KavaLytics™ provides clients with a fast, affordable, accurate and validated quality assessment method for Kava using a state of the art handheld Near Infrared spectrometer, paired with a powerful, cloud-based artificial intelligence machine-learning platform that is intended to implement quality controls and traceability throughout the Kava value chain. Simply point the scanner at a bag of Kava powder and get instant lab quality results on any blue tooth enabled IOS device. Print, save, or email the analysis results on the spot, or create a label to apply to the product, complete with a QR code for quality control, traceability compliance and added value at each step of the value chain.
KavaLytics™ V5.2.1 Model Features
- Determine the total Kavalactone content
- Determine if Kava is Beverage Grade or Non-Beverage Grade
- Determine if Kava is above 0.25% Total Flavokavain Level
- Predict moisture content
- Predict the first three numbers of the chemotype i.e. “4-2-3”(Beta)
- Predict total Flavokavain A+B content (Beta)
- Detect common adulterants of Kava:
- Stump and stem peelings or "Black Kasa"
- Peeled upper stems or ”White Kasa"
- Spent kava or “Makas”
- Green aerial parts such as stalks and leaves
The new standard in kava quality testing and traceability
KavaLytics™ is the result of listening to the quality control and traceability needs of kava industry stakeholders at all levels of the value chain in 7 kava producing regions. Based upon industry feedback, KavaLytics™ proudly partnered with the data science experts at Sagitto in Hamilton New Zealand, to build a machine-learning platform that was trained on over 500 analyses of kava performed by a highly credentialed ISO-17025 analytical lab with vast experience in kava analysis. The result is the most technologically advanced, comprehensive and robust kava analysis system ever built. Thanks to machine learning, the method continues to learn from each new kava sample it is given and will always be at the forefront of fast, affordable, and accurate kava analysis. As you read this paper, the KavaLytics™ method has already learned from new data acquired by a user scan.
Frequently Asked Questions
How accurate is the current KavaLytics™ model?
KavaLytics is currently able to correctly determine if a sample is Beverrage Grade or Non-Beverage Grade Kava in over 95% of samples. The model can also predict the total kavalactone content of kava powder within 0.7% of a properly calibrated HPLC at a highly credentialed ISO-17025 Lab. It can determine if a sample is above 0.25% Total Flavokavains in over 92.5% of instances and it can predict moisture content down to 0.2%.
The next version of the model will include another 50-60 samples and we expect to improve upon this model's accuracy. With the continued support of kava exporters and governments, the system continues to progress and new features are added with model updates.
How often is the Model Updated?
The model is updated every other month and generally includes an additional 50-60 samples. We prioritize sample sets that will lead to the greatest improvement of the model. For example, if the most inaccurate predictions are coming from 100% lateral roots from Tonga, then we train the model with an additional 4-5 samples of Tongan lateral roots so that we can bring the prediction accuracy above and beyond the average accuracy of the current model.
The updated version of the model is automatically utilized by the device, requiring no action on the users part. You always have the most refined version of the model every time you use it.
Advantages over traditional HPLC and "Other" quality assessment methods
There are several advantages to implementing KavaLytics™ as an initial quality assessment model as opposed to HPLC, TLC, and UV/Acetone methods.
No more waiting weeks for HPLCs to come back.
Receive the Certificate of Analysis on your phone, email it as a PDF from the cloud, or print a sticker via any IOS compatable network printer.
No wet chemistry!
Say goodbye to harsh chemicals like acetone, diethyl ether, methanol, and acetyl nitrile.
No complicated Sample preparation
Simply point and shoot at a bag of powder.
Low initial cost
Low initial cost compared to HPLC, GC/MS, and other methods
Low staff training costs
Very low staff training costs – Simpler than the Acetone method.
Low ongoing testing cost
$25 per sample compared to paying lab tech hours to operate sophisticated equipment.
High degree of accuracy
Highly accurate - Repeatable results.
Whether deploying 1 or 100 units, they work out of the box and provide consistent results.
Units automatically utilize the latest version of the model as revisions are made
Built in Traceability
Scannable QR Codes Show movement through the supply chain