How Card Scanner Apps Get Prices (and How to Build One on Real Sales)

Every card scanner app does the same two things: figure out what card it's looking at, then put a price on it. The first part gets the attention, but users judge the app on the second. Show a price that's 40% too high and they'll stop trusting every scan.

#The pipeline

1. Identify. Photo → card name, set, number, language. If it's slabbed, read the grader and grade from the label.
2. Build the query. One search per card version: raw, PSA 9, PSA 10, and so on.
3. Price. Recent sold listings → drop lots and outliers → median.
4. Show it. The median, the middle range, how many sales it's based on, and a few recent sales the user can check.
5. Cache. Most scans are of the same few thousand cards. Don't re-fetch the same card every time.

#Where scanner prices go wrong

Asking prices instead of sold prices. Active listings are what sellers hope to get. Many never sell. A price built on them skews high, often by a lot. Builders who started on active-listing APIs found the same: "The browse API only showing list comps."

Raw and graded mixed together. Searching the card name returns raw copies, every grade, and lots. One average of all that is wrong for every version. Search each version separately. How.

Stale prices. A guide updated monthly misses fast moves. Price from the last 90 days, and show the recent trend for hot cards.

No sample size. "$412" from 2 sales and "$412" from 40 sales aren't the same confidence. Show the count.

#Step 2 in code: one query per version

const versions = {
  raw:    `${name} ${number} -psa -bgs -cgc -sgc -graded`,
  psa9:   `${name} ${number} "psa 9"`,
  psa10:  `${name} ${number} "psa 10"`,
};

async function priceCard(version) {
  const url = new URL("https://api.soldcomps.io/v1/sold");
  url.searchParams.set("q", versions[version]);
  const res = await fetch(url, { headers: { Authorization: `Bearer ${process.env.SALEHISTORY_KEY}` } });
  if (!res.ok) throw new Error(`${res.status}`);          // never show 0 for a failed lookup
  const { count, stats, items } = await res.json();
  return { median: stats.median, low: stats.p25, high: stats.p75, sales: count - stats.excluded, recent: items.slice(0, 5) };
}

When the user scans a raw card, call priceCard("raw"). When the label reads PSA 9, call priceCard("psa9").

#Step 5: don't price the same card twice an hour

Scans cluster on popular cards. Cache each card version's price for a day, and refresh popular cards in the background. With a catalog of thousands of cards, this is the difference between a few thousand lookups a day and tens of thousands. The math is in pricing thousands of items daily.

#What users want to see

Scanner users are skeptical, especially experienced collectors. Show your work:

• The median and the range, not a single magic number
• How many sales it's based on
• The last few sales with dates and links
• "No recent sales" when there aren't any, instead of a made-up number

That transparency is what separates a scanner people trust from one they double-check every time.