The policy number on an insurance card is the member ID

Count the numbers on a US health insurance card. There are usually six or seven: a long one under your name, a shorter one labelled Group, two or three beginning with Rx, a bracketed number beside the word Issuer, and one or two more on the back. Exactly one identifies you as a person, and on most cards none is labelled “policy number”.

The number people mean by policy number is the long one under your name — printed as Member ID, Subscriber ID, Member Number, Enrollee ID or Contract Number depending on who issued the card, all the same field wearing different labels. It is what a clinic types into its system, what the pharmacy needs, and what you read out on the phone. Everything else describes your plan.

What each field is actually for

CMS publishes a one-page explainer of a generic card, Your Insurance Card, as part of its Coverage to Care programme. It labels the field Member Number and defines it without ceremony: “This number is used to identify you. It tells your provider how to bill your health plan.” Group Number gets an equally short definition — “this number is used to track the specific benefits your plan offers.” Neither is called a policy number.

Label	What it identifies	Who uses it
Member ID / Subscriber ID / Policy Number	You	Front desk, member services, claims
Group Number	Your employer’s plan, not you	Front desk, claims
RxBIN or RxIIN	Which processor receives the pharmacy claim	Pharmacy only
RxPCN	Which route inside that processor	Pharmacy only
RxGrp	Which pharmacy benefit within the plan	Pharmacy only
Payer ID	Where the electronic medical claim is sent	Billing office only
Issuer (80840)	That what follows is a US health plan identifier	Card systems

The Rx fields never come up at a doctor’s office, and the payer ID never comes up at a pharmacy. If someone asks for “the number on the card” without qualifying it, they want the member ID, with the group number as the usual follow-up.

The member ID has structure, and the structure matters

Blue Cross Blue Shield of Michigan’s 2025 provider brochure describes its own card precisely: the Subscriber ID is an “alphanumeric identifier composed of a three-character Blue Cross prefix followed by a Blue Cross-issued contract number”, and its sample card shows XYQ888888888. That prefix is not decoration — across the Blue system it routes a claim to the right Blue plan when a member is treated outside their home state. A wrong character there does not fail cleanly; it points the claim somewhere else.

Cigna’s provider guide shows the same shape from a different insurer: ID: U23456789 01 1. One letter, then digits, then a suffix.

UnitedHealthcare’s EDI guidance gives the outer bounds: the member ID is “required on all claims” and “can be a minimum of nine or maximum of 16 alphanumeric or numeric characters”. Nine to sixteen characters, letters or digits, no further rule. That is the entire specification.

The three Rx fields

These belong to a different system. NCPDP — the National Council for Prescription Drug Programs, an ANSI-accredited standards developer — sets what a pharmacy card must carry, building on the ANSI INCITS 284 standard for health care identification cards.

RxBIN is the routing number. NCPDP describes its Processor BIN as “a six-digit sequential number we assign that begins with ‘0’ as in ‘000010’”, which is why BINs on real cards look like 610011. The name is a leftover: BIN stands for Bank Identification Number, borrowed from the payment card world. NCPDP has since moved to an eight-digit Issuer Identification Number, but cards still overwhelmingly print six digits under the label RxBIN.

RxPCN, the Processor Control Number, is the secondary routing field — “a secondary identifier that may be used in routing of pharmacy transactions”. It has no fixed shape, and NCPDP is blunt about why: “There is no registry of PCNs. The PCN is alphanumeric as defined by the PBM/processor.” An RxPCN can be four characters or twelve, digits or letters or both, and nothing in its appearance tells you whether you read it correctly.

RxGrp identifies the pharmacy benefit group. NCPDP requires the routing block to carry the RxIIN plus “a Processor Control Number (RxPCN) and/or a Group Number (RxGRP) when required by the benefit administrator” — which is why some cards show all three and others only RxBIN and RxGrp. The card is not incomplete; that plan does not need it. NCPDP also fixes the print order, so on a photo the three always appear in sequence.

The back of the card

Nothing on the back identifies you — it is the half people photograph second and need first.

The payer ID is the one identifier there. UnitedHealthcare’s EDI guidance defines it for its billing partners: “all payers receiving electronic claims have one or more Payer ID numbers that indicate where claims are routed, similar to an electronic mailing address”, and “a Payer ID must be indicated to file a claim electronically”. It is typically five characters, though not always, and can be letters or digits. It belongs to a practice’s billing software, not to you.

The rest is contact detail and a postal claims address, and NCPDP requires that placement: the benefit administrator’s name and address go “on the back where non-electronic prescription claims, patient or provider correspondence can be sent”, help desk numbers beside them.

Subscriber, dependent, and the suffix

The contract holder is the subscriber; a spouse or child on the same coverage is a dependent. The distinction is usually encoded in the last two digits of the member ID rather than in a separate number. NCPDP’s sample card shows exactly this — an ID of 123456789 SFX 01, with a table beside it listing dependents at 02, 03, 04. CMS puts it as a warning rather than a spec: “if your spouse or children are also on your coverage, your member numbers may look very similar.”

Very similar is the problem. Two cards from one household can differ in a single digit at the end, and no amount of squinting catches that, because both values are valid. Medicare is the counterexample — CMS states that spouses and dependents each get their own randomly generated Medicare Beneficiary Identifier, sharing nothing.

None of these numbers can tell you it is wrong

An IMEI carries a Luhn check digit, a VIN a mod-11 check character, an IBAN a mod-97 remainder. Each contains arithmetic proof of its own integrity, so software can say yes or no — the divide that separates checkable numbers from uncheckable ones everywhere else.

A member ID has nothing of the kind. UnitedHealthcare’s published rule is a length range. CMS describes MBI characters as “non-intelligent” so that “they don’t have any hidden or special meaning” — deliberately random, with no relationship between positions to check. RxBINs are assigned sequentially, and an RxPCN has no registry at all. There is no checksum anywhere on the card.

What the standards bodies did instead is revealing. NCPDP addresses it at the printer:

A font size of eight (8) points or greater must be used when printing mandatory data; font style should clearly differentiate letters vs. numbers (e.g., alpha “O” vs. numeric “0” and alpha “I” and numeric “1”).

CMS went further and removed the characters from the alphabet: the MBI uses “numbers 0-9 and all letters from A to Z, except for S, L, O, I, B, and Z. This will help the characters be easier to read.”

Both are precautions taken by whoever designs the card, and neither helps you after the fact. If a card was printed in a face where a capital I and the digit 1 are the same vertical stroke — plenty were — the ambiguity is in the ink, and no amount of looking harder resolves it.

Which is why a letter prefix followed by digits — XYQ888888888, U23456789 — is the worst case. At that boundary a reader must decide whether a character belongs to the alphabetic part or the numeric part: IO against 10, S5, B8, Z2.

Getting it off a photo

The easy case first: your phone already does this. Open the photo in Photos on an iPhone and select the number with Live Text, or open it in Google Photos on Android and tap Lens. Both put a member ID on the clipboard in two taps, free, and for a single number nothing is faster. Their limits — pixel height, contrast, long strings with no dictionary behind them — are covered in copying text from a screenshot.

What none of them does is tell you the result is right. With a phone number, structure catches a misread: the numbering rules rule out an area code starting with 1. A member ID has no such rule, so the only control is reading the characters back against the card before you submit anything.

Build that step in. Our number extraction tool returns the digit runs from an image into an editable field beside the picture — slower than tap-and-copy, deliberately, because the point is seeing the result next to its source rather than pasting it blind. SwipeScan is our app and does the same from the camera, which helps with a stack of cards; it cannot validate a member ID either, because nothing can.

Two habits help most. Shoot the card square-on and filling the frame in ambient light, because flash on a glossy card puts a hotspot across the ID. And check the letters, not the digits — digits are where attention goes, the prefix is where the errors are.

The short version

• The policy number is the member ID: the long one under your name, also printed as Subscriber ID, Member Number or Contract Number.
• The group number identifies your employer’s plan, not you. A clinic normally wants both.
• RxBIN, RxPCN and RxGrp are pharmacy routing fields and nobody else uses them. NCPDP assigns six-digit Processor BINs starting with zero; RxPCN has no registry and no fixed format.
• The payer ID on the back routes the electronic claim. It is for the billing office, not the front desk.
• Dependents usually share the subscriber’s ID with a different two-digit suffix, so two cards in one household can differ by a single character.
• No number on the card carries a check digit. NCPDP’s answer to the O/0 and I/1 collision is a typography rule for the printer, so verification is yours: read it back against the original.