APP REVIEW: COINSNAP

PREV ARTICLE NEXT ARTICLE FULL ISSUE PREV FULL ISSUE

V28 2025 INDEX E-SYLUM ARCHIVE

The E-Sylum: Volume 28, Number 49, 2025, Article 12

Back in March 2023, CoinWeek published an article reviewing the CoinSnap app. See E-Sylum's excerpt at the link below. Andrew Crellin of Sterling & Currency in Australia published a new review of the app this week. Here's an excerpt - see the complete article online. -Editor

This article is intended to help me think through and explain why I believe Coinsnap (as well as the many other similar apps) aren't fit for anything more than having a bit of fun.

Don't get me wrong, I love the possibility that an app can be used to identify and value coins, but the reality is, the technology isn't there yet. Coinsnap has been around since late 2022, I downloaded the free version shortly after it was released and thought the technology showed real potential. I don't want a potential solution though - I want an actual solution! So after a few fails, I've never bothered with it.

If you haven't heard of Coinsnap before, let me bring you up to speed - it's described as the

"TOP 1 Coin ID App" in the Apple App Store. The developer claims: "CoinSnap is a companion for coin enthusiasts. Whether you're identifying a rare error coin, uncovering the value of a hidden gem, or organising a collection of hundreds of coins, CoinSnap makes it fast, easy, and fun to explore the fascinating world of numismatics."

Sounds fantastic in theory, but in reality, it's a little like having a year 10 geography student act as your guide on a trip into the Amazon jungle. They'll be earnest and will have a basic grasp of the knowledge required, but they will be woefully inadequate when the stakes are high.

The developer claims to have more than 22 million "users" worldwide, and independent information from Sensor Tower (a platform that provides intelligence on the market for mobile apps) estimates the app generated US$700,000 in income in the past 30 days alone. More than US$8,000,000 a year is big business, so this solution clearly resonates with people right around the world. There's clearly an itch that needs scratching here!

The developer behind the app is Next Vision Limited, they're based in China. There's very little publicly available information about them online, but Sensor Tower reports they have a suite of 24 apps that have generated about US$3 million in revenue last month alone.

A few of the other apps in their suite include such useful titles as: "Rock Identifier: Stone ID", "VinylSnap: Scan & Value Record", "Picture Mushroom: Identifier", "StarSnap: Sports Card Scanner" and "Picture Fish - Fish Identifier".

The common technology across CoinSnap, Rock Identifier, Picture Insect, NoteSnap, VinylSnap and others is an image recognition AI tool. The Next Vision platform is built around this central engine as well as a library of millions of images (of coins and apparently rocks and records) that users' images are compared against.

In each app, the user takes a photo, which is immediately sent to a cloud-based server where the AI model processes it. The AI's accuracy in identification (or lack thereof) relies entirely on the training data it has access to, and this is where the app falls short.

I believe the coin apps are being trusted with 4 tasks by those that use them - identification (broad and granular), grading and valuation.

It's a well-written review and I appreciate Andrew's thorough approach. As a lifelong techno-optimist I also believe the basic technology shows real potential, and could greatly improve with continued investment and greatly expanded and well-classified image libraries. But I haven't seen evidence of serious investment. The app's performance on the most basic task of broad identification hasn't improved since the CoinWeek review, where it did well at identifying basic U.S. coins and failed on world and ancient. Here's what Andrew found for "Task 1: Broad Identification."

The first task is broad identification - finding out what a coin is, in general terms. This isn't the only job people use an app like Coinsnap for, but it is the first task it's asked to do.

If we're talking about Australian users of the app, I think most of us know what a 2-dollar coin is. We know how to tell what year an Australian coin was made and can even have a basic idea about a commemorative coin just by looking at it. Foreign coins are another thing altogether - the letters are in different languages, they have different images on them, and even use different numbering systems, so conceivably that's where an app like CoinSnap can save a lot of time in identifying something that'd otherwise take hours to work out.

Unfortunately, my experience of the accuracy of CoinSnap's system when it comes to foreign coins is much lower than their stated accuracy rate of 99%. I've snapped a Tongan coin and it has said it was Mexican, I've snapped an Iraqi coin and it said it was from Switzerland! On both of those occasions, I could see the AI algorithm had matched the photos I took with images that looked similar in very broad terms, but the identification was completely wrong. An app that fails this basic task doesn't even pass the first test as far as I'm concerned.

I've seen the CoinSnap app develop capability over time (it's pretty neat the way you take 2 photos, it crops out the background and spins them around while it does the data matching in the background), so their image library will no doubt build over time, but it's sure inadequate at the moment.

You have to walk before you can run, and you have to crawl before you walk. These apps need to get the basics right before touting high-level capabilities. They're selling a mirage into a strange online feedback loop of bizarro articles claiming that common quarters "could be worth up to" a zillion dollars, and ads promoting apps like this that claim to be able to accurately tell you what any coin is worth. That makes downloading CoinSnap sound like a bargain, but it's a lottery ticket with a payoff that will never come.

I do think we'll one day see an app that does a good job of basic broad identification, and that will be useful and fun. But again, that will require investment and hard work, and higher levels of granular identification, grading and valuation might only work after much, much more investment and deep, deep specialization. -Editor

"I have to say I was shocked when I saw the revenue that the app generates.

"The game-changing app is going to be one that marries that image-matching technology with an accurate image library and accurate market values.

"With $10m a year flowing through their coffers, if they choose to invest that in some numismatic assets to build an app that actually works, that will be a significant development for our industry."

To read the complete article, see:
CoinSnap - A Fantastic App That's Not Fit for Purpose (https://www.sterlingcurrency.com.au/blog/news-research/the-fine-art-of-numismatics/coinsnap-a-fantastic-app-thats-not-fit-for-purpose/)

To read the earlier E-Sylum article, see:
APP REVIEW: COINSNAP (https://www.coinbooks.org/v26/esylum_v26n12a07.html)