Owner.com 
-
The problem
Owner sells to independent restaurants, where reaching the owner on the phone is the hard part. Its system for scoring leads and routing them to outbound reps had stopped running. Sales managers were building lead lists by hand in Salesforce, stacking around 30 filters, which took hours every week and still didn't produce good leads. Reps had stopped trusting the old scores: one lead scored -6,998,200 and still closed, and nobody, including the person who wrote the scoring script, could say where the point values came from.
-
What they built
A new score, built from scratch, that predicts which restaurant leads are most likely to put a rep through to a decision-maker when they call. It feeds the routing system that hands leads to the outbound team. YG had no machine learning background and built it with Claude over two weeks.
-
What changed
2× rate of reaching a decision-maker
In a test, three reps worked only the top-scored leads while the rest of the team worked their usual leads. The three reached decision-makers at twice the rate. Because the score can run every day at no cost, the team dropped its week-long call cadences and now replaces each rep's leads daily with the best-scored ones.
-
So what
Reps spend their dials on leads that reach someone who can buy, managers stop building call lists by hand, and a lead that didn't connect today is swapped for a better one tomorrow.
Under the hood
For GTM engineers
Claude Code was connected to Slack, Notion and Google Docs to read how the old routing system was meant to work, and to Snowflake over MCP to explore the data. The model is a gradient boosted tree predicting one yes/no outcome, a decision-maker connect, with data split into training, validation, calibration and test sets. For feature selection, a script reported the connect rate for every candidate variable and each was sense-checked by hand. Time zone was left out: Pacific leads connect more often only because the East Coast team calls them at better hours. On held-out data the top tenth of leads connected at about 2.2 to 2.3 times the average rate. The score runs daily.
More examples
Other systems GTM engineers have built, or see all of them.
-
Octave Turned the cold sales emails his head of sales receives into an automated campaign where the sender's own pitch becomes the demo
Nalin Krishnan · GTM Engineer -
Cognition Anyone on the marketing team can ask in Slack for a campaign, an audience or an event review, and an AI agent runs it through tested code
Will Hunter · GTM Engineer -
Modal Forecasts each customer's usage and alerts the team when a customer drifts from it, where closed-won and churn don't have clean definitions
Chris Prinz · GTM Engineer
Hiring someone who can build like YG?
We search for GTM engineers who have shipped systems like this one, and Joe speaks to every candidate before you do.