Sterling B2B Integration

Sterling B2B Integration

Come for answers, stay for best practices. All we're missing is you.

 View Only

How to get more from your BTI investment with AI-powered Anomaly Detection

By Kiran Rao posted Mon May 06, 2019 04:12 PM

  

In Lori Webber’s recap of Think, one of the areas she promised you’d hear more about is Anomaly Detection. Business Transaction Intelligence (BTI), an SCBN capability, uses Watson AI to provide business users with insights into B2B transactions quickly and easily. If you watched this webcast, then you’re aware that we continue to make significant investments in AI and on December 8, 2018 we released our new Anomaly Detection capability.

Built-in to BTI, Anomaly Detection monitors day to day document trends between you and your trading partners. When it discovers an anomaly, it provides an early warning signal so you can investigate and proactively address the problem, instead of hearing about it from a customer. It works behind the scenes, tracking the contents of EDI-based supply chain documents and learning what is typical activity across three categories of data:

  1. Document counts – how many specific document types are exchanged
  2. Document value – the actual dollar value based on the contents within specific documents types, such as purchase orders or invoices
  3. Document volume – the amount of data being processed in kilobytes across the network

Anomaly Detection develops a baseline understanding of expected activity across these three categories for various time periods (e.g., specific days, weekly, monthly, seasonally, etc.) and shows you if the document count, value or volume is out of the expected range. Using the search bar, you can drill down with key words such as “volume”, “invoices”, “850”, “order value”, etc. [see Figure 1]. This allows you to quickly hone in on the specific document type – a purchase order, invoice or Advanced Ship Notice (ASN) – and the trading partners contributing to an anomaly. You can learn more about a potential problem and take corrective action sooner to preempt disruptions to the business.  


Figure 1. You can use the search bar to quickly view anomalies.

Anomaly Detection provides business users, including customer support reps and vendor/supplier managers, greater insights from procure-to-pay and order-to-cash so they can improve customer service and meet their business objectives. Technical users also benefit because they can quickly see and fix EDI configuration problems to ensure EDI documents are sent and received smoothly. Let’s look at a couple of use cases.


Use Case #1: Business user

Imagine a scenario where invoices are sent out every Monday, but this particular Monday there is a problem. As a customer support rep using Anomaly Detection within BTI, you quickly see a document value anomaly this Monday. You can pull up the list of top trading partner contributors and see that the typical value for RBG Company is in the $200,000 - $250,000 range, but today it is $0. You wonder if they cancelled their order or submitted a change order.

To get more information you type in the search bar of BTI: “Show me RGB Company’s cancellations” or “Show me RGB Company’s change orders”. If you don’t see evidence of either, you can drill down further and search: “Show me invoices for RGB Company”. You now see that invoice values for the day were zero. You look at your flows (the status progress of transactions) and discover that all flows stop at the invoice milestone. You can see an order, an acknowledgement, a shipment and an invoice created.

You click on one of those flows and find that when the invoice was sent a 997 EDI document was received, rejecting the invoice because the system couldn’t read it. You quickly involve your EDI expert who sees that the document wasn’t configured properly and fixes it so that the invoice can be resent, accepted and processed in a timely fashion.

Without BTI and Anomaly Detection, the situation would not have been noticed for days and could have had spiraling effects. Your Days Sales Outstanding (DSO) would have increased and the trading partner would have been unhappy since the invoice was late and automatically prevented you from getting their 1%/10/30 terms. You might need to involve the CFO to make a manual adjustment, and your EDI experts would have to spend hours investigating the situation to determine the scope of the problem and how to fix it.


Use Case #2: Technical user

In this scenario, as an IT expert using BTI and Anomaly Detection, you receive a notification of a volume anomaly. There’s a spike in network traffic above the usual kilobytes expected. You do a search within BTI for EDI 214 document volumes and discover an unusually high number of shipment status messages. You look at the list of top trading partner contributors and see that the volume with Carrier West is unusually high. Your system is caught in a loop, repeatedly asking Carrier West the status of your shipment, and the carrier’s system continues to reply to each request. With these insights you can quickly go in and configure the EDI 214 request back to normal so that you don’t surpass bandwidth limits. Without BTI and Anomaly Detection the loop could continue to run for days before being discovered, likely resulting in overage fees which can run in the thousands of dollars.


Under the hood of Anomaly Detection

Anomaly Detection is preconfigured using average normal range settings out-of-the-box. From the moment BTI is activated, Anomaly Detection begins to adjust these ranges based on historical data in your BTI solution and within two weeks starts to project expected ranges for your organization. Once these values appear, these predictions/ranges should be accurate, and will continue to improve over time as more data is collected and time passes. The model considers day, week, month, season, and year time series and builds a model to forecast around these recurring periodic occurrences.

To adjust for seasonal variations, it uses changepoint detection so that it doesn’t trigger anomalies in error, for example during the winter when orders for snowblowers spike. As values go back down in the spring Anomaly Detection adjusts the expected range to reflect the shift. Months later it will remember the seasonal trend and adjust ranges upward.  


How to get started

If you haven’t had the opportunity yet, watch this 30-minute webcast for more details on what you can do with Anomaly Detection today and what’s on the roadmap. After that, your local IBM sales rep, Customer Support, or IBM Business Partner can help you get started.

Let us know what you think about Anomaly Detection! Feel free to ask questions or provide feedback in the discussion forum.

 


#SupplyChain
#B2BIntegration
0 comments
72 views

Permalink