Research eBay Sold Prices Without Mixing Bad Comparables

You want a sensible resale price for a console. Search results include the console, a controller, an empty box, a damaged unit, and a bundle with games. Averaging them together produces a number, but it answers the wrong question.

Sold-listing research becomes useful when you define a comparable item before calculating a price summary.

My eBay Sold Listings Scraper collects listing observations with source provenance and coverage information. Use those records as evidence to review, with careful attention to displayed-price limits.

Define the comparable before searching

Write down model, storage or size, condition, included accessories, marketplace, and the observation window you need.

Keep parts-only units, accessories, and bundles in separate groups. If the exact model is unclear, place the record in a review queue rather than letting it enter the main comparison.

For example, "Nintendo Switch OLED, used, console and standard accessories" is a much clearer cohort than "Switch."

Run a bounded request

This sample input keeps the first run small and sets an explicit upstream budget:

{
  "keywords": ["nintendo switch oled"],
  "ebaySites": ["ebay.com"],
  "daysToScrape": 30,
  "maxResultsPerKeyword": 5,
  "maxResults": 5,
  "retrievalMode": "hybrid",
  "upstreamBudgetUsd": 0.05,
  "maxConcurrency": 1
}

The current actor documentation discloses that hybrid mode can call Caffein's sold-listings actor through your Apify account when native retrieval fails. Upstream charges are additional and have a separate budget control. Check both actors' current pricing before scaling.

The US marketplace is tested; UK coverage is beta. Keep marketplace status visible in your research notes.

Distinguish displayed price from transaction value

A sold-listing observation does not independently verify what a buyer finally paid. Accepted offers, shipping, taxes, fees, returns, and source omissions can change the economic meaning of the displayed amount.

Use the term "displayed price" when that is what the source supplies. Do not rename it "net proceeds" or "verified sale value" in your spreadsheet.

Retain item identity, title, source date, marketplace, currency, displayed price, available shipping information, condition, and provenance. If a field is unavailable, leave it unknown.

Clean the sample in explicit stages

Keep an exclusion reason for every removed row. You should be able to explain why a strikingly cheap listing did not belong in the main group.

Avoid quietly deleting inconvenient observations. A record can be genuine and still be a poor comparable for your question.

Use a median only after defining the cohort

For illustration, suppose three sufficiently comparable displayed prices are 180, 195, and 210 in the same currency. Their median is 195. Adding a 25-unit empty box would distort the sample because it describes a different product.

Report the comparable count, observed range, and exclusion criteria beside any summary. A tiny sample should stay a tentative reference point. It cannot establish a guaranteed resale price or likely profit.

Keep your own fees, shipping costs, item condition, and business assumptions in a separate worksheet. The listing export does not calculate those for you.

Read the run report before concluding there were no sales

Inspect RUN_SUMMARY, per-query coverage, and upstream diagnostics. An empty response caused by a retrieval failure is different from observed empty search evidence.

Similarly, a completed processing scope does not establish exhaustive eBay history. Budget or result caps can leave useful partial data, provided you preserve the stopping reason.

Start with a small sold-listings export, label every row's comparability, and review the source records. The first useful result is a clean cohort you can defend.