solving the limitations of standard forecasting methods, we deliver on the promise of the newest technology.  We bring together retailers, brands, and ecosystem partners in their need of transforming towards an AI-driven business.

It all starts with exceptionally precise SKU-shopfloor or channel demand forecasts, ranging from a single day to a 6-week span. This helps multiple use cases such as replenishment, channel reservation, and stock redistribution.  

WAIR’s demand forecasts are translated to stock-level recommendations for initial distribution, replenishment and redistribution. We move your processes from static into highly dynamic. Based on the reality of your customer demand. 

Forecast per aggegration level

Brand Shopfloor Store Channel Category Item SKU

Forecast per moment in time

5 Weeks Month 3 Weeks 2 Weeks Week Day

Exceptional
forecast accuracy

Our forecasts accurately predict up to 96% of all SKU-store sales (1 week). But in inventory management, it’s the ROI that counts. We push to find the optimal balance between over- and underprediction, acknowledging that no prediction is ever perfect.

Ultra-
responsive

With our forecasts pushed directly to your ERP, there is no interference in your current processes. Experience higher responsiveness in replenishment, redistribution, and channel reservation. 

Easy setup &
fair pricing

Our forecasting goes hand-in-hand with smooth ERP integration. With minimal implementation costs and a near plug-and-play setup, there’s little need for ICT involvement. Our usage-based pricing makes our technology accessible for businesses of all sizes, and best of all, there’s no vendor lock-in.

True domain
expertise 

Without thorough domain knowledge of the data and its business context, reaching breakthroughs is hard. In the WAIR team, we have more than 100 years of retail experience,  deeply understanding the retail industry. 

Our forecast
GPT-2.5 model

Deep learning forecast GPT-2.5
Training unified
model
Pooled data

7.5m parameters WAIR GPT-2.5

Pooled data

Unified model

100+ Features

Computational power

Domain expertise

Initial distribution
Replenishment
Re-distribution
Forecast per moment time per aggregation level
Deep learning forecast GPT-2.5
Unified model
Deep learning forecast GPT-2.5
Unified model
External data features
Internal data features

Breaking Down Silos with Pooled Data

The times of isolated models, each trained on one client’s dataset is behind us. Large and diverse datasets increase accuracy and minimize bias. The future is in pooled industry data. As the world grows increasingly interconnected, so too should our approach to data analysis and processing. 

This idea is as simple as it sounds: we process all our client’s data as one entity to train a unified model. No single business has enough data to capture a complete view of local, regional, and national trends to forecast demand accurately. Therefore, the Industry needs to work together in order to change. With our approach, we bring together retailers and brands, by breaking down the silos and creating a mutually beneficial network.

The future of forecasting

We are in the midst of a seismic shift. The intelligence of AI models is advancing at an unprecedented pace, and it’s changing everything we thought we knew about model development. Models like ChatGPT have set a new standard for smart technology.

Our own ForecastGPT-2.5 already outsmarts every alternative on the market. We’re combining pooled industry data, advanced deep learning technology, powerhouse computing, domain expertise, and insightful features to build an AI model that’s best in class.  This is reflected in the number of parameters our models contain. It indicates its ability to learn, adapt, and make accurate predictions. They encapsulate the ‘knowledge’ that the model has acquired from the training data. 

Unified Model: The Way Forward

Smart models like chatGPT are capable of handling multiple tasks. The old method of using separate models for different use cases (like replenishment, redistribution) has shown severe limitations and lacks further evolution. We are at the forefront of this change with our unified model approach. One model to handle multiple tasks, eliminating the need for customization and maintaining a holistic view of your business. The transformation of the industry is well underway, with pioneers like Shein and Walmart leading the charge- we follow their path. Bringing enormous potential of lost profits back to our clients wallets.

The power of features

In the world of deep learning AI, features are the key elements for success. They’re the pieces of information that the model uses to make a prediction. We don’t have to limit ourselves when it comes to the use and selection of features. We are harnessing the power of both internal and external data sources and work with over 100 of different features. Some examples: product sales history, prices, sizes, location, day of week, first sales, weather, seasonality, etcetera

 

Deep learning forecast GPT-2.5
Training unified
model
Pooled data

7.5m parameters WAIR GPT-2.5

Pooled data

Unified model

100+ Features

Computational power

Domain expertise

Initial distribution
Replenishment
Re-distribution
Forecast per moment time per aggregation level
Deep learning forecast GPT-2.5
Unified model
Deep learning forecast GPT-2.5
Unified model
External data features
Internal data features

The AI revolution in
forecasting

We redefine the boundaries of conventional forecasting models. Through innovative technology and unifying the industry (retailers & brands), we provide precise SKU sales forecasts, from a day up to six weeks.

Succes stories

Case #1 Shoeby

The AI Replenisher makes replenishment recommendations for 240 physical Shoeby stores at SKU level. This led to a 4% increase in inventory turnover, a 2% reduction in end-stock, and a 3% growth in total revenue.

Case #2 OFM

OFM was using a traditional replenishment system to keep the initial distribution intact without adapting stock levels based on demand.

Case #3 Daka

Daka reduced overstock by 47% percentage point and solved the edge case of seasonality

Case #4 Van Dal mannenmode

Van Dal used to employ a traditional static push strategy, where all stores received identical stock replenishment. Now, they’ve transitioned to a hyper-personalized approach, optimizing stock levels for each SKU and store based on actual daily customer demand.

Ready to make the difference?

Join the network of brands, retailers and ecosystem partners

Unleash the power of AI within your processes. Let us do the work we are good at.