In simple terms, a smart alert or anomaly is a data point that differs from others to a degree that is statistically significant. Data points in a cluster or pattern may have one or more data points outside those boundaries. These are outliers that indicate there may be a problem, such as a phony credit card in use, a data breach at a company or a doctor discovering that a patient has a serious disease.
For example, a retailer might find believe that conversion is down due to a lack of quality traffic to the store's website. However, sales might have dropped due to customers having issues stemming from a website update released the previous night. Or perhaps sales at an individual location are lower than usual or profit is less than expected despite overall sales remaining constant. Other problems could include higher returns, an inordinate amount of sick staff or product inventory missing from the warehouse.
In the past, your executives would uncover these anomalies using historical data. With today's sophisticated anomaly detection software, they can act as soon as unexplained anomalies occur.Discounted Flights
Here is another example. In 2015, American Airlines inadvertently sold some tickets for up to 90 percent off the regular price due to errors in the booking program. There was a problem with transposed currency values affecting the price of the tickets when the information entered the system. The result was that an international ticket that might have sold for $4,000 (USD) was sold for 10 percent of that amount. Earlier that year, an almost identical situation happened to United Airlines. Travelers flying across the Atlantic could buy tickets for a huge discount. Some fortunate souls were even able to grab first class tickets for under $100 (USD).Domino Effect
These two stories are interesting to examine because of the nature of the airline industry. In recent years, airlines have expanded their computer operations into large, complicated systems that instantly tie to a massive network of online websites, smartphone apps and airport terminals. Due to the significant consolidation of airlines over the last 10 years, if any major carrier has a technical glitch, it affects travel on almost every other airline.
In 2016, thousands of travelers relying on Delta Airlines were stranded due to a computer outage that canceled 650 flights and delayed another two thousand. The system was fixed in less than 24 hours, but the domino effect took hold and rippled through the commercial aviation world.Interdependent Networks
Although the most significant problems took place at Atlanta's Hartsfield Jackson Atlanta International Airport, where a large majority of passengers use Delta, the outage affected multiple flights at Los Angeles International Airport and others around the globe. Aviation industry experts said the cascading outages were due to the elaborate, interconnected data systems—for which the airlines have neither the funds nor the sufficient staff to operate at peak efficiency.Benefits of Anomaly Detection
It's easy to pick on the airline industry because we all have experienced less than ideal travel conditions. However, today almost every industry relies on intricate networks of high-speed computers and huge data streams in the form of website traffic, e-commerce, customer account management, order processing and tracking, and much more. This is where anomaly detection software can really help IT teams. It can:
- Provide near-real-time results. No matter how much IT staff you have, you'll never have enough manpower to provide extremely close monitoring of the massive amounts of big data used by today's organizations. Anomaly detection programs provide near-real-time monitoring of your data streams, allowing your technicians to adjust as problems are uncovered.
- Reduce dependence on thresholds. How do you know what anomalous behaviors you should monitor to trigger alerts? Let's say you have a large enterprise with hundreds of applications, thousands of users and dozens of servers. How can you know which thresholds to establish? Anomaly detection applications automatically establish normal behavior baselines.
- Detect problems on the fly. Instead of waiting for problems to happen, smart alerts find problems as they occur, giving you an advance warning. This gives your IT team more time to respond to difficulties before they negatively affect your end users and customers.
- Speed up troubleshooting. Anomaly detection software lets you rapidly detect and manage emerging problems. These programs use machine learning to understand the normal state of affairs in your organization. They monitor several domains, including systems, networks, middleware and servers, alerting you to which domain might be having trouble. By isolating trouble spots, you reduce troubleshooting response time.
- Cut down on expert manpower. Emerging problems often demand the attention of your most talented subject matter experts. Anomaly detection software eliminates the cost of removing your most valuable personnel from their primary job duties.
An automated system that adapts on its own is the key to effective anomaly detection. Machine learning algorithms are easier to use than ever before and more effective at understanding how anomalies affect systems.
AVORA allows you to use current data to quickly resolve potential problems and lets you make faster business decisions. It is not a substitute for your current business intelligence programs, but it is a powerful enhancement. You create a few simple metrics, and the system will do the rest, potentially saving your organization hundreds of thousands of pounds.