The need for data observability, or the ability to understand, diagnose and orchestrate the health of data across IT tools, continues to grow as organizations adopt more apps and services. (Nearly 10% of businesses now have more than 200 apps to manage, according to a recent Okta study.) on a large scale survey of IT decision makers published last September, 75% of respondents said they expect to “significantly” increase their observation spend by 2022 to better plan, deploy, and run software.
Observability tools to capture and analyze data from IT tools are not new – and today they bring in a respectable amount of capital. Monte Carlo, whose platform uses machine learning to infer what data looks like and assess its impact, became a unicorn last May with $135 million in funding. rival cradle reaffirmed its unicorn status with a new round of funding — $150 million — also in May. Other observability providers with substantial support include Manta, Observe, Better Stack, Coralogix, and Unravel Data.
That may sound like a lot of competition. But it doesn’t deter Metaplane, a data observation startup founded by MIT graduate Kevin Hu (CEO), former HubSpot engineer Peter Casinelli, and ex-Appcues developer Guru Mahendran in 2020. The three co-founders originally launched Metaplane as a product for “customer success”. who analyzed a company’s data to prevent churn. After going through Y Combinator, and when the pandemic hit, Metaplane turned around, but continued to build data analytics-focused tools.
“After several of the tools we built couldn’t be used by prospects due to data quality issues, we realized we could apply many of the lessons learned from software observability to help data teams spot these issues early,” he said. Hu to australiabusinessblog.com in an email interview. “Every day, executives make decisions based on incorrect data. More confidence in data saves time and money for the executive, the data team and the company.”
Metaplane monitors data using anomaly detection models trained primarily on historical metadata. The monitors try to account for seasonality, trends and customer feedback, Hu says, to minimize warning fatigue.
“Each ‘monitor’ we apply to a customer’s data is trained on its own. Unlike most anomaly detection schemes built on Meta’s Prophet library, we have our own proprietary approach that we’ve proven to be more effective for this domain as we can observe data very regularly and make assumptions based on the type of data that is monitored.” (For context, Prophet is an open source algorithm for generating time-series statistical models, often used to predict events.)
Metaplane also tries to establish lineage of data in a data warehouse – the systems used for reporting and data analysis – and notify stakeholders of issues via their tool of choice (e.g. Slack, PagerDuty, email). Those tools allow users to mark any alert as an expected change, and Metaplane will learn over time, Hu said.
“Metaplane is the data dog for data,” he added. “Data teams at fast-growing companies use our observation platform to save engineering time and increase confidence in data by understanding when something breaks, what went wrong and how to fix it – before an executive sends them a message about a broken dashboard.”
According to Hue, Metaplane is currently used by more than 140 teams at companies, including Imperfect Foods, the aforementioned Appcues and Reforge. That shot apparently caught the attention of investors, including Khosla Ventures and Y Combinator, Flybridge Capital Partners, Vercel CEO Guillermo Rauch, and HubSpot CTO Dharmesh Shah.
Metaplane recently closed an $8.4 million funding round from those and other backers, the company announced today. Hu says it will be used to invest in platform integrations and discovery features.
“With strong traction proving the self-service model can work, we felt it was the right time to raise,” Hue said. “We plan to invest in … creating resources for data engineers to find us.”
Hu did not answer questions about earnings but said Metaplane, which has a team of 10 employees, has not yet faced any roadblocks due to the wider technical slowdown. He attributes the success in part to Metaplane’s freemium model, which allows customers to sign up for self-service and then opt for a premium subscription.
“Our price points fall below budget freezes and allow us to make deals with our competitors, who must determine the cost of their sales forces,” Hu continued.