SigOpt, a Y Combinator grad, has raised a $2M seed round from Andreessen Horowitz and Data Collective to provide world-class data science and optimization as a service. SigOpt’s promise is to optimize everything and anything — from digital ads, to shaving cream, to synthetic biotech products — cutting the need for manual (and costly) trial and error.
SigOpt was founded by Scott Clark, CEO, while studying applied mathematics at Cornell. Clark realized that the majority of his time spent on experiments was used for very simple but time-consuming tweaks. It occurred to him that subtle “dial-turning” wasn’t a particularly efficient use of a PhD’s time.
Clark used and refined an early version of SigOpt while working with Yelp’s ad-targeting team on the Metric Optimization Engine (MOE), an open-source machine learning tool designed to target advertisements based on a user’s location. SigOpt is essentially a commercial package for MOE. You can compare Clark’s vision for SigOpt and MOE with how Cloudera distributes Hadoop and provides practical services around the technology.
If this sounds like A/B testing on steroids, it kind of is. But it’s not just about testing predetermined, user-loaded scenarios. Instead, SigOpt examines your data and recommends what scenarios to even build next before continually building and testing new ones.
SigOpt is already in use in a number of ways:
- A/B Testing: SigOpt allows for users to import tests from Optimizely and determines which tests are promising to build and run next.
- Consumer Packaged Goods: Afterglow, a hair care development company, uses SigOpt to optimize new formulas for its products more quickly.
- Machine Learning: OptimoRoute uses SigOpt to suggest new field service delivery variations and improve its route-planning services, saving hundreds of hours of engineering time.
- Biotech: Pembient, a company driving biotech innovation to fabricate wildlife products, such as rhino horn and elephant ivory, is using SigOpt to more efficiently develop product formulas. Interestingly, illegal wildlife is a $20 billion black market — the fourth largest after drugs, arms, and human trafficking. In this case, the implications of helping create comparable synthetic products at scale in legal trade is massive.
While an avalanche of data has created opportunities for companies in every industry to continually improve through testing and optimization, the whole process has become a bit of a resource hog. Highly skilled and highly paid data scientists wind up spending most of their time tweaking processes for sometimes marginal gains. SigOpt potentially cuts the manual part of trial and error significantly.