Journey Foods reduced costs, boosted accuracy

Industry

Business

Use Case

Product

Who is Journey Foods

Journey Foods is a portfolio intelligence and lifecycle management software for food development and innovation. Journey Foods addresses and manages the complete lifecycle of products from ideation to the marketplace.

Main Findings

  • Used its existing team and resources to implement the Machine Learning project. 
  • Implemented ML quicker and easier than with traditional ML solutions.
  • Implemented cost prediction analytics for CPG customers
  • Applied ML directly to data in MongoDB, reducing MLops costs

Who is Journey Foods

Journey Foods is a portfolio intelligence and lifecycle management software for food development and innovation. Journey Foods addresses and manages the complete lifecycle of products from ideation to the marketplace.

Main Findings

  • Used its existing team and resources to implement the Machine Learning project. 
  • Implemented ML quicker and easier than with traditional ML solutions.
  • Implemented cost prediction analytics for CPG customers
  • Applied ML directly to data in MongoDB, reducing MLops costs
Learn how it was built:
Link↗
Check out the code:
Link↗
Try out:
Link↗

Challenges

Journey Foods holds a database of 130,000 food ingredients and 22,000 suppliers and sources with analytics tools and dashboards.  Journey Foods wished to implement cost prediction for its customers - accurate costs for ingredients projected into the future 1, 3, 6 and 12 months.

Ingredients and suppliers are constantly changing, and predictive analytics requires mapping complex, sometimes ‘many to many’ relationships.  For a traditional ML solution, data prep and transformation pipelines would need to be maintained and frequently updated, in addition to the requirement for frequent, manual re-training of ML predictors for forecasting.  Additionally, the ingredient data is held in MongoDB collections, and organizing those collections to reflect the cost and ingredient category prediction is cost and labor prohibitive.  For example, in the case of predicting cost changes to ‘Coconut Oil’ you would need to include categories ‘Coconut Oil Extract’, ‘Refined Coconut Oil’, ‘Unrefined Coconut Oil’, etc.  For oil-based ingredients, many-to-many category mappings could include various coconut oils and derivatives, unrefined and refined oils.  Each would require a Mongo collection containing the category members, and would significantly increase the storage and data prep requirements for successful prediction.

MindsDB Solution

MindsDB allows Journey Foods to connect its AutoML cloud service directly to MongoDB and to perform data prep in addition to training and querying machine learning models directly from Mongo using the native query language (MQL).   Journey Foods React JS front end uses code to query the Mongo back-end for rendering dashboards in exactly the same way both for ML-generated predictive and non-predictive analytics, without the need for building and maintaining custom data pipelines for ML predictions.  Additionally, MindsDB can be used to predict the ingredient category, rather than creating additional Mongo collections to account for all categories and ingredients.  Thus, in the case that historical pricing is not available for a particular ingredient, MindsDB predicts the ingredient category, and then predicts the future price for that category.

Results

The overall solution that the Journey Foods team has developed offers a few key benefits to their business:

MindsDB powered predictive analytics offers accurate cost predictions for food ingredients.

MindsDB allows Journey Foods to build and maintain ML-based predictive features into their product quickly and inexpensively

Journey Foods reduced costs, boosted accuracy

MindsDB operational costs are considerably lower than a homegrown ML solution, and significantly reduces MLops overhead

Make the most of your data. Use AI to power prediction and analysis capabilities, and create actionable strategies.

Get Started for FREE