Digital, network-connected systems are transforming every aspect of business — from your mission-critical workloads to your most rarely used applications. But the increases in scalability and cost efficiency come at a cost.

Because every system is so reliant on network connectivity, unplanned downtime is becoming increasingly expensive. What happens when you have total system outages? Or when you have a minor glitch that’s preventing eCommerce checkouts? Or, when you have a software bug causing micro-glitches that lead customers to find new solutions?

Even the smallest issues can have major implications for your bottom line. But spotting slow business leaks is often easier said than done. Rather than flying blind on slow business leaks, you can integrate real-time anomaly detection to help stay ahead of glitches, bugs and downtime. These four companies could have avoided serious business disasters if they had done the same.


Disaster #1: Knight Capital

Outages are usually credited as the most costly business disasters. And at an average cost of $300,000 per hour, that’s no surprise. That’s a drop in the bucket compared to what happened when global finserv firm Knight Capital in 2012 experienced a computer glitch in a new trading software.

Within 45 minutes, the company’s trading systems bought and sold millions of shares in hundreds of different stocks before the issue was resolved. Meanwhile, they traded overvalued shares back to the market and lost the company $440-million.

The incident lasted less than an hour, but there were opportunities to identify operational anomalies the day of the glitch and in the days following Knight Capital’s rollout of new trading software.

Every second counts in situations like this one, and autonomous detection could have saved the company hundreds of millions of dollars.


Disaster #2: Wells Fargo

In November 2018, Wells Fargo said that it identified a computer glitch that had been quietly impacting the business (and its customers) for more than eight years. You read that right – EIGHT YEARS.

The glitch in its underwriting software resulted in 870 loan modification rejections — 545 of which caused homes to go into foreclosure.

According to the company, the glitch included calculation errors that disqualified loan modification candidates who should have been approved. The total amount of money Wells Fargo paid out as a result is unclear, but one customer reportedly received $25,000 in damages.

It’s possible that the analytics team was searching for anomalies manually. But whatever processes were in place, it’s clear that a small leak got past the team and led to this widespread issue.

If the analytics team had been backed up by AI-driven, real-time anomaly detection, they might have caught this glitch before it had such costly consequences.


Disaster #3: Macy’s

It seems every holiday season, there’s at least one mega-retailer who experiences a data glitch of massive proportions. This year it was Macy’s. On Black Friday, the department store chain’s credit card system went down which triggered revenue loss both in stores and online.

No retailer is immune to the small glitches that lead to revenue loss and lost revenue and damaged reputations. Even though the outages from high-volume traffic on Cyber Monday and Black Friday receive all the headlines, other problems like incorrect pricing, mistargeted advertising, and missed upsell opportunities all hurt the bottom line.

If you look closely enough, something like a widespread credit card outage doesn’t happen suddenly. You can track the buildup of small transaction failures, increased latency, and other warning signs that a problem is on the horizon. But even if you have a full-time team dedicated to looking for these issues, high-volume situations make it impossible to track them manually.

Real-time anomaly detection can automatically present these warning signs to system admins so they can proactively address issues before there’s an outage. For Macy’s, this might have meant some slow response times for the first few transactions rather than a full-blown outage with potentially millions of dollars in damages.


Disaster #4: The IRS

The use cases for real-time anomaly detection aren’t limited to financial transactions like in banking or eCommerce. Behavioral anomalies can plague any system in any industry — and the government is no exception.

Back in 2006, the Internal Revenue Service (IRS) was entering its tenth year using the same software for fraud detection in tax returns. They were promised a new version of the software, so they shut down the original and waited. Except the new software wasn’t implemented before the deadline and glitches in the system resulted in 66% of fraudulent claims making it through screening.

The inability to spot fraudulent claims cost the government between $200 million and $300 million that year. Now, rather than baking anomaly detection into the tax software, it can exist as its own entity, spotting suspicious claims in real time and alerting officials before approval.


Automated Real-Time Anomaly Detection: Prevent Losses, Spot Opportunities

The first key to any real-time anomaly detection system is to gain a baseline of your entire business. That means learning the normal behavior of each system, across all business metrics, and constantly monitoring against that baseline.

Automating this process is essential for your success, because there’s just too much information for any team to manually process. But once you have automated real-time anomaly detection in place, you can avoid disasters like the ones listed above with the help of:

  • Immediate alerts sent via text, email, Slack, Jira, and other channels to ensure admins act on anomalies before it’s too late
  • Deep integration that ensures accurate data across applications databases, storage, CRM, IT infrastructure, and more
  • Automated root cause analysis that links anomalies to business incidents, cutting through the clutter of metrics to solve real issues.

These benefits help real-time anomaly detection prevent issues that will negatively impact your bottom line. But they can also be used to spot positive opportunities. For example, real-time anomaly detection can pinpoint when an upsell opportunity is working, giving you the insights necessary to invest more in that existing campaign or tactic.

Whether you’re fixing negative business incidents or trying to do more of what works, real-time anomaly detection gives you the insights necessary for success.

To learn more about how this can work for your business, download our eBook on AI-powered analytics for anomaly detection.

Written by Nir Kalish

Nir Kalish is the senior director of solution engineering and customer success at Anodot. He has more than 15 years of experience in the software industry, leading groups of QA automation, QA infrastructure and solution engineering. Nir is passionate about the two complimentary worlds of business and technology, and especially enjoys developing strategies to resolve data mysteries while upholding exceptional user experiences.

You'll believe it when you see it