Valuation lab
The live cascade. Four sources produce an internal estimate each; they are blended geometrically, modulated by swap demand, then quantised to a band. Only the band and the confidence ever reach a user — the amounts below are visible because this is a sandbox you are driving yourself.
Inputs
Σshare², not a trade count. A sweetener riding along in big bundles accrues hundreds of trades and almost no evidence.
What a user sees
The range is the honest output: at 52% confidence, claiming a single band would be false precision. It narrows as sources agree and evidence accrues.
| Source | Estimate | Weight | Share |
|---|---|---|---|
| RRP + depreciation RRP ×0.55 decay over 4.5y ×0.82 condition | £384 | 0.40 | 39% |
| Completed swaps Revealed preference, effective n=6.0 from completed swaps | £320 | 0.63 | 61% |
| Demand modulation 4 wants vs 2 available → +15% | ×1.15 | — | — |
| Blended weighted geometric mean × demand | £396 | — | Band 7 |
What completed swaps teach it
Set a true value the model cannot see. It starts at the RRP-and-feed anchor and only ever observes that a swap cleared. Watch the revealed term pull the band onto the truth — and watch how much slower it is when the item is small.