What supports the numbers.
The beta shows category-and-condition scenarios with explicit assumptions. Here is what the evidence can support today, what remains uncertain, and how we will compare original assessments with complete pallet outcomes.
What the beta can support today
PalletMath does not have a completed user-sales dataset today. Category-and-condition scenarios are based on displayed assumptions. An external item price reference can support a selling-price estimate when available, but it does not establish the fraction of an untested pallet that works or sells.
The conservative, base and optimistic cases are scenarios under their inputs. They are not guaranteed bounds, prediction intervals or calibrated probabilities of profit. A suggested bid limit is conditional on the selected scenario, included costs and surplus target.
These live counts come from the existing recorded-outcome system. Sold-through counts alone do not measure the accuracy of the new category scenarios. No launch-ready badge or no-loss promise follows from them.
How we will measure the new scenarios
Before outcomes are known, preserve the manifest, categories, stated condition, benchmark version, input assumptions and scenario outputs. The original quantities and retail totals stay fixed, including stock that turns out missing, damaged or unsold. Later receiving corrections are recorded separately.
Evaluate cash recovery at a fixed horizon from receipt, initially 90 days. Include every enrolled pallet at that date, even if it has not sold through. Report missing follow-up separately instead of treating it as a successful sale. Display any unfinished refund period as provisional.
Cash recovery includes item receipts after refunds, buyer-paid shipping where recorded, and the relevant selling costs. Compare it with purchase price, buyer's premium, acquisition taxes, inbound freight and other included costs. Supplier refunds are separate from resale revenue. Unsold inventory is recorded as inventory, not counted as cash received.
This is the validation protocol we are working toward. The existing aggregate endpoint below is a historical record of recorded sold-through pallets; it does not yet report a complete 90-day cohort or prove the new scenarios are accurate.
What useful accuracy looks like
Report the dollar error and whether estimates tend to run high or low. Show results by category, declared condition, resale channel and supplier where the sample supports it. Compare the approach with a simple retail-percentage baseline, and test on whole pallets held out of model development.
Track how often recorded outcomes fall outside the displayed scenarios as well as the width of those scenarios. We will only describe a range as a prediction interval after testing its coverage. A future probability score would need its own calibration against the stated target and horizon.
A small privacy threshold can permit publishing an aggregate; it cannot establish precision. Five pallets with no observed losses would still leave substantial uncertainty about the next pallet. We will show sample size, completeness and limitations alongside results.
Historical records help build benchmarks
Past pallets can provide valuable receiving counts, defect rates, selling prices, fees, refunds and remaining inventory. Preserve the original manifest denominator and include unsuccessful or incomplete pallets too. A list containing only items that sold would overstate pallet recovery.
Historical entries are development evidence. An estimate produced after the outcome is known is not a prospective prediction, so historical records stay separate from future validation results.
What a sale record proves
Manual entries, a seller's own marketplace orders and supported forwarded receipt emails have different evidence sources. A valid email signature confirms the message's origin; it does not prove that the payment remained final, the item avoided a return, or every fee was included.
Receiving records, settlement statements, refunds and cancellations matter alongside sales. Confirmed records will help evaluate future estimates as the dataset develops. They are not an existing pool of PalletMath sold prices.
Your decision still needs a review
Check the category, actual testing status, quantities, retail basis and freight quote. Open packaging does not establish working condition. Unknown descriptions do not establish a category. A precise-looking dollar result does not make an uncertain assumption precise.
PalletMath D&D LLC · 502 W 7th St Ste 100, Erie, PA 16502 · hello@palletmath.com · Privacy · Terms · palletmath.com