Case Studies
Successful B2B case studies for various business use-cases
o1.
classify



ProfilePrint was developed to address the fragmented food ingredient industry’s challenges of rampant adulteration, quality inconsistencies and taste inaccuracies.
Specimen type: Green Coffee Beans and Freeze-Dried Instant Coffee
Background: Client is a reputable OEM of premium blended coffee products for global brands. Headquartered in Switzerland with manufacturing facilities in South East Asia.
Objectives: Using ProfilePrint’s QC model, to rapidly ascertain the consistency of raw ingredients and finished products instead of the existing complex and onerous methods via instrumentations and sensory panels.
Results: ProfilePrint was able to predict with an accuracy of 100.0% for green beans and 96.3% for freeze dried coffee within seconds.













o2.
classify



Specimen Type: Cocoa Mass, Compound Chocolate, Real Chocolate
Background: Client is a Singapore-based company that produces both chocolate covertures and compounds. With 6 brands under its wing, they export their products to over 55 countries all over the world and source supplies globally.
Objective: Using ProfilePrint QC module to conduct rapid check for consistency of its incoming ingredients to ensure that they meet the company’s internal QC requirement during product manufacturing and product development.
Results: ProfilePrint was able to predict with an accuracy of 98.2% for Cocoa Mass, 100.0% for Compound Chocolate and 94.4% for Real Chocolate within seconds.



o3.
PROFILE



Specimen Type: Tea leaves
Background: Client is a listed company in China. They specialise in the production and distribution of instant noodles and RTD beverages.
Objective (1): Using ProfilePrint’s Profile Prediction module to shorten New Product Development cycles.
Results: Client provided thirty-three (33) samples of oolong tea leaves and a specimen of the target oolong tea standard with sensory data.
The model consists of 8 parameters – four describing taste and four describing aroma. The AI models were successfully established on the ProfilePrint platform, with a low mean relative error value of 7.9% and 9.0% respectively.











o4.
blend



Specimen Type: Tea leaves
Background: Client is a listed company in China. They specialise in the production and distribution of instant noodles and RTD beverages.
Objective (2): Using ProfilePrint Blend Recommendation module to proposed blend percentage of all available samples to achieve the target taste profile provided by the client, greatly reducing time and effort required for new product development.
Results: To verify the blend prediction, samples of the proposed blend were mixed using existing samples of dry oolong tea leaves. The proposed blend has close accuracy of 92.7% match against the target within seconds instead of the traditional process that requires expert tasters over few hours.


