Request POC

The purpose of the Free PoC (Proof of Concept) is to review the feasibility of applying AI-based demand forecasting using some of the customer's product data. The results provided include predicted values for each product, visualization materials, and prediction accuracy metrics per SKU, allowing you to pre-examine the potential for AI adoption and expected effects.

The Free PoC utilizes the minimum data required to configure a time series prediction model. This includes basic item information and sales (shipment) history data.The Paid PoC utilizes additional data for advanced prediction accuracy, in addition to the Free PoC data. This may include promotion information, external variables (e.g., weather, macroeconomic indicators), inventory status, and other business-related metrics.

Sales history is essential as demand forecasting is based on product-specific time series data. A minimum of 24 months of data is required for monthly forecasting, and a minimum of 12 months is required for weekly forecasting. The longer the data period, the more favorable it is for improving prediction accuracy.

The possession of product-specific time series data is the most important requirement. If sufficient data for prediction is secured, you can receive the results within 2 weeks after the data is received.

There are no industry restrictions, including manufacturing. However, due to the nature of the data, industries with Sell-out data, such as food, retail, F&B, and distribution, tend to show higher average prediction accuracy.

The predicted values and accuracy provided in the Free PoC are verification results offered within the scope of model evaluation standards, so there are limitations to applying them directly to real-world tasks. However, it has sufficient review value for judging the form of AI-based prediction results and the feasibility of demand forecasting for each product, so it is recommended as a pre-verification step before adoption.

We can suggest other forecasting approaches or research methods besides the time series model, or conduct additional feature verification by selecting suitable product groups or categories. If more in-depth verification is required, precise modeling and analysis are conducted through a Paid PoC.

The Free PoC is conducted through AutoML-based automatic verification using over 200 models applied within DeepFlow. It is a stage of exploring prediction feasibility for each product without human resource investment.

The Paid PoC involves an AI engineer directly participating to design customized models to achieve the target accuracy. Sales data and additional variables are learned, and precise tests are performed based on the agreed period and product group.

A cost discount is provided upon official adoption for redundant efforts incurred during the PoC process.

Yes, it is possible to apply, but each PoC must utilize a different dataset.

The person in charge will contact you within 3 business days after applying. After the data is received, the results will be provided within approximately 2 weeks.