Artificial Intelligence utilizing true unsupervised machine learning to identify anomalous patterns
Data that has been ingested and normalized paves the way to apply machine learning models for anomaly detection that uncovers bad behaviors, human errors, and data glitches in a TMS, WMS, OMS, or another system. Armed with this knowledge, you can:
Intelligent Audit’s proprietary machine learning algorithms are an ensemble of auto-encoders, boosted decision trees, and LSTM (long short-term memory) deep learning frameworks, which can give invoice-level results in real-time.
Training the algorithms includes a combination of self-supervised learning and supervised learning, which uses fully labeled datasets and leverages important insights on historical data made by account managers and other company experts.
The input data includes dozens of variables over hundreds of millions of entries and leverages both known KPIs (for example, the amount charged or the total weight) and the more abstract, auto-generated features.
Due to the relative rarity of anomalous events, the algorithms are tuned to deal with heavily imbalanced datasets. The end result is a suite of detection algorithms with high sensitivity and specificity in the face of enormous amounts of data.
Once you clearly understand that something has changed, you can begin to assess whether it truly is a problem and determine the impact it is having on your business. The visibility Intelligent Audit provides makes this possible.
The problem could be as simple as a misplaced comma in data entry, transforming a 10-pound shipment into 1,000 pounds.
Most shippers have metrics in place to monitor the standard Key Performance Indicators (KPIs) - spend, volumes, cost/shipment, cost/kg, cost/mile, etc. For a shipper spending millions on transportation annually, it is unlikely for an anomalous pattern to occur that rises to the degree of changing any of those KPIs at a macro level. When they do figure it out, it's after they have spent irrevocable money that could have been an easy fix had they been made aware of it quickly.
This is where true unsupervised machine learning is impactful. Artificial intelligence is able to detect the slightest new pattern in your transportation—data across any type of KPI—and makes you aware quickly before you end up spending millions of dollars on a technical glitch or human error.
An e-commerce initiative with hopes of high profits incurred high costs instead. Our system quickly detected that large-package fees were to blame. Armed with this info, the client pivoted to regain profitability.
A new carrier for an infrequently used lane’s rates were 50x the agreed-upon rates, causing costs to skyrocket. Our system quickly detected the issue and the cause, enabling the client to quit using the carrier and save thousands.
New team members keyed in weights European-style (5,00 vs. 5.00) causing 5 lb shipments to become 500 lb shipments. Our system quickly detected the issue, enabling the client to stop spending thousands.
A carrier quoted lower rates, concealing that beyond charges would be separate, causing costs to unexpectedly spike. Our system quickly detected the fees, enabling the client to stop using the carrier before costs got out of hand.
A client with a contract on detention fees had an anomalous spike. Our system quickly detected the issue and worked with the client to sort out the ports causing the problem, plan better, and control costs.
A new pattern of accessorial fees caused a client’s cost-per-shipment to rise without warning. Our system quickly detected the problem and helped the client come up with an action plan to reduce future fees.
A frequent same-day shipper suddenly started incurring excess mileage fees. Our system detected the anomaly and dove deep to determine the charges were unprecedented. This enabled the client to receive a full refund from the carrier and prevent future erroneous charges.
