Moving Away from the ‘Red Sea’ of Rule Based Alerts
Traditional alert systems for optimizing business performance tend to favor a rule-based approach. ‘If this then that’ rules are based on past experiences, such as anticipated operational slowdown of machines, or cash flow issues at the end of the month. as well as human intuition, for example choosing which person should be the first to know if a delivery doesn’t arrive on time.
However, in a complex business landscape, rule-based alerts are being increasingly usurped by intelligence-led models, leveraging Machine Learning and Artificial Intelligence to drive business competitive advantage. Here are the top ways that Machine Learning thrives where rule-based models fall short.
A common frustration when it comes to rule-based alerts are the amount of notifications that turn out to be false positives. A maintenance crew is sent to fix an error but the alert was triggered by adverse weather conditions. A tenant seemingly didn’t log into a resident portal because they were using a new device. Individually they are each a minor inconvenience, but a slew of false positives can lead to staff circumventing the alerts system altogether, bored of the long list of red items that they don’t trust to be correct.
According to Bloomberg, enterprises like Microsoft and Amazon, as well as start-ups and SMBs are moving away from rule-based technology when it comes to detecting anomalies and intrusions. One example is a recent rule-based Microsoft system that aimed to prevent false logins. Due to a 2.8% rate of false positives, (equal to millions of user logins in some larger companies) Microsoft moved over to a Machine Learning approach. Once the new system was rolled out, the percentage dropped to just .001%. Rule-based models are not accurate enough to rely on.
Another consideration is that rule-based alerts can only protect based on existing information. When it comes to a dynamic environment with new technological adoptions, or for sensitive areas like cybersecurity, this is never going to be enough to gain a competitive edge.
“Machine Learning is a very powerful technique…it’s dynamic, while rules-based systems are very rigid,” says Dawn Song, a professor at the University of California at Berkeley’s Artificial Intelligence Research Lab. With Machine Learning, your algorithms can learn from normal protocol and create baselines to spot anything out of the ordinary, even where behavior has not been seen before. While rule-based alerts might allow a new behavioral anomaly to pass by undetected, Machine Learning would be able to identify anything unusual, and alert the relevant stakeholders in real-time.
Machine Learning has eliminated a lot of frustrating IT admin and streamlined the idea of alerts for operational excellence. When it comes to rules-based systems, Song continues, “it’s a very manual intensive process to change them, whereas machine learning is automated, dynamic and you can retrain it easily.” This means that rather than stay fenced in with inflexible policy that may not suit your new technology or ever-changing work environment, you can trust that your algorithms will scale and adapt as you do.
Machine Learning also removes the operational burden for the key personnel on the ground. Your staff, from maintenance crews to managers have less false positives, and are relieved from the list of high priority action items that never seems to end. Instead, they can enjoy a move ‘back to black’ – more accurate and actionable alerts that are targeted to their roles and work towards overarching business objectives as a company.
Rule-based alerts are not built to handle a complex set of data points, or spot patterns in the trends that you’re experiencing as a company. At the same time as providing actionable analytics for stakeholders, Machine Learning studies your data and opens up new disruptive opportunities for your company.
Take predictive analytics for example. Rule based alerts might send your engineers to fix dozens of machines over the course of a week, working in crisis mode. Only Machine Learning can use parameters such as the machine type, room sensors, the amount of workers and more, to organise for proactive maintenance on machines before they break down, pre-empting a problem before your tenants have a chance to complain. Unlike the simple use of specific information to trigger an action, maximizing your data’s potential with Machine Learning is a true competitive advantage.
Today’s Dynamic Landscape
Cloud technology and Artificial Intelligence are here to stay, and there are benefits to be found in every industry. You can bet that if you aren’t taking advantage of the latest advances, your competition will be. With Machine Learning algorithms, you can combine cloud technology and AI to collate data from multiple streams, including internal datasets, third-party data such as weather or traffic, and external information such as building sensors and IoT devices. This enriched data leaves rule-based systems head and shoulders below the curve.
At Okapi, we use this dynamic data with 4,000 metrics and algorithms to build targeted action points for each member of staff in your company. Alerts are accurate and enriched, ensuring the only red items are truly actionable and urgent, and your personnel can always be one step ahead.