This page used to claim we made your store $312 more per vehicle retailed. We investigated whether that number had any real backing and found none — no real closed deals exist yet anywhere in the platform to measure a PVR lift from. We'd rather tell you that than keep the number up. Full findings below.
We previously said we observed +$312/unit. That figure is retracted (see the hero above and /transparency). What follows is the attribution framework we intend to apply once a real PVR delta exists — kept here so the reasoning is visible, not as a claim that it produces a real number today.
Full methodology: louieauto.com/attribution →
The 83% first-look approval figure and the simulation engine behind it aren't guesses — they come out of a model with two inputs: 800K+ AI-generated deal simulations and 30 years of floor-operator knowledge encoded as the rules the simulations run against. The $312 PVR-lift figure that used to be described the same way has been retracted (see hero above) — it did not, in fact, trace back to real deal data the way this section implied. Below is how the simulation engine's other outputs are built.
Synthetic customer profiles (FICO, income, LTV, vehicle type, state) are sampled from CFPB enforcement data and ACS demographic distributions — not made up, drawn from federal datasets. Each profile is run through the routing engine, F&I sequencer, and desk-pencil math. The outcome (funded / declined / re-pencil / stip cycle time) is recorded in the simulation DB.
simulation.db as sim_runs / sim_scenarios.Every rule the simulator uses came off the floor: lender weight priors (which bank buys subprime today vs. last quarter), stip patterns (which deals require pay stubs first), F&I product sequencing (when to lead with VSC vs. GAP), desk-voice scripts (how to re-pencil without losing the customer). These are not LLM-generated guesses — they are operator-authored rule sets a 30-year GM wrote and reviewed.
src/lender-submission.js, stip logic in src/fi-compliance.js, desk voice in src/ai-deal-structure.js.800K+ profiles run through 30 years of encoded rules produce a distribution of outcomes. The median outcome across that distribution is the number we publish. PVR lift = (front gross delta + reserve delta + F&I attach delta) at the median. Each headline number on this page is the simulation's median output calibrated against NADA national benchmarks.
$312 PVR lift: retracted September 2026. This used to be shown as a simulation median of (front gross delta $84) + (reserve delta $147) + (F&I attach delta $81), but investigation found no real closed-deal data anywhere in the platform to validate it against — see /transparency.
83% first-look routing accuracy: simulation-modeled figure calibrated against the platform's real 13,046-row lender_outcomes table — % of deals where the first lender submission funded (no re-pencil to next tier). Calibration anchor: industry baseline ~67% (NADA 2024).
Stip turnaround ("47→9 min"): retired as a claim on this site. No production table currently carries a stip resolution-time column to substantiate it — see /transparency for the full "what we don't claim yet" list.
If you want to inspect the rules: the source files cited above are in the codebase. The simulation engine is in simulation/ with the Anthropic API key configurable per acquirer. Run the simulator on your own dealer parameters and the outputs map to your store, not ours.
We used to say "$312 comes from these three levers." That $312 figure is retracted (see above). What's still true and still worth showing: the three mechanical levers below are real, designed features of the platform, and the logic for why each should move PVR is sound. We just don't have a validated dollar figure to attach to them yet.
Every deal declined and re-submitted loses reserve on the way down the tier ladder. Louie looks at your last 847 deals, your lender weights, and the actual box each lender is buying right now. It tells you who to hit first, before you submit. Higher first-look approval rate = reserve kept.
Most F&I managers present the same menu the same way on every deal. Louie reads the deal — FICO, term, vehicle type, down payment, customer history — and tells your F&I manager what to lead with, before the customer sits down. Higher penetration on the right products = more back.
Equity alerts in your service lane catch customers in positive equity while they're already on your lot. Morning briefing flags aging inventory before it costs you $150/day floor plan. Deals that fall through the cracks stop falling.
We claimed: +$312 total PVR lift documented across 1,167+ deals. That claim is retracted as of September 2026 — a direct investigation of the platform's deals tables found zero real closed deals to document it from. The retraction and full methodology are at /transparency.
We don't claim: isolated attribution for each mechanism above. Those are the levers we believe drive the lift — mechanically sound, directionally consistent — but we haven't run a controlled study separating each one. If you need sub-component attribution for diligence, that's an honest gap. We flag it.
If you find a number on this site you can't verify, email brian@louieauto.com. We will show you the math or we will remove the claim. That's the deal.
This calculator used to multiply your monthly volume by a $312/unit PVR-lift figure. In September 2026 we investigated whether that figure had any real backing — checked every deals table in the platform for real closed transactions with real gross data — and found none: zero real closed deals exist yet anywhere in the platform. Full findings at /transparency. Rather than keep running your volume through a number with no real backing, the calculator below now shows that state plainly instead of a dollar figure. The one number on this site that is real today is the routing-approval lift (8–17 more approvals per 100 deals, from 13,046 real routing decisions) — see /proof.
The $312 PVR-lift figure and every dollar projection derived from it (40%/65%/80% attribution cases, the table below) have been retracted, not re-estimated. We will not fabricate a replacement number. See /attribution for the full retraction and the attribution framework we intend to apply once a real number exists.
Join the Waitlist →The table that used to project monthly/annual dollar lift at 50–80% attribution has been removed — it was built entirely on the retracted $312 figure. It will return, with real numbers, once real deployment data exists.
Live demo. Your numbers. No login required. Five minutes.