Customer ID formats, compared
Sequential, random, prefixed and pattern-based customer IDs — with examples, entropy math, and guidance on when each one fits.
Every customer ID you have ever seen belongs to one of four families. Knowing which family you are looking at — and why it was chosen — makes you dramatically better at designing your own.
1. Sequential IDs
CUS-0001, CUS-0002, CUS-0003… The oldest and simplest scheme: a counter with a fixed-width, zero-padded number. Strengths: trivially sortable, human-readable, easy to read over the phone. Weaknesses: fully predictable — anyone can guess the next ID — and they leak business information (a competitor can estimate your customer count from ID ranges). Use for internal records, never for anything access-related.
2. Random IDs
CUS-7KQ2-9XD4 — unpredictable strings drawn from a large character set. Strengths: unguessable, no information leakage, no coordination needed between servers generating them. Weaknesses: harder to read aloud, and collisions — while astronomically unlikely at sane lengths — must still be guarded against with a database unique constraint. The generator's Random tab produces these; the entropy gauge shows exactly how much unpredictability your settings buy.
3. Prefixed IDs
USR-00001 — a short type prefix plus a number. The prefix is doing real work: it tells humans (and code) what kind of record this is at a glance, which prevents the classic bug of passing an order ID where a customer ID was expected. Common prefixes: CUS, USR, MBR (member), CLI (client). Keep prefixes short, uppercase and unambiguous — avoid lookalikes like CUS vs CUST in the same system.
4. Pattern-based IDs
CUS-2024-A8K2 — structured templates mixing literals, dates and random segments. Strengths: they encode useful context (year, region, batch) while keeping a random segment for uniqueness. Weaknesses: every encoded field is a future constraint — a year segment becomes awkward for multi-year records, and region codes rot when regions change. Encode only facts that will stay true forever.
Entropy cheat sheet
| Format example | Random part | Approx. entropy |
|---|---|---|
| CUS-#### (4 digits) | 10,000 combinations | ~13 bits |
| CUS-######## (8 digits) | 100M combinations | ~27 bits |
| XXXXXXXX (8 × A-Z0-9) | 2.8T combinations | ~41 bits |
| XXXXXXXXXXXXXXXX (16 × A-Z0-9) | — | ~83 bits |
These are computed from the character-set sizes, not measured — the same math the generator's gauge uses live. Rule of thumb: 40+ bits is plenty for non-security identifiers; anything access-related wants 80+.
Choosing between them
- Internal records, support desks: prefixed sequential — readable, sortable, phone-friendly.
- Public-facing references (order tracking, share links): random — unguessable by default.
- Multi-tenant or multi-region systems: prefixed + pattern — the prefix prevents cross-type mixups.
- Test data and demos: any of the above — which is exactly what the generator is for.