Ranges, not verdicts.
Every Valumech number is built from tracked manufacturer prices and a transparent depreciation model, expressed as a range with a confidence rating — and tightened as real sales accumulate. Here is exactly how the estimate is made, where the data comes from, and what we deliberately don't do.
1 · What we measure
For every model we track the new price (verified against the manufacturer or an official sales channel), then model how it depreciates with age and condition to produce an estimated used value. A used-value estimate is shown only for models that are both publicly priced and actually buyable (shipping or pre-order) — there's no point valuing a robot you can't buy.
2 · The depreciation model
Used value is a geometric (compounding) decline from the new price:
Each model carries its own depreciation rate where we have one; otherwise it inherits a category default by tier — heavier for cheaper, fast-moving hobbyist hardware, gentler for enterprise machines with longer useful lives:
Hobbyist
48% per year
≈52% of value retained after one year
Consumer
40% per year
≈60% of value retained after one year
Research
36% per year
≈64% of value retained after one year
Enterprise
30% per year
≈70% of value retained after one year
The estimate is then adjusted for condition, and published as a range (the modeled midpoint ±10%):
13,500 × (1 − 0.40)2 × 1.0 ≈ $4,860 in good condition (range $4,374–$5,346), ≈36% retained.
3 · Confidence & the “Placeholder” label
We're honest that this is a modeled estimate, not a data-backed appraisal. The used market for humanoid robots barely exists yet, so there are almost no resale transactions to calibrate against. Every used-value figure carries a clearly-labeled Placeholder badge and a low-confidence note. As evidence accumulates — first observed new-price declines, then real resale comps — we replace the tier model per robot and drop the badge as confidence rises.
4 · Where the data comes from
New prices are tracked automatically each day and stored with full provenance (source URL, date, and the exact figure). The hard rule across the whole catalog is sourced or left blank — never guessed: a price counts only from the manufacturer's own site or an official sales channel. Reseller and aggregator quotes, subscriptions/RaaS, and aspirational future targets are not recorded as a price. Specs are verified against the primary source the same way.
5 · Known limitations (and the roadmap)
Being upfront about what the current model can't do yet:
- The ±10% range is too tight for an asset class with this little resale history — it will widen as we model real variance.
- No new→used “cliff.”The geometric curve doesn't yet capture the steep drop the moment a unit becomes “used,” which real comps will add.
- Thin markets. Some priced enterprise units have essentially no resale market, so an estimate there is more illustrative than tradeable.
The fix is data: the daily collector builds the price history, and observed declines then resale comps progressively replace the tier model — per robot, transparently.
6 · What we'll never do
- Guess a price or a spec — it's sourced or left blank.
- Publish a single hard “appraisal” that implies false precision.
- Invent reviews, ratings, or resale figures we can't stand behind.
Frequently asked
How does Valumech estimate a humanoid robot's used value?
We start from the tracked new price and apply a tier-rate geometric depreciation model — used value ≈ new price × (1 − annual depreciation rate) raised to the robot's age in years, then adjusted for condition. The result is shown as a range with a confidence rating, never a single hard appraisal.
Are the used-value estimates based on real resale data?
Not yet. The used market for humanoid robots barely exists, so the estimate is a clearly-labeled modeled placeholder based on category depreciation rates — not data-backed resale comps. As real new-price declines and resale transactions accumulate, we replace the model per robot and drop the 'Placeholder' label.
Why a range instead of a single number?
A single figure implies a precision we don't have. Every estimate is a range (currently the modeled midpoint ±10%, widened by condition) so it communicates uncertainty honestly.
Where do the new prices come from?
Manufacturer pages and official sales channels, tracked automatically each day and stored with full provenance. Every price is sourced or left blank — we never guess. Reseller/aggregator quotes, subscriptions, and aspirational targets don't count as a price.
Want the curves, the index, and the provenance behind every figure as data? That's the licensing offering. · Browse the directory → · Anatomy of a robot →