We observed real improvements across three dealer metrics. This paper explains exactly how we measured them, what market factors we controlled for, and why we conservatively attribute 50–80% to LouieAuto.
What this paper is: A full, public accounting of how we derived our attribution estimates — including the data we used, the factors we considered, and the range of outcomes if our assumptions are wrong. What this paper is not: An audited financial statement. The underlying DMS data and ActivityLog event tables are available to verified acquirers under NDA. This methodology document is public because the assumptions should be scrutinizable by anyone.
Three dealer outcomes were tracked before and after LouieAuto deployment across franchise dealerships operated by the founder. All data is from the operator's DMS P&L exports and the LouieAuto ActivityLog event tables. The measurement period is 18 months: a 12-month pre-deployment baseline and a 6–18 month post-deployment window.
Per Vehicle Retailed. Source: DMS P&L rollup, per-deal gross report. Units: all retail deals through AI routing period.
Deals funded on first lender submission. Source: DMS funding log + LouieAuto lender routing table. 1,167+ deals sampled.
Time from lead received to first human or AI contact. Source: BDC contact log + LouieAuto 60-Second Lead Responder event table.
These are the raw before/after numbers from a single operator's stores. They are not adjusted for market conditions and not attributed to any cause. They are the starting point of this analysis, not the conclusion.
The question this paper answers: of the observed $312/unit PVR lift, how much was caused by LouieAuto versus other factors? The same question applies to approval rates and lead response.
Founder-data risk: All data comes from the operator's own stores. There is no external customer data to confirm these results generalize. We are not hiding this — it's the most significant limitation of this dataset. We disclose it here and in every public summary. A third-party pilot study and independent audit are in progress. See Section 8.
Even with clean before/after data, determining how much of the lift came from the software is not straightforward. A dealer who installs new tools in March 2025 and measures results in September 2025 is not running a controlled experiment. Many things change simultaneously:
Any honest analysis must separate what the software did from what the market did. This is the attribution problem, and it's why we do not claim 100% credit for the observed results.
A true controlled study would require two identical dealer groups — one with LouieAuto, one without — running simultaneously over 12+ months. We do not have that. We have a single operator who deployed LouieAuto across all stores at a similar point in time.
The consequence: we cannot fully separate the LouieAuto effect from concurrent market changes. Our methodology attempts to estimate this separation using available market data, but it is an estimate, not a measurement. We state this plainly so acquirers and diligence teams can apply their own discount.
We derived a 50–80% attribution range (65% midpoint) using three inputs applied independently to each metric.
For each metric, we asked: "How much of the observed change could be explained by market-wide forces alone, with no software change?" We used public data sources — Federal Reserve, FRED, AFSA, NADA — to estimate upper bounds on market-driven improvement.
We set a conservative floor (minimum LouieAuto credit) by assuming market factors explain as much as they plausibly could. We set an optimistic ceiling by assuming market factors explain as little as plausible.
Rather than use each metric's own high-end attribution, we applied a uniform conservative discount of approximately 10–15% to all metrics to produce the 50–80% overall range. The 65% midpoint is a weighted average, skewed toward caution.
This means we are intentionally claiming less credit than our metric-level analysis supports. We would rather a diligence team find our numbers understated than overstated.
We applied an additional discount for three unquantified risks: (1) unknown operator-specific factors in the founder's stores that may not generalize to other dealers; (2) the possibility that the lender mix in the pre-period was worse than average, making any post-period look artificially strong; and (3) the absence of a randomized control group. These risks are real and they are not fully quantifiable. We discounted for them rather than ignore them.
The deployment period coincided with meaningful macroeconomic shifts. Here is how each factor was evaluated and what credit we assigned to it.
The Federal Reserve cut interest rates three times in Fall 2024: −50bps in September, −25bps in November, and −25bps in December — ending 2024 at a 4.25–4.50% federal funds rate target, down from a peak of 5.25–5.50% in 2023.
Impact on lender approval rates: lower benchmark rates reduce the cost of lender capital and can modestly increase consumer buying power, leading to improved approval outcomes at the margin. For near-prime and subprime borrowers (the majority of LouieAuto's routing focus), rate sensitivity is lower than for prime borrowers, since subprime rates are driven more by risk appetite than cost of capital.
+2 to +4 percentage points from rate improvements. Sub-600 FICO approval is primarily driven by lender risk appetite, not rate levels.
Subtracted 3pp (midpoint) from the 15pp approval lift before computing LouieAuto attribution. This produces a 12pp software-attributed improvement, which we then discount further to arrive at 65% overall.
Used vehicle prices normalized in 2024–2025 after the pandemic-driven spike of 2021–2022. The Manheim Used Vehicle Value Index declined approximately 8–12% from its 2021 peak but stabilized in the 2024–2025 window, ending the period approximately flat year-over-year.
Impact on PVR: a flat used vehicle market during the measurement period means we cannot attribute the $312/unit PVR lift to rising wholesale values. If anything, a flat market creates a slightly more challenging PVR environment (less free appreciation), which means market conditions worked against us, not for us, on this metric. We do not take credit for a headwind but note that the neutral-to-negative market environment makes the PVR lift harder to dismiss as market-driven.
Broadly flat to slightly declining. Manheim Index: approximately neutral year-over-year in the 2024–2025 window. This is a headwind for PVR, not a tailwind.
No market-driven credit applied to PVR for used vehicle prices. A 0% market contribution here is probably conservative — a declining market would normally reduce PVR, suggesting the software impact was even larger than attributed.
Industry-wide F&I penetration has trended up over the past 3 years as dealers have improved their product menus and training programs. NADA data shows average F&I income per unit rose approximately $80–$120 from 2022 to 2025 across franchised dealers, reflecting improved menu selling and VSC penetration gains that preceded AI tools industry-wide.
This is a real market-driven component of our PVR improvement. We estimate $50–$90 of the $312/unit lift could be attributed to industry-wide F&I gains that would have occurred without LouieAuto.
The founder-operator made intentional changes to BDC staffing and response protocols during the measurement period. While LouieAuto's 60-Second Lead Responder drove most of the 41-minute to 52-second improvement through automation, behavioral coaching and cleaner lead routing contributed. We attribute 5% of the lead response lift to non-software factors (staff changes, process discipline) to avoid overclaiming.
Auto retail follows predictable seasonal patterns: Q1 is slower, Q2 and Q4 are stronger. The before-period and after-period were aligned to similar calendar windows to the extent possible given the 18-month dataset. No material seasonal bias was identified, but with only one before/after cycle, we cannot fully rule it out.
What happens to the dealer value case if our attribution assumption is wrong? The table below shows outcomes at four attribution levels for a 100-unit/month rooftop.
| Attribution Scenario | PVR Lift (per unit) |
Approval Rate Lift |
Lead Response Lift |
Annual Value (per rooftop) |
|---|---|---|---|---|
| Conservative (40%) Most of the gain is market / operator — LouieAuto is a minor contributor |
$125/unit | +6pp | −39 min | $120,000–$150,000 |
| Midpoint (65%) — Our Published Estimate What we use in all public claims; includes a conservative buffer beyond our metric analysis |
$203/unit | +10pp | −40 min | $180,000–$250,000 |
| Metric-Supported (75%) What the per-metric analysis actually produces before the cross-metric conservative discount |
$234/unit | +11pp | −40 min | $215,000–$290,000 |
| Optimistic (90%) All market factors ignored; near-total credit to LouieAuto |
$281/unit | +13.5pp | −40 min | $260,000–$380,000 |
| Rooftop assumption: 100 retail units/month. Annual value calculated from PVR lift + approval rate improvement (fewer re-submissions, fewer lost deals) + incremental gross from lead response. Ranges reflect variation in unit volume and F&I product mix. We publish the 65% midpoint figure only. | ||||
Even at the most conservative 40% scenario, the dealer ROI is $120K–$150K/year against a $9,995 perpetual price — a 12–15× value-to-price ratio. The conservative case is still a compelling dealer purchase. Attribution uncertainty affects the magnitude of the uplift claim but does not change the fundamental value proposition.
For an acquirer evaluating strategic value, the relevant question is not "is the uplift exactly $312/unit or $203/unit?" but rather "does LouieAuto meaningfully improve dealer economics at a price dealers will pay?" The sensitivity analysis suggests yes, across all scenarios.
How does LouieAuto's attributed improvement compare to documented industry baselines?
| Metric | Industry Baseline | LouieAuto Observed | Source |
|---|---|---|---|
| Average dealer PVR (franchise) | $2,200–$2,800 | $2,635 (pre) → $2,947 (post) | NADA 2024 Dealership Financial Profile |
| Average lead response time | 15–45 min (industry average: ~20 min) | 41 min (pre) → 52 sec (post) | Gubagoo / CarGurus lead response benchmarks (2023) |
| Lender first-look approval, non-prime | 60–70% (AFSA industry benchmark) | 68% (pre) → 83% (post) | AFSA State of Auto Finance 2024; DealerSocket subprime approval data |
| Sub-600 FICO approval rate | 40–55% (varies by lender mix) | Improved proportionally with first-look rate | CFPB Auto Lending Data (2024); lender-specific NMLS data |
| Average F&I income / unit (franchise) | $1,650–$2,100 | Material improvement tracked; backend figures in data room | NADA 2024; DealerSocket F&I benchmarking report |
The pre-deployment numbers are consistent with industry norms for a multi-rooftop franchise group — they do not represent an operator who started from an unusually low baseline. A 41-minute lead response is slightly worse than the industry average, suggesting room for improvement, but the lender approval rate of 68% sits within the typical AFSA range. This context matters: the post-deployment 83% approval rate represents an above-industry outcome, not merely a return to average.
The PVR comparison is trickier: NADA numbers blend franchised and independent dealers, high-volume and low-volume stores, and vary significantly by brand mix. We note only that the pre-deployment PVR of $2,635 is within the franchise range and the post-deployment $2,947 is at or above the upper end of that range — a meaningful shift upward.
That these results will replicate exactly at a different dealer. Operator skill, lender relationships, market geography, unit mix, and staff quality all affect outcomes.
That the 65% attribution is precise. It is a conservative estimate. The honest range is 40–90% depending on which market factors you believe were causal.
That external customers would see the same results. This is founder-operator data only. No external customer data exists yet.
That this methodology has been independently audited. It has not. An agreed-upon procedures engagement is planned. See Section 8.
The entire dataset comes from one operator who built the software and deployed it in their own stores. This creates an inherent selection bias: the operator had deep knowledge of the tool, ran it with discipline, and had strong pre-existing lender relationships that LouieAuto could optimize but not create. A dealer with weaker lender relationships or less disciplined deal desk may see smaller improvements.
We do not have a way to quantify this bias without external customer data. It is the primary reason we are running a paid pilot with an independent dealer and pursuing a third-party audit. Until those exist, this limitation stands and any prudent acquirer should apply their own discount.
The metrics in this document reflect the 18-month deployment period. Market conditions will change. Federal Reserve policy, used vehicle pricing, and lender risk appetite are all time-sensitive. An acquirer evaluating this in a materially different macro environment should apply their own market adjustment to the approval rate and PVR figures.
We are actively pursuing two external validation efforts. Neither is complete. We publish their status so there is no confusion about what exists today versus what is in progress.
We are sourcing a paid pilot dealer — an independent with 30–50 units/month who is not affiliated with the founder — at 50% discount ($5,000 pilot price vs. $9,995 standard). The pilot will run for 90 days with the same three metrics tracked (PVR, approval rate, lead response time) before and after deployment.
Results will be published publicly under the dealer's real name and location with their consent. No anonymization. No cherry-picking. If results are worse than the founder-store data, we will say so.
Status: Pilot dealer outreach began May 2026. Active conversations with 3 independent dealers in the 30–50 unit/month range. Target: pilot deployment by end of Q2 2026, 90-day results published Q4 2026.
We plan to engage an automotive-specialized accounting or advisory firm to perform an Agreed-Upon Procedures (AUP) engagement on the underlying DMS data and ActivityLog methodology. An AUP is not a full audit — it is a structured review of specific claims against defined procedures. The output is a report stating whether the firm found exceptions to the methodology.
The engagement would cover: (1) tracing ActivityLog events to DMS records, (2) validating the pre/post period definitions, (3) reviewing the market factor adjustment methodology described in this paper, and (4) issuing a finding on whether the 65% attribution assumption has a "reasonable basis" given the evidence provided.
Status: Planned. Estimated cost $40K–$60K. Timeline: 60 days from engagement start. Not yet initiated.
An acquirer who needs audited numbers before proceeding should wait for Effort 2. An acquirer who can evaluate the methodology on its own merits and apply their own discount has this document and can request the underlying data under NDA. We believe the methodology is honest and conservative, but we also believe an acquirer has every right to demand more — and we intend to provide it.
On the relationship between honesty and valuation: We believe publishing this methodology — including its limitations — increases rather than decreases the credibility of our claims. An acquirer who reviews this paper and finds the methodology reasonable has more confidence in the numbers than an acquirer who is handed a single-page "proven results" summary with no methodology behind it. We are betting on transparency.
DMS P&L exports, ActivityLog event tables, and per-deal routing records are shared with verified acquirers. No scheduler gate — email directly and receive access within 24 hours.
LOUIEAUTO ATTRIBUTION METHODOLOGY · V1.0 · MAY 2026
This document is published without NDA. It may be shared, quoted, and distributed freely. If you identify a methodological error, email brian@louieauto.com — we will correct it and note the revision. This document does not constitute a financial statement or audited claim.