Case study · Irth

Bennett built the risk model behind Irth's North American recurring revenue product

Bennett built and operates the underground-utility damage-prevention model that Irth sells at scale across North America. The model ranks risk tightly enough to put roughly half of all incidents in the riskiest 8% of tickets, with 83% more damage risk surfaced vs. manual field review, and $9M+ in savings on track for a single deployment.

Bennett built and operates the risk model inside Irth Insights, scoring 811 tickets before field work begins. The model became a direct recurring revenue line for Irth. Customers see the business case quickly: fewer damages, lower operating expense, & measurable payback.

How Bennett's model turns risk into a decision

Tickets and damage potentialIncidents captured vs. tickets inspected
0%25%50%75%100%0%25%50%75%100% 8% of tickets · 50% of incidents Safest quarter · clears automatically No model · inspect at random % of tickets · highest risk to lowest % of incidents captured

Source · Irth Insights · 2026 · Curve shape illustrative between the two stated points; replace with the measured curve

Every 811 excavation ticket is scored and ranked from highest risk to lowest, and damage is rare, in only 1 in 1,000 to 2,000 tickets.

The curve answers one question: if a team inspects only the riskiest slice of tickets, what share of incidents does that slice contain?

An extreme imbalance: finding the ~0.05–0.1% of tickets that go wrong is a problem where naive models fail by default. The lift below is the result.

Left end: half of all incidents concentrate in the riskiest 8% of tickets. A lean field team focuses there first.

Right end: the safest quarter of tickets carries almost none of the incidents, so it can clear automatically with no site visit.

The diagonal is the baseline with no model: inspecting 10 percent of tickets at random covers 10 percent of incidents. Each operator picks its operating point by weighing inspection cost against the cost of a strike.

Three deployments of Bennett's model

Next case studyBennett built and operates the D2C analytics platform for seven brands at a Fortune 250 consumer-goods company
Talk to an engineer