Case Study: Goldbelly builds a trustworthy experimentation culture with Eppo

Goldbelly transforms experimentation culture with the transparency and rigor of Eppo

Goldbelly, an online delivery platform for specialty foods, faced a significant challenge in its experimentation program. Their previous A/B testing service provided questionable results that did not match internal calculations, causing a trust deficit with leadership and forcing the ML team to spend hours manually verifying every experiment. This process undermined confidence and slowed down the development of their machine learning systems.

To solve this, Goldbelly implemented Eppo, a warehouse-native experimentation platform. Eppo connected directly to Snowflake to provide transparent, trustworthy results, eliminating discrepancies and saving the team hours of manual recalculation. This renewed confidence in the data allowed stakeholders to trust the results without question. The ML team now focuses on shipping models faster, and they plan to use Eppo's CUPED feature to further reduce experiment runtimes by an estimated 10-20%.

Goldbelly

Mun Kim

ML Engineering Manager