Portfolio demo3,000-appraisal portfolio walkthrough · seed 4283. Every chart on this page is computed from a deterministic dataset committed at/datasets/r83_secondary_market_demo.json. Production analytics are generated from your submitted review package.
← Back to Home

Secondary market — portfolio review

Run DeepV across a 3,000-appraisal portfolio

Pre-delivery QC at portfolio scale. Heatmaps, defect-category roll-ups, repurchase-risk segmentation, and per-loan tiles all driven by the same deterministic Prime Brain runtime that powers AR review. Containerised services scale concurrency horizontally so a portfolio sweep is just per-review pricing multiplied by file count.

Portfolio KPIs

Demo

Portfolio size

3,000

appraisals

Demo

Findings emitted

8,612

across 21 sections

Demo

Black-ball hard fails

354

critical files

Demo

Repurchase-mapped

1,418

findings

RISC distribution

Demo dataset

Materiality (value-impacting 1.0 · compliance-only 0.6 · cosmetic 0.3) · severity 0.90 / 0.70 / 0.40 / 0.15.

Low
1,140 · 38%
Mild
660 · 22%
Moderate
540 · 18%
Elevated
360 · 12%
Critical
210 · 7%
Unknown
90 · 3%

12-month findings + black-ball trend

Demo dataset

Blue area = total findings emitted that month. Red area = black-ball hard fails (subset).

Interactive Portfolio Visualization

Portfolio Health Heatmap

Interactive visualization of your entire loan pool with color-coded risk levels. Click any loan to drill down into detailed appraisal analysis.

1,000-file walkthrough drawn from the deterministic 3,000-appraisal portfolio demo

Low Risk

459 files

Mild Risk

277 files

Moderate

152 files

High Risk

98 files

Critical

14 files

Defect-category roll-up

Demo dataset

Findings per defect family. The right-hand bar shows what fraction of each category is value-impacting (versus compliance-only or cosmetic).

  • Subject Property940 findings · 72% value-impacting
  • Comparable Selection1,720 findings · 81% value-impacting
  • Adjustment Defects1,380 findings · 66% value-impacting
  • Market / Time Adjustment920 findings · 58% value-impacting
  • Photo / Condition1,240 findings · 41% value-impacting
  • Compliance / Form1,180 findings · 18% value-impacting
  • Title / Parcel540 findings · 74% value-impacting
  • Zoning / Use380 findings · 61% value-impacting
  • Other312 findings · 32% value-impacting

Blue bar = total findings (relative to category max). Orange bar = value-impacting share within the category.

Repurchase-risk segmentation

Demo dataset

Findings carrying a structured repurchase-defense mapping, grouped by category.

  • origination qc408 · 29%
  • secondary qc552 · 39%
  • investor review218 · 15%
  • post close audit142 · 10%
  • rep warrant defense98 · 7%

Pay-per-review portfolio calculator

Per-review pricing scales linearly with volume. The DeepV per-review price is shared on the contact us page once we have scoped your pilot — until then this calculator uses whatever value you enter below.

Manual-review labour assumption: 45 minutes per file. Substitute your own internal cost when scoping a pilot.

Manual review cost (estimate)

$255,000

DeepV pay-per-review cost

$0

Cost delta (manual − DeepV)

$255,000

Actual ROI depends on your portfolio scope, file mix, and review process — run a real pilot to see your numbers.

DeepV estimated cycle time

≈ 6.3h

Cycle-time estimate assumes containerised concurrency; actual time depends on payload size and review depth. Compared to current 8 days.

How DeepV processes a 3,000-file portfolio

DeepV's brain fan-out runs in containerised services. Concurrent reviews scale horizontally with autoscaling. Throughput depends on payload size, brain set, and review depth; DeepV does not publish a fixed reviews-per-second figure.

  • Cloud Run-style containerised services for the brain fan-out
  • Prime Brain orchestration is a single review run, replicated across containers per review
  • Per-review pricing means cost scales linearly with portfolio size
  • Containerised secrets via Secret Manager (no .env in repo)

Run DeepV across your own portfolio

Pay-per-review pilot on a real slice of your portfolio. We share the full RISC distribution, defect catalogue, and repurchase-risk segmentation — built on your data, not simulations.