What the score estimates, the data behind it, how confidence and staleness are shown — and what it cannot verify or guarantee.
Every score is a calculation, not a measurement. It combines what a part recently sold for, what the yard charges for it, what the marketplace takes, and how long the pull takes — and ranks the result so you know where to point your Saturday first. If any input is missing or weak, the score says so instead of guessing quietly.
Part view ranks individual pulls by modeled net per estimated pull minute. Vehicle view ranks donor cars by total extractable opportunity. Both views draw from the same ledger; they answer different questions — "what should I pull?" versus "which car deserves my time?"
Profit per minute exists to order candidates, not to predict your hourly earnings. Pull minutes are heuristic estimates, not measured times. Your actual pace, tools, and the vehicle's condition change the real number. Use the rank to prioritize; do not multiply it by eight and call it a day rate.
Inventory changes constantly. A part the model valued yesterday may already be gone, and the model cannot see the current state of the vehicle.
The model never sees the part. Cracked tabs, cut wires, water damage, and wrong option codes are invisible to it and decisive in reality.
A 90-day sold median summarizes what comparable parts sold for. It does not promise your part sells at that price, or this month.
Counter totals include sales tax estimates for planning only. Nothing here is tax or legal guidance.
No score replaces eyes on the donor, a part number check, and a condition inspection before removal.
Year, make, model, and row/space location from the yard's inventory feed, where available. See the freshness table below for current source status.
A part catalog maps candidate parts to the vehicles they fit, with option and configuration rules (engine, drivetrain, body style, trim). Cross-model and unverified-option matches are flagged verify, not treated as confirmed.
90-day eBay sold listings: the median sold price and the number of sales behind it. Listings are matched to the specific part — not the model family in general.
Posted yard price lists per store, verified October 4, 2026: Pick Your Part (pyp.com) and Pull-A-Part Birmingham/Nashville (pullapart.com — both stores publish identical lists). Includes core charges and environmental fees where the yard itemizes them.
eBay final value fee 13.6% (Motors Parts & Accessories, no Store — per eBay's fee page, checked Oct 4, 2026) plus $0.40 per order, applied once per listing. Shipping is buyer-paid (calculated) for shippable parts, so no shipping deduction in the net. Packaging, cleaning, travel, and return allowances are not in the modeled net — see the calculation section.
Heuristic minutes per part based on part type and access. Estimates, not measurements — your mileage will vary.
Older lot arrivals are more likely picked over. Any availability adjustment is labeled as an adjustment, not a fact — and an old vehicle can still hold a valuable part.
| Component | Source | Last retrieved | Status | What can go wrong |
|---|---|---|---|---|
| Yard inventory — Pick Your Part (sample market) | pyp.com inventory feed | Oct 4, 2026 (787 vehicles) | Blocked — 403 from bot protection since Oct 4, 2026; automatic daily retry | Vehicles may leave, arrive, or be picked over while the feed is down |
| Yard inventory — Pull-A-Part Birmingham (5) / Nashville (6) | pullapart.com frontend data endpoints | Oct 4, 2026 (location data verified) | Available | Same staleness and pick-over risk as any lot listing |
| Market evidence | eBay 90-day sold listings, manual pulls | Rolling; 116 of 748 catalog parts anchored as of Oct 4, 2026 | Partial — remainder modeled | Sales may be sparse, old, or non-comparable; modeled values are labeled as modeled |
| Cost assumptions | Posted yard price lists + tax rates | Oct 4, 2026 | Current | Yards change prices, fees, and tax; counter beats the website |
| Score model | RowGold scoring engine | v1, Oct 4, 2026 | Active | Logic, classifications, and estimates change — see version history |
When a source is blocked or stale, results are labeled with the last successful retrieval — never presented as current.
Net = sold_median × (1 − 0.136) − $0.40 − ship − (price + env + core) × (1 + tax_rate)
Sold median: 90-day eBay sold median for the specific part. 0.136: eBay final value fee, Motors Parts & Accessories without a Store, on the total sale — per eBay's fee page, checked Oct 4, 2026. $0.40: eBay per-order fee, once per listing. Ship: shipping cost only when the seller pays it (our shippable parts use buyer-paid calculated shipping, so usually $0). Price + env + core: posted yard price, environmental/services fee, and core charge. Tax: the yard's sales tax rate on the counter outlay.
Counter prices come from each yard's posted price list (see table above). Tax rates: Pick Your Part 9% (confirmed at the counter), Pull-A-Part Birmingham 10% and Nashville 9.75% (from Pull-A-Part's location data, Oct 4, 2026). Core deposits are treated as sunk cost — a flipper has no core to return, so no recovery is ever modeled.
The modeled net does not subtract packaging materials, cleaning supplies, travel cost, tool wear, storage, or a return reserve. It is therefore optimistic at the edges by a few dollars per part. We show the lean version because it is the version every car on the lot is scored with — consistently, not completely.
So you can reconcile the number. A score you cannot recompute from displayed inputs is a claim, not a calculation. Every opportunity card links its inputs; this page defines them.
Net ÷ pull minutes, ranked top-down. Time is the binding constraint on a yard trip, so time-efficiency leads.
A vehicle scores as its best non-overlapping set of pulls — not the sum of everything on it. Harvesting a loaded door assembly consumes the glass, regulator, and shell you might otherwise count separately. The model picks the best extraction path per car and counts only what that path yields.
Parts missing sold data are marked needs analysis — never silently zero, never silently skipped. They are excluded from ranked totals until researched, and shown as needing work.
Candidates whose fitment rests on cross-model matching or unverified options carry a verify flag and are excluded from rankings and headline totals until confirmed.
| State | Meaning | Ranking behavior | Your move |
|---|---|---|---|
| Verified | Recent, relevant sold evidence plus structured fitment support | Standard ranking | Verify the physical part and its condition |
| Limited data | Some comparable evidence, but thin sample or aging data | Shown with lower confidence; kept out of headline totals | Review the comps manually |
| Inferred | Value estimated from related data, not direct sold evidence | Never presented as equivalent to verified data | Treat as a research lead |
| Needs analysis | Required price, fitment, or vehicle detail is missing | Never treated as zero; clearly flagged | Research before relying on it |
| Verify fitment | Candidate may match, but a key option or attribute is unknown | Not called compatible | Check the part number and options |
| Blocked | A known constraint contradicts the candidate | Excluded from scoring | Do not pull on this basis |
Model and year alone are often not enough. Engine, drivetrain, body style, trim, and option packages decide whether a part actually fits — a headlight can differ by halogen vs HID, adaptive vs non-adaptive, left vs right. Cross-model compatibility follows documented interchange rules, and where the rules are uncertain the candidate carries a verify flag instead of a confident score. When the model cannot responsibly score a candidate, it refuses to — that refusal is a feature.
Lot age is an imperfect proxy for what's left. A car that arrived months ago has probably been worked over — but the part you want may be exactly what nobody else wanted. Any availability adjustment is labeled as an adjustment. What the model cannot see from inventory data: condition, completeness, whether the part was already pulled, and whether the donor was optioned the way you need.
A 7.3L Powerstroke PCM on a 2000 Ford F-350:
Use scores to decide what to research first — then verify the exact part number, option codes, condition, counter price, and current sold market before pulling or listing. Never treat a modeled net as expected income. When a score and your eyes disagree, trust your eyes.
No. It is a modeled estimate from historical sold data and posted yard prices. Actual profit depends on condition, fitment, listing quality, shipping, returns, fees, taxes, and market changes.
No. Inventory feeds show what arrived, not what remains. Always verify at the yard.
90-day eBay sold listings, pulled and matched per part. 116 of 748 catalog parts are anchored to real sold medians as of Oct 4, 2026; the rest are modeled estimates, labeled as such.
Completed eBay sales from the last 90 days for the specific part. We record the median sold price and the number of sales — the count tells you how much to trust the median.
The evidence is thin or indirect. Treat it as a research lead: check the actual listings yourself before spending a Saturday on it.
A required input — price, fitment detail, or vehicle data — is missing. It is flagged visibly and never silently treated as zero or skipped.
eBay's 13.6% final value fee plus $0.40 per order; buyer-paid shipping (so no shipping deduction on shippable parts); yard price plus environmental fees plus tax at the yard's rate; core deposits as sunk cost. Packaging, cleaning, travel, and return reserves are not in the net.
New sold data, yard price changes, fee changes, corrected fitment, or a model update. Scores are snapshots, not constants.
No. Fitment follows documented rules with verify flags where options are uncertain. Confirm part numbers and options yourself.
As current as the last successful retrieval, which is always labeled. Pick Your Part's feed has been blocked since Oct 4, 2026 — results there rest on the last good snapshot until access is restored.
Last reviewed: October 4, 2026. This page is updated when the scoring model, data sources, fees, or yard pricing change.
RowGold ranks every car on the lot by modeled profit per minute — with every input shown. Join the early-access list and get the yard dashboard while it's in beta.
Join the early-access list