peshkov - stock.adobe.com
Customer experience software vendor Gainsight had struggled to surface anomalies in its data using open source software. So, about two years ago, Gainsight turned to another software vendor for its business monitoring and AI in forecasting needs.
Gainsight has been using Anodot, a 2014 startup based in Redwood City, Calif., for the past two years. The relationship with Anodot began because of a DevOps type of need, said Mickey Alon, founder and CTO of Gainsight.
AI in forecasting
Gainsight, which sells several customer experience products, including platforms for customer success analytics, had been using log analysis tools that provide basic, and not particularly accurate, machine learning and anomaly detection elements, Alon said. Those tools have the potential to work well, but companies without dedicated data scientists for anomaly detection, such as Gainsight, often struggle to have them work to the best of their ability, he said.
"With Anodot, it was almost plug and play," Alon said.
"It's very easy to understand what's going on," he continued. The Anodot platform shows graphically normal usage compared to what is happening in real time, without the need for a dedicated data science team.
Anodot enables Gainsight to better monitor its customer usage data, one of the company's key metrics. Gainsight is able to better monitor its customer usage data to, say, understand if there is an increase in use because of a Black Friday spike, for example, or because of an opportunity to upsell.
Mickey AlonCTO and founder, Gainsight
"We are able to deliver a better business experience. We are able to see if something is changing and notice that very close to when it is happening and then understand the impact of that on the customer," Alon said.
Usually, Anodot serves as a replacement for a system that doesn't use AI for forecasting, such as an open source dashboarding system that doesn't automatically report anomalies, said Ira Cohen, chief data scientist at Anodot.
In machine learning, "there is always an error," always false positives and negatives, Cohen said. Anodot, although it uses analytics and machine learning itself, is meant to minimize those errors.
"Nobody tells you this is what anomalies look like," Cohen said. So, Anodot has a database filled with anomalies users have tagged early on, and new users can continue to build out that database by uploading new tags and data. Using unsupervised learning, Anodot can then identify anomalies, as well as patterns that lead up to anomalies.
From an engineering perspective, Anodot saves Gainsight a lot of time, Alon said. For example, in the past, Gainsight might have pushed for a software release and put it into production, then a sudden memory leak might have compromised the release and set the project back. With Anodot, Gainsight gets advance notice of a potential leak, enabling engineer teams to patch it before it becomes a problem and slows down production.
"We get signals very quickly before it's too late," Alon said. Anodot "makes us more proactive," he continued.