It’s real: Data-Driven Product Authenticity™ with Machine Learning

One of our big goals at EVRYTHNG is to change the way supply chain integrity and product authenticity issues are tackled. Issues like counterfeit products, gray market imports and the production of  backdoor goods. Today most of these challenges are resolved using ‘material science’ techniques: putting ‘hard to copy’ tags on products. For example, tags with patterns that decay when copied or encrypted RFID or NFC tags.  While these tags are valid mechanisms to distinguish counterfeit products from genuine products or detect fraud in supply chains — they also present several disadvantages: 

 

  • These product tagging techniques are significantly more complex than simple QR codes;
  • They often require the use of proprietary tags or printing techniques which equates to vendor lock-in;
  • Consumers need special applications to interact with the tags; and
  • All of these factors together have an impact on cost, scale and usability making these techniques completely impractical for industries like, consumer product goods 

Our approach is to resolve these problems through the new currency: data

 

Data-driven Product Authenticity™ 

The principles of Data-driven Product Authenticity™ are simple: serialize your products using Digital Link QR codes or low-cost NFC tags and gather as much data as you can throughout the product’s supply chain journey and into the hands of the consumer. Supply chain integrity decisions can then be made based on real-time data intelligence. The highest level of integrity is achieved by combining data-driven techniques with material science techniques. In the past however, cost has been an issue in taking this approach.

 

The EVRYTHNG platform until now has delivered Data-driven Product Authenticity™ through either our Reactor  business rules engine or Factory Activation (authenticating products at source)

 

We are excited to announce however, we’ve pushed the boundaries, breaking through previous limitations to cost-effectively add machine learning to our Data-driven Product Authenticity™ offering.   Welcome to Predictor™, our new machine learning engine

 

Read on to get the details…

Why Machine Learning?

If we already have an effective tool for data-driven authenticity, you may be asking — why invest in machine learning? What additional value can it bring to companies in the consumer products industry?

 

Achieving better supply chain integrity begins with the mass-scale serialization of your products on the EVRYTHNG platform. Several intelligence engines are then configured:

 

  • Redirector  – this component uses contextual inputs, geo-location, user’s profile, time, date, etc. to dynamically determine what URL the QR code or NFC tag redirects to after a product scan. This is a very powerful tool, perfect for creating application workflows for expected interactions or usages.  
  • Reactor  – is a Javascript-based rules engine for creating bespoke business logic.  For example, in the case of a product recall due to a safety issue with product packaging, Reactor™ can automatically determine whether a unique item was part of a faulty batch by analyzing the product hierarchy to see component parts and production data. 

Given the number of parameters, rules can become complex. Real world scenarios are difficult to predict and/or react to. The onus is on humans to understand and update workflows as events change.  This is where machine learning comes into play.  

 

  • Predictor  – This is EVRYTHNG’s new machine learning engine. It learns from historical data to predict outcomes and identify anomalies.  With Predictor, the EVRYTHNG platform can adapt on its own, automating the decision-making.   For example, applying the determination; “Based on similar products, with similar metadata, and scan patterns,  this item is likely to be counterfeit.”

 

EVRYTHNG Product Developments 

The possibilities of machine learning are endless. We’ve focused our initial developments on a specific set of use cases where we believe machine learning will play an increasing role. 

 

  1. Counterfeit detection
  2. Gray Market detection 
  3. Spotting data integrity issues with supply chain data
  4. Predicting consumer engagement rates
  5. Dynamically altering brand content and experience delivered to consumers on scanning products

We’re working towards making machine learning easily deployable for our customers, creating off-the-shelf models available via our self-service portal. 

Activating machine learning via the EVRYTHNG dashboard

Think of it like an App Store for machine learning, tailored (initially) for resolving supply chain integrity issues.  It’s all about moving machine learning from the back-office domain of data scientists to become a central component in any connected-product application, accessible and valuable to any business user.    After all, the more products our customers digitize, the more data is generated. The more data generated, the greater the value of machine learning when applied.  

 

virtuous circle, the benefits of which we expect will surprise even us! 

 

Beta Trialists Welcome

We are excited to share we’ve officially implemented an early stage version of Predictor™ and are ready to roll out machine learning pilots in the latter half of 2019. We’re working with several consumer product brands on machine learning initiatives and welcome others. Please reach out directly if you’d like to get involved.  And if you’d like more technical information, check out our developer tutorial and API documentation too.  

 

Thanks for reading!

 

Dom

 

About the authorDom is Co-Founder and CTO of EVRYTHNG.  An IoT expert, he spent more than a decade working on projects for Oracle, the Auto-ID Labs, Nokia and SAP. Currently a faculty researcher for the Blockchain Research Institute, Dom is working on the potential of blockchains for the Internet of Things (IoT). Dom was a Web of Things (WoT) pioneer at ETH Zurich and MIT, and his Ph.D on the WoT was granted the ETH Medal.

image_pdfimage_print