Harte INFINITI

The dogfooding posture

Harte INFINITI is Rooftop's founding pilot — the Hartford, CT store where our founder Ken Criscione runs F&I. AutoCurb is being battle-tested inside an active luxury F&I desk with 25 years of operator instinct on the deal.

This isn't an arms-length pilot. AutoCurb is being built inside the same store where Ken sits the deal every day, watched by the used-car director and BDC managers and accounting staff. Every rough edge gets found by an operator with the desk's number on the line and yesterday's receivables to sign off on. Every fix ships within the week.

That posture — operator-led dogfooding before any external pilot — is the deliberate filter. The version of AutoCurb any other dealer touches has already survived a working F&I desk running it as a primary acquisition channel.

What AutoCurb is doing in the store

All three AutoCurb channels are wired into the Harte INFINITI floor:

• Off-street, on the dealer-branded subdomain, taking VIN submissions from the open web.
• Service drive, with the soft SMS opt-in firing on every appointment.
• In-store trade, used by the desk for trades-not-on-the-lot — the customer is in the showroom but the trade-in is in the driveway.

Every submission flows into the same pipeline. Manager approval is gated at the database level. Every consent is versioned and timestamped. The CFO can pull a per-VIN audit packet from any closed deal.

What they're measuring

The pilot has four metrics it's tracking, and these are the same metrics every external pilot will track:

1. Units acquired direct, per rooftop, per month. The headline. Every direct unit is one less car bought at auction with auction fees, transport, and reconditioning surprises.
2. Cost per direct unit vs. comparable auction unit. The savings number. Currently modeled at ~$2,000 per unit; the pilot will publish the actual.
3. Time-to-offer (median). From customer submission to offer in the customer's hand. Speed wins close rate.
4. Customer NPS on the offer page. Does the customer feel respected by the experience, or do they bounce? This is the leading indicator on whether the channel scales without paid acquisition spend.

Why we publish at pilot end

We've made a deliberate decision not to advertise mid-pilot numbers. Two reasons.

First, every honest measurement window matters. Thirty days of data isn't enough to control for seasonality or for the particular mix of vehicles the store happens to be buying that month. Sixty to ninety days is the floor for a number we'll defend in a sales conversation.

Second, the dealer-tech industry has a credibility problem with case-study math. We're not going to add to it. When the pilot results land here, they'll be auditable: methodology, sample size, comparison cohort, the whole packet.

If you're a 1-to-10-rooftop group considering Rooftop and you want to talk to an operator who's been running it day-in-day-out, reach out. Ken will get on a call.

Live results land here once the pilot completes its first measurement window.