Unique Error Code Generator

The unique error code generated below is formed using crypto.randomUUID() converted to base 62. This method produces 5.3 undecillion possible values.

XXXXXXXXXXXXXXXXXXXXXX

The Pros and Cons of using these unique error codes.

Pros

Collision Resistance

The odds of two codes ever colliding, having the same value, are vanishingly small. A discussion on the chances of collision is here. The algorithm, codeGenerator.js, can be viewed here on github.

Identification of the Point of Origin

Unique error codes help developers quickly locate a specific issue which occurred in the codebase. When the codebase is modified or reorganized, the unique error error code can potentially still communicate the same issue, even if multiple releases exist in the field.

Cryptographic Quality Randomness

It uses the platform's CSPRNG, so the output is unpredictable and well-distributed, which is much better than Math.random() based approaches.

Compact Representation

A raw UUID is 36 characters (with hyphens) or 32 hex digits. Converting the 128-bit value to base62 yields roughly 22 characters, which is meaningfully shorter while remaining URL-safe and human-typeable.

No Special Characters

Base62 (0–9, a–z, A–Z) is URL-safe, filename-safe, and readable in logs, emails, and support tickets without escaping or encoding.

Stateless Generation

No database sequence and no coordination between servers is required. Any instance can generate a code independently, which scales well.

Universally Available

crypto.randomUUID() is available natively in Node.js 14.17+, all modern browsers, and Deno. No dependencies are needed.

Cons

Not Truly 22 Characters

Due to how BigInt conversion works with the full 128-bit UUID value, you'll get a variable-length string which is up to ~22 chars.

Loses UUID Structure

A raw UUID v4 has a well-known format that tools, databases, and libraries understand natively. Once converted to base62, you lose that interoperability.

Case Sensitve

Base62 includes both upper and lowercase letters, so ABC123 ≠ abc123. This can cause user reporting errors in phone support contexts.

No Intrinsic Meaning.
The unique error code carries zero metadata. It has no timestamp, no service identifier, no severity. Systems like Sentry embed structured info in error IDs. If you ever want to decode "when did this happen" or "which service threw this," a pure random code won't help.

Overkill for Many Error Reporting Volumes

If your application generates thousands, not billions, of errors, even a simple nanoid or sequential ID would have negligible collision risk. The full UUID to base62 pipeline is robust but adds complexity that may exceed what the problem actually demands.