Overview: Van Valuator - Camper van market price estimator
The Van Valuator calculator estimates what the market is currently willing to pay for a van based on
your specific features and configuration. Behind the interface is a multi-stage statistical and machine-learning
pipeline trained on a large historical database of secondary-market listings — DIY conversions and
professional builds — collected across multiple marketplaces. Each prediction represents a best
estimate of the fair market price for a van matching your specifications, adjusted for recency,
geography, and evolving market dynamics. The calculator also quantifies uncertainty, and accounts for price adjustment post-listing while providing
model-derived explanations of the top features influencing your price.
Model fundamentals: Price prediction
Your estimate is produced by a gradient-boosted ensemble model that learns a function based on bespoke van or vehicle input features. These models construct many shallow decision trees sequentially, each correcting the residual error of the
ensemble so far. This results in a flexible, nonlinear function capable of representing:
interactions between van attributes (e.g., mileage affecting high-roof vans differently than low-roof models),
threshold effects (such as price steps at common mileage cutoffs),
nonlinear value patterns (e.g., diminishing returns from larger electrical systems).
The displayed value is the model’s point prediction.
The range you see reflects model uncertainty, variance in historical listings of similar vans, and natural
differences in how sellers represent and maintain their builds. It also provides a conservative representation to account for expected negotiation between buyers and sellers, which may cause actual transaction prices to deviate from listing prices (which our dataset is trained on).
GBDTs model the price as:
F(x) = Σm=1..M γm hm(x)
where each hₘ(x) is a small decision tree trained to minimize the remaining prediction error.
This approach is well-suited for diverse tabular data and handles missing fields gracefully while capturing
nonlinearities without manual feature engineering - perfect for custom van build data.
To decompose your specific estimate, we compute SHAP (Shapley Additive) values—a game-theoretic method that decomposes a
prediction into additive contributions from each input. SHAP shows:
features increasing your price,
features decreasing it,
a ranked list of the strongest contributors.
Model fundamentals: General market insights
To understand broad market patterns, we also fit a regularized linear model estimating average marginal
effects of individual features:
minimizeβ (‖y − Xβ‖² + λ₁‖β‖₁ + λ₂‖β‖₂²)
Elastic Net Regression blends feature selection and stability, helping estimate directional guidance values such as
“a shower typically adds $X–$Y.”
This complements the nonlinear predictions of the main model.
Data sources, time period, and cadence
Coverage: Dataset begins in mid-2025 and includes continuously scraped listings from multiple public marketplaces.
Variables captured: year, mileage, drivetrain, roof height, layout and bed configuration, electrical
capacity, water systems, bathroom components, appliances, and over 150 structured and unstructured features.
Quality controls: removal of implausible values (e.g. $0 or >$800k); standardization of categorical
fields; unknown technical specs (battery, solar, water) are treated as missing rather than zero to avoid penalization.
Weighting: sold/pending listings are weighted more heavily, with adjustments for typical negotiation
discounts relative to listing prices.
Update cadence: new listings are ingested regularly and the model/database update on a scheduled basis to
reflect evolving market conditions.
Typical Van
The “typical van” snapshot is computed from the overall distribution of vans in our current database, independent of any individual user’s inputs. For key numeric fields (such as price, mileage, year, sleeping capacity, and seating), we report the median value across all listings. For other features like showers, toilets, 4x4, solar, inverters, and other core amenities, we show the share of listings that include each feature. Taken together, these medians and prevalence rates provide a high-level picture of what is most common in the market right now.
Top features that drive prices
The top features displayed in the “Top features that drive prices” section are chosen based on their global SHAP importance — the average absolute contribution of each feature across the entire training dataset — rather than being tailored to any single van. They highlight which attributes explain the most variation in list prices in the market as a whole.
The length of each bar reflects the feature’s relative importance, using its global SHAP importance compared with the other features shown. The arrow and accompanying text summarize the direction and approximate dollar magnitude of the effect estimated by the Elastic Net model — for example, whether a feature typically increases or reduces prices, and by roughly how much per unit or when present versus absent.
Dollar ranges such as “a shower adds $X–$Y” come from the Elastic Net regression, which estimates global average
marginal effects while controlling for correlated build variables.
Regional price differences
Regional price differences are calculated directly from recent listing data, rather than from model residuals. Over a recent time window, we compute the median listing price in each state and compare it with the overall median price across all states. This produces a simple descriptive view of which regions currently have higher or lower asking prices.
The card shows these premiums as both dollar and percentage differences relative to the overall median. Because this view is based directly on recent median listing prices (and is not model-adjusted), it is best interpreted as a snapshot of how asking prices differ across regions in the current market.
Market trend
We estimate recent market movement using the Elastic Net model’s time coefficient on listing age. Intuitively, this captures how prices tend to change per day after controlling for van features. Over a chosen window (for example, the last 90 days), we scale that per-day effect by the window length and express it as an approximate percentage change relative to the current median listing price.
The resulting percentage change is converted into a dollar estimate by applying it to the observed median listing price in the dataset, yielding an approximate “typical” dollar increase or decrease over the selected window, after controlling for van features.
The sparkline is built directly from recent listing data. For each day in the selected window, we compute the median list price of all vans created on that day and plot those daily medians as a time series. This provides a real, data-level view of how asking prices have moved over time, with smoothing coming from using daily medians rather than individual listings.
Uncertainty & interpretive notes
All estimates come with uncertainty stemming from listing variability, feature ambiguity, DIY build variation,
seller presentation, and regional supply/demand patterns.
The range shown reflects these uncertainties as well as the fact that we model listing prices, which may differ
from final sale prices. We expect that our fair value confidence ranges account for this fact.
Legal notes & disclaimers
This tool provides informational estimates only and does not constitute an appraisal, valuation, or financial
advice. Predictions are modeled from public listing data and may not represent actual transaction prices.
Conversions Consulting makes no guarantees of accuracy or completeness. Use at your own discretion.