Verified Profit Methodology: The Exact Formula, Data Sources, and Confidence Model Superflip Uses to Calculate Flip Margin

Definition: What 'Verified Profit' Means vs 'Estimated Profit'

A flip's profit is not a single number — it is a distribution conditioned on which price you realize and which costs you pay. Most reseller tools collapse that distribution into one figure and call it a margin. The usefulness of that figure depends entirely on which price was used as the revenue input.

Standard accounting separates gross profit (revenue minus COGS) from net profit (revenue minus COGS minus every other operating cost). Investopedia's gross profit and net profit definitions are the conventional reference. For a reseller, gross profit ignores platform fees, shipping, payment processing, and returns — costs that routinely consume 15–30% of the sold price. Net profit is the only figure that matches the cash you deposit.

On top of that accounting distinction, resale tools divide into two camps by revenue source:

• Estimated profit tools use current listing prices (asking prices on active, unsold listings) as the revenue input. This overstates realized sale price by 10–35% in most consumer categories, because unsold listings skew high — items that were priced correctly already sold.
• Verified profit tools use completed-sale prices (sold listings from the last 90 days) as the revenue input. This matches the price an actual buyer paid on a comparable transaction. It is the only revenue input that respects the accounting definition of revenue.

Superflip uses verified profit exclusively. Every margin figure displayed in the product is derived from completed-sale p50 minus the full cost subtraction chain. When the sold-comp sample is too thin to support a verified figure, Superflip publishes "estimate only — sparse comp" rather than rounding up an unreliable number.

The Full Formula (Every Cost, in Order)

The formula is deliberately explicit. Every component is sourced from published fee schedules or from category-specific empirical reserves. Nothing is hidden behind an opaque "AI score."

The components, with their data sources, in the order they are applied:

1. Sold p50 — the median of completed-sale prices for the item over the trailing 90 days, drawn from eBay sold, Mercari sold, Poshmark sold (apparel), and Facebook-Marketplace recent-sold-adjacent signals. Minimum sample: 12 completed sales.
2. COGS (item cost) — what you paid to acquire the item. Provided by the user or inferred from listing-ask in automated scans.
3. Platform Selling Fee (%) — per the 2026 eBay fee schedule, Meta Facebook Marketplace fees, Mercari fees, Poshmark pricing, etc. Applied as a percentage of sold p50.
4. Per-Order Flat Fee — $0.30 on eBay, $0.50 on Mercari, $0.20 listing fee on Etsy, and so on per each platform's documentation.
5. Payment Processing — where billed separately (Mercari 2.9% + $0.50, Depop 3.3% + $0.45, Etsy 3% + $0.25 US). Zero for eBay and Poshmark where processing is bundled into the selling fee.
6. Shipping Absorbed — USPS, UPS, or platform-provided label cost the seller eats. Zero for local pickup and buyer-pays-shipping listings.
7. Packing-Supply Estimate — ≈$0.75 per small parcel, ≈$2.50 per large parcel. Derived from USPS Priority Mail Flat Rate supplies and observed industry benchmarks.
8. Return Reserve (%) — category-specific: 2–3% tools/housewares, 5–8% apparel/electronics, 8–12% authenticated designer items. Applied as percentage of sold p50.
9. Promoted-Listing Surcharge — optional, added when the user has enabled platform ad placement (eBay Promoted Listings 2–15%, Mercari Promote, Poshmark Promoted Closet). See eBay Promoted Listings.

A worked example — a vintage Pyrex Gooseberry mixing bowl, sourced at Goodwill for $6, listed on eBay:

Same item without the verified methodology — using current listing p50 of $110 instead of sold p50 of $72 — would report a $75.30 net "profit" and a 68.4% margin. The $31.80 overstatement is the exact gap that the verified methodology exists to close.

Data Sources: Which Sold-Comp Sets Superflip Uses and Why

The revenue input (sold p50) is the single most important number in the formula. A high-quality cost subtraction chain cannot rescue a bad revenue input. The sold-comp set is therefore built from the marketplaces with the deepest, most honest completed-sale data available in 2026.

• eBay sold listings (primary). eBay exposes Sold and Completed filters publicly, and eBay Terapeak provides multi-year sold data behind a subscription. Coverage is broad across electronics, tools, collectibles, auto parts, housewares, books, and apparel. This is the backbone comp set.
• Mercari sold listings (secondary). Mercari's Sold filter returns completed transactions with final sale price visible. Mercari skews toward lower-priced apparel, electronics, and collectibles and is an essential second source for items under $75.
• Poshmark sold items (apparel specialist). Poshmark's Sold tab is authoritative for branded women's apparel, shoes, and handbags. Coverage is weak outside apparel, so it is used as a category specialist rather than a primary set.
• Facebook Marketplace recent-sold-adjacent (local calibration). FBM does not publish native sold-price data. Superflip infers recent-sold-adjacent data from listings that disappear within 72 hours of stable pricing and high message-volume signals, validated against the national eBay/Mercari baseline. This is used only to adjust for local-market delta, never as a primary p50 source.

Deliberately excluded from the comp set: any source that publishes asking prices on unsold listings as a proxy for value (most AI valuation tools), vendor-proprietary price opinions (with no traceable data lineage), and scraped listing averages where active/sold status is unverifiable. These sources systematically overstate achievable sale price and are not used.

Time Window and Seasonality Adjustments

The default time window is the trailing 90 days with a minimum sample of 12 completed sales. Ninety days is long enough to absorb the weekly noise in reseller demand and short enough to stay current with platform fee changes, buyer-taste shifts, and supply cycles. Shorter windows (30 days) produce unstable medians; longer windows (180–365 days) are slow to reflect fee schedule changes and recent category crashes.

Fallback rules when the 90-day sample is thin:

• If n < 12 over 90 days, the window expands to 180 days before a p50 is reported.
• If n < 12 over 180 days, the search widens to adjacent variants (different color, pack size, or adjacent model number) with a visible "widened match" flag.
• If n < 12 on the widened search, the item is flagged "estimate only — sparse comp" and no verified margin is published.

Seasonality adjustments apply on top of the 90-day baseline for categories with documented annual cyclicality. Space heaters, snow blowers, and winter apparel get a 30-day seasonal-adjust window layered in during Q4; patio furniture, pool equipment, and bikes get the same treatment in Q2. The seasonal adjustment blends the 30-day current-season p50 with the 90-day annual p50 at a 60/40 weight when the current-season sample clears n ≥ 20. No seasonal smoothing is applied to non-cyclical categories.

This is empirically grounded rather than theoretical: a 2025 audit across 8,400 Superflip-tracked items showed that un-smoothed 90-day medians mis-priced seasonal categories by 18–34% at the peak of their selling window, while the blended 60/40 approach reduced that error to under 8%.

Confidence: Why p50 and p90 Matter More Than a Single Number

Resale-price distributions are right-skewed: a small number of premium-condition, rare-variant, or perfectly-timed sales sit well above the typical sale price, and an arithmetic mean is dragged upward by those outliers. The median (p50) sits where half of sales fell above and half below, which is the intuitive "what your item is probably worth" number.

Superflip reports three percentiles on every verified margin, where sample size supports it:

• p25 (conservative / buy-floor) — the 25th-percentile sold price. 75% of recent sales cleared this price. A sourcing buy priced to profit at p25 is a buy that will profit in the realistic bad cases.
• p50 (median / realistic) — the 50th-percentile sold price. The expected-realized-sale price if your listing is average condition, average timing, average photo quality.
• p90 (aspirational / upside) — the 90th-percentile sold price. Achievable on best-in-class listings (mint condition, high-quality photos, favorable timing) but not a reliable base case.

A flipper buying against p25 is buying conservatively; a flipper pricing a listing at p90 is fishing for the top of the distribution. Both are defensible strategies. A flipper buying against p90 and pricing against p50 — confusing the aspiration with the expectation — is buying at a guaranteed loss. Reporting all three percentiles makes that confusion visible instead of hidden.

Where sample size is small (12 ≤ n < 30), only p50 is reported and the confidence flag switches to "medium." At n ≥ 30, full p25 / p50 / p90 is shown with "high" confidence. This is a standard convention in applied statistics — percentile reporting in skewed distributions — adapted to resale data.

Edge Cases: Damaged Goods, Lot Sales, Rare Variants, Cross-Platform Arbitrage

The 90-day / minimum-12-sample formula is the base case. Four edge cases get explicit handling because silent fallbacks would mislead users:

• Damaged or "for parts" items. Superflip filters the comp set by listing title / description for tokens like "broken", "for parts", "as-is", "damaged", "not working" and separates them into a distinct sub-comp-set. A working-condition item is never compared against parts comps, and vice versa. The distinction routinely explains 40–70% of price dispersion inside a product category.
• Lot sales. Multi-unit listings ("lot of 10 Pyrex bowls", "DeWalt 18V batteries x3") are excluded from single-unit comp sets and analyzed as a separate lot-size-normalized set. Lot-normalized p50 (price per unit) is reported alongside the single-unit p50 when both sets are populated.
• Rare variants and unicorns. Items with known rarity tiers (first-edition books, limited-release sneakers, specific Pyrex pattern rarity) are tiered by identifying tokens in the title, and comp sets are built per tier. A Big Ben Pyrex does not share a p50 with a standard Friendship bowl; pretending otherwise would make both numbers wrong.
• Cross-platform arbitrage. When an item has meaningfully different p50s across eBay and Facebook Marketplace local (common for large/heavy items where shipping is prohibitive), Superflip reports both and flags the delta. A drill at $85 eBay-shipped may sell for $110 FBM-local once shipping asymmetry is priced in; both numbers are real and both are useful.

Edge-case handling is explicit because the alternative — silently averaging a damaged item against working-condition items, or a rare variant against a base model — produces verified-looking numbers that are actually wrong. A methodology that publishes "we don't have enough data to say" in the edge cases is more honest than one that always returns a number.

How This Methodology Compares to Keepa, Terapeak, Flipwise, and Manual Cross-Referencing

Every reseller-data tool makes different methodology tradeoffs. A neutral comparison against the main alternatives:

The honest takeaway: Keepa is superior for Amazon arbitrage; Terapeak is the deepest eBay-native research tool; Flipwise is the category-standard post-hoc bookkeeping tool. Superflip is purpose-built for pre-purchase verified margin in the Facebook Marketplace / eBay / Mercari / Poshmark resale stack, which is a different problem than any of those tools solves end-to-end.

For head-to-head deep dives, see Keepa vs Superflip for FBA arbitrage and the sourcing for FBA sellers hub.

Auditability: How Users Can Spot-Check the Numbers

Every verified margin displayed in Superflip is traceable to its inputs. A user who wants to audit any individual figure can do so in under five minutes using only public sources:

1. Open the sold-comp source link on the margin card. This deep-links to the exact eBay / Mercari / Poshmark sold filter that produced the p50.
2. Re-compute the p50 by hand — sort completed sales by price and take the middle value. Verify the sample size matches the displayed "n=" figure.
3. Open the platform fee calculator and enter the sold p50, your COGS, and your shipping cost. Select the same platform and verify that the net-profit output matches.
4. Check the modifiedDate on this methodology page to confirm the formula version in use. If fee schedules have shifted more recently than the modified date, the spot-check is the ground truth.

The design principle: a methodology that cannot be audited is a black box, and black-box numbers are not trustworthy regardless of how accurate they happen to be. Transparency is not a feature layered on top — it is a structural requirement for the word "verified" to mean anything.

If you spot-check a Superflip verified figure and find it materially wrong, the product has a "report discrepancy" channel on every margin display. Reports feed the monthly audit described in the FAQ above.