Goji Berry

Before writing a single line of code, Roman sent ~20,000 cold emails to businesses across the USA and multiple industries offering "high-intent leads" for $200–$300/month, with a money-back guarantee. He got on calls, closed customers, then manually delivered leads by having people search LinkedIn for relevant profiles (e.g., searching "dental clinic nurse" to find commenters), extracting contact info, and emailing the lead lists. He and his co-founder did this manually, then hired Fiverr workers with a detailed blueprint to scale fulfillment. This generated $7,000 MRR before any product was built. Once too many customers made manual delivery unmanageable, they paused sales, rebuilt everything as an automated platform, and re-launched as Goji Berry V3. The prior two versions (a UGC AI tool that hit $500 MRR and a Fireflies-style meeting notetaker) both failed to gain traction and were shut down quickly. The initial customer base came almost entirely from cold outreach — Roman targeted the whole world, focused on businesses that would benefit from better leads, and leveraged his prior experience running outreach for Coco AI to identify high-intent prospects (e.g., people commenting on LinkedIn posts about WhatsApp marketing or email marketing).

Prior exit from Coco AI (a Shopify WhatsApp marketing app grown to $500K ARR and sold 7 figures) gave Roman capital to invest in paid channels, influencers, and tooling from day one. He also had 5+ years of hands-on cold outreach experience and a co-founder with an established affiliate/course audience totaling ~$2.5M in personal revenue shown on TrustMRR, lending social proof. He was also the core user of his own product.

cold_email_and_linkedin_outreach

Goji Berry V1 (AI UGC tool) hit only $500 MRR with no conviction and was shut down. Goji Berry V2 (meeting notetaker, Fireflies clone) was built in 2 weeks but was "really really hard to sell" and abandoned. Paid influencers who turned out to be fake lost some money. Newsletter sponsorships were "okay" but not incredible — only profitable over 2–3 months and hard to attribute. Acquire.com and Flippa were tried for the Coco AI exit but offered only 1/4 to 1/3 of eventual sale price due to lack of 1-year revenue history.