# How a model year is holding value, month by month

Canonical URL: https://ai.marketcheck.com/reports/value-retention-curve

> Value-Retention Curve: How one model-year cohort held its asking price over 12 months, with the odometer beside it and the whole market for comparison.

**Price:** $0.99 per cohort, with a $4.99 minimum per run (up to 200 cohorts); a larger run is quoted before it runs. Runs inside Claude or ChatGPT (MCP), or as a shareable hosted web report. Pay per report, no subscription needed.

The value-retention curve for one model-year cohort (year, make and model, optionally trim): twelve months of sold-market data, month by month, with the units that left dealer websites, the median asking price at exit, the median odometer and the median days to sell, pinned to ONE model year so the curve moves only when values move, never when the mix of vintages moves. States the retention (the last month's median as a percent of the first), the odometer drift in miles, the low point of the window and the units, and draws the whole used market (every vehicle of the same inventory type, nationally) as a comparison line rebased to the same start. With a powertrain (e.g. BEV) the cohort is filtered to it and a combustion comparator is added: the same nameplate on combustion when it has sales, else every vehicle of the same model year on combustion, stated on the report. National by default; a US state narrows the cohort. From the Sold Summary API: a sale is inferred when a listing leaves a dealer website, every price is the advertised asking price at exit (never a negotiated transaction price), and the cars in a cohort age and add miles across the window, so it is a cohort trajectory, not same-car depreciation and not a residual forecast. PORTFOLIO RETENTION SHEET: paste cohorts instead (one per line: year make model, optional trim) and every cohort gets its own row (first and last median, retention, odometer drift, units) in one table sorted lowest retention first, with a CSV of every row and the whole-market retention beside it. Use when a lender, insurer or analyst asks how a specific model year is holding value, what residual or advance-rate cushion a cohort needs, whether an EV cohort is eroding faster than combustion, or which cohorts in a book are losing value fastest. Do NOT use to value one VIN (use Vehicle Valuation Report), to pick where to remarket units (use Remarketing Liquidity Map), or for a seasonal buy and sell timing question (use Seasonal Buy-Sell Calendar). The sold data has state grain, not ZIP or radius. PRICING: $4.99 for one cohort. A portfolio sheet is metered at $0.99 per cohort with a curve, with a $4.99 minimum, up to 200 cohorts per run; lines that cannot be read or that have no sold data are not billed. A sheet that could cost more than $4.99 requires cost confirmation: call first WITHOUT confirm_cost_usd for an up-to quote (this charges nothing), then again with confirm_cost_usd set to the quoted cap once the user approves.

## Inputs

### Recommended
- **year** (e.g. `2022`): Model year of the cohort (the 2022 in "2022 Toyota Camry"). Needed for one cohort; leave blank when you paste cohorts. A curve is pinned to one model year.
- **make** (e.g. `Toyota`): Make (e.g. Toyota). Needed for one cohort.
- **model** (e.g. `Camry`): Model (e.g. Camry). Needed for one cohort.

### Optional
- **trim** (e.g. `XSE`): Trim (e.g. XSE) to pin the cohort to one trim. Leave blank for every trim of the model.
- **state** (e.g. `TX`): US state, as a 2-letter code (e.g. TX), to narrow the cohort to one state. Leave blank for the national curve. The comparison line is always the whole US market.
- **inventory_type** (e.g. `used`): Which market: used (the default), new, or certified pre-owned.
- **months** (e.g. `12`): Months in the curve, 2 to 12, ending at the latest data month. 12 by default.
- **powertrain_type** (e.g. `BEV`): Filter the cohort to one powertrain (BEV, PHEV, HEV, MHEV or Combustion) and pair it against combustion: the same nameplate on combustion when it has sales, else every vehicle of the same model year on combustion.
- **cohorts** (e.g. `2022 Toyota Camry XSE`): A portfolio sheet instead of one cohort: one cohort per line as "year make model" with an optional trim (e.g. "2022 Toyota Camry XSE"). Up to 200. State, inventory type and months apply to every line. Metered at $0.99 per cohort with a curve, with a $4.99 minimum.
- **confirm_cost_usd** (e.g. `39.6`): Only for a sheet that could cost more than the $4.99 floor. After quoting the user the up-to cap and getting approval, call again with this set to the quoted cap. Charged only for cohorts with a curve, never above the cap. Leave empty on the first (quote) call.

## Questions

**Where does the sold data come from?** The Sold Summary API: a sale is inferred when a listing leaves a US dealer website, and each price is the advertised asking price at that moment, never a negotiated transaction price. Auction and private-party sales are not in it. The data runs through the latest complete month.

**Why one model year?** A curve that blends vintages moves when the mix moves, not only when values move. This curve is pinned to one model year (and optionally one trim), so the 2022 cohort is tracked as the 2022 cohort. It is a cohort trajectory: the cars age and add miles across the window, so the odometer is shown beside the price.

**How is retention computed?** The last month's median asking price at exit as a percent of the first month's, both at whole dollars, counting only months with at least 5 priced sales. Under 30 priced sales at either end the curve is marked thin. The whole used market (every vehicle of the same inventory type, nationally) is drawn beside it, rebased to 100 at the same start.

**Can I compare an EV cohort with combustion?** Yes. Pass a powertrain (BEV, PHEV, HEV, MHEV or Combustion) and the cohort is filtered to it, with a combustion comparator beside it: the same nameplate on combustion when it has sales, otherwise every vehicle of the same model year on combustion, and the report says which.

**Can I run a whole book?** Paste cohorts one per line (year make model, optional trim) and each gets its own row with the first and last median, retention, odometer drift and units, sorted lowest retention first, with a CSV. $0.99 per cohort with a curve, $4.99 minimum, up to 200 cohorts; lines that cannot be read or have no sold data are not billed.

**What does it cost?** $4.99 for one cohort and $0.99 per cohort in a portfolio sheet. Connect the MarketCheck AI Desk to Claude or ChatGPT once and ask in plain English, or run it from this website. You pay per report from a credit balance; no subscription is needed.

## Data semantics

Every price is a **listed (asking) price**, not a transacted sale price. "Recently off-market" comps show a vehicle's last list price (a sold proxy). The ML "MarketCheck Price" predicts the off-market exit price: what comparable cars were last listed at right before they left the market. It is a listing price, not a sale price.

## How to run

Connect MarketCheck to Claude or ChatGPT at https://ai.marketcheck.com/install, then ask in plain English. Or open the report page and run it.

[Open the Value-Retention Curve](https://ai.marketcheck.com/reports/value-retention-curve) · [All reports](https://ai.marketcheck.com/) · [llms.txt](https://ai.marketcheck.com/llms.txt)
