Sold-comp pricing: real eBay data, not an AI guess.

Most AI eBay listing tools have the same pricing flaw: they ask a chatbot what the item is worth and call the answer a price. That works for items with a stable retail price — newer electronics with a current MSRP — and fails everywhere else. For used clothing, vintage decor, collectibles, raw cards, estate inventory, and the broad universe of one-off resale items, there is no MSRP. The only honest price signal is what an actual eBay buyer paid for a similar item in the last 90 days. FlowLister's sold-comp pricing engine is built around that idea.

How sold-comp pricing works

The pricing pipeline runs after the AI listing draft is built and after the item has been identified from photos. Once the identity is known — brand, model, size, condition, category — FlowLister constructs a query plan weighted toward the strongest buyer-facing identifiers and sends each query to one completed- sale authority: the SoldComps API.

• Exact identity first. UPC, card number, model, product line, or the clearest compact title is tried before any broader comparable-market query.
• Identity-safe broadening. If the exact search is thin, FlowLister relaxes only nonessential words while retaining category, condition, brand, model, size, department, part-number, and item-kind protections that matter.
• Row-level truth checks. A pricing row must be marked sold, include a positive sold price and end date, use the expected currency, and survive relevance and outlier filters. Catalog quotes and active asks cannot pass this gate.

The result is a relevance-filtered completed-sale set. That set is passed through identity, condition, lot, accessory, auction, and outlier checks before it can produce a recommended price, range, and confidence score. When those checks leave no credible sold evidence, FlowLister says so instead of presenting a model guess as a sold-comp price.

Why sold comps beat Buy It Now prices

The naive alternative — averaging the active Buy It Now prices on eBay — is what most non-AI eBay tools do, and it's why most non-AI eBay tools systematically overprice. Active listings measure seller hopes, not buyer behavior. Many listings on eBay never sell at their current asking price; they sit until the seller drops it. Averaging asking prices anchors your price to hopes rather than to what buyers actually pay.

Sold comps measure cleared transactions. A sold comp is, by definition, a price that found a buyer. Average across enough of them and you get a number that already accounts for the negotiation, the price drops, the auction sniping, and the market reality. For resale inventory, that's the only honest price.

Configurable pricing strategy: Quick Flip, Balanced, Premium

FlowLister recently shipped user-configurable pricing strategies — three modes that bias the final price recommendation off the raw sold-comp median:

• Quick Flip. Prices at roughly the 35th percentile of the comp set. The listing is intentionally under-market so it sells in days, not weeks. Best for high-volume thrift flippers who care more about turn than margin.
• Balanced (default). Prices near the median. This is the recommendation for most sellers — fair to the market, fair to your margin, and the listings clear in a reasonable window.
• Premium. Prices above the median (roughly the 65th percentile). The listing sits longer but captures more margin per sale. Best for sellers with patience and strong inventory.

The strategy is set per-account on the pricing preferences screen. The listing review always shows which strategy produced the number, so you can override on a per-listing basis when an item is unusual.

How FlowLister handles thin or noisy comp sets

Real-world sold-comp data is messy. A single bad outlier — a PSA-graded version of an ungraded card, a brand-new version of a used item, a wholesale lot of 20 listed as one — can drag the raw median up or down by hundreds of percent. FlowLister applies three safeguards:

• Outlier removal. Comps that fall more than two median-absolute-deviations from the centre of the comp set are excluded before the price is computed. The full comp list is still visible in the listing draft so you can verify the call.
• Confidence widening. When the comp set is small (fewer than 4 relevant comps) or the spread is wide, FlowLister widens the displayed price range and lowers the confidence score, flagging the listing for manual review rather than confidently picking a number from too little data.
• Identifier discipline. Recent backend fixes made the pricing engine refuse to use a bare Model or MPN number as a hard match anchor without a confirmed Brand, because vintage catalog numbers and generic part numbers were producing wildly wrong matches. The system is conservative about which identifiers are trustworthy enough to drive the price.

Accuracy vs competitor approaches

The honest pitch: sold data is stronger evidence than asking prices, and FlowLister shows the comps behind every price so you can verify the number yourself — and see when the data is thin. AI-only pricing tools — the ones that ask GPT what an item is worth — perform reasonably on items the model memorized (current-year electronics, top-100 sneakers) and poorly on the long tail of resale inventory. They also can't tell you they're uncertain, because the model is happy to guess.

FlowLister's engine is designed to make uncertainty visible. Every listing shows the comp count, the price range, and the confidence score, so you can see at a glance whether the number deserves trust. The same pipeline also powers Worth It — the photo-to-value sourcing tool — so the pricing evidence behind both features is identical.

Real eBay sold comps, configurable strategy, visible evidence, honest about uncertainty. List my first item on eBay and audit the comp data on a real item. See the plans.