Goava: ranking the next company a seller should call.

The situation

Goava helps B2B teams find and prioritize companies from CRM data plus an external graph. Sellers need a short, trusted list of who to call this week.

The hard part

A ranking that is "generally right" still fails if the seller does not understand why a company surfaced. Trust dies faster than precision. The system also had to stay fast enough that a morning list felt instant.

What I worked on

• Backend services that turned company and CRM signals into a usable queue.
• Recommendation / ranking paths a seller could act on the same day.
• The unglamorous work: APIs, data shape, and making the next engineer's change cheap.

The decision that mattered

Placeholder: optimize for a list a seller will work. That usually means fewer features in the ranker, a reason attached to each row, and a hard latency budget.

Outcome

Replace this with one public number or a qualitative result you are allowed to share. Example shape: "Sellers started the day from the ranked list instead of a raw CRM export." If you cannot name a metric, name the behavior that changed.