FERMÀT is a San Francisco-based, AI-native commerce platform founded in 2021 that helps brands embed personalized shopping experiences directly inside the marketing touchpoints where consumers first discover a product: emails, paid ads, influencer posts, editorial content, and agent-driven chats.
"The SkyBright crew felt like an extension of our engineering team. They translated every marketing requirement into rock-solid pipelines, cut our bill in half, and documented everything so we can run it ourselves if we ever bring ETL in-house."— Shreyas Kumar, Co-Founder & CTO, Fermat Commerce
As a fast-scaling commerce platform, Fermat Commerce ingests click-stream and conversion data from dozens of campaigns every day. With aggressive employee growth and a 24/7 experimentation culture, its Fivetran bill had tripled in a year, eating into runway and forcing trade-offs between new features and data reliability.
Fermat's data team set three goals:
Phase | What SkyBright delivered |
---|---|
Discovery | Joint workshop to catalogue tables across Facebook, GA4, Shopify & TikTok; cost benchmarking and cut-over plan. |
Build | • Data model translation and schema parity validation across all sources. • Custom-typed GA4 connector development to capture unsampled funnel metrics. • Historical data transformation pipelines to preserve business continuity. • White-glove app onboarding process for Facebook, Google and other platforms with tailored authentication flows. • Custom connector development for specialized data requirements. |
Validate | Dual-run mode: Fivetran & SkyBright ingested in parallel; automated row-level diff checks reached 99.98% parity. |
Cut-over | One-click feature flag cut-over with pricing protection throughout transition. |
Throughout the project, SkyBright and Fermat held weekly standups and remained available on-demand via Slack to both US and Indian engineering teams.
With costs under control, Fermat now provisions three times as many experiments per sprint — feeding more purchase events into its AI conversion models and releasing funnel variants in hours, not days. Engineering hours once spent reconciling webhook spikes and API schema changes are redirected to new revenue features.
SkyBright constantly monitors Fermat's pipelines and helps onboard new customers when needed as Fermat continues to scale.