QR vs Data Matrix vs PDF417 vs Aztec: Picking a Symbology

The label on a prescription box, the boarding pass on a phone screen, the train ticket in someone's wallet, and the poster in the station concourse all carry two-dimensional codes. Usually only one of the four is a QR code, and the other three are not there by accident or by legacy.

Each of these formats was designed around a specific reader in a specific situation. Picking between them is mostly a question of who is holding the scanner.

Start from the reader, not the code

The single most useful question is what device will point at the symbol, and whether you control that device.

If the answer is a phone belonging to someone you have never met, the decision is already made and the rest of this is background reading. If the answer is a fixed-mount imager on a production line, a handheld unit issued to your own staff, or a gate reader at a station you operate, then the field opens up and the trade-offs become real.

Everything downstream of that question, capacity, footprint, error correction, print method, is negotiable. Reader support is not.

QR: the only one a stranger's phone will read

Three finder patterns in the corners, an alignment pattern in the fourth for all but the smallest sizes, Reed-Solomon error correction at four selectable levels, and encoding modes tuned for numeric, alphanumeric, byte, and kanji content. The structure and its history are well documented, and the read sequence explains why the corner squares carry so much of the work.

The decisive property is not technical. Current iOS and Android camera apps detect QR codes natively, with no separate app and no user configuration. No other 2D symbology has that. It means the cost of asking a stranger to scan a QR code is close to zero, while the cost of asking them to scan anything else includes finding and installing something first.

Capacity is comfortable for typical use. URLs, contact records, structured payloads, plain text, and Wi-Fi credentials all fit without difficulty at ordinary sizes. Producing them at volume is well supported too, whether that is a handful for a campaign or thousands of unique codes at once.

The cost of all that is footprint. QR needs a four-module quiet zone on every side, and its three finders consume a fixed share of the grid regardless of how little content you encode.

Data Matrix: for when the part is smaller than the label

Data Matrix replaces the three corner squares with an L-shaped solid border on two adjacent sides and an alternating row and column on the other two, which the reader uses to establish orientation and module pitch. Current implementations use Reed-Solomon correction. Both square and rectangular variants exist.

The advantage is density and margin. Data Matrix requires a far narrower quiet zone than QR, and its finder geometry has less overhead, so at a given module size and a small payload it takes up meaningfully less area. That is why it has become the standard choice for direct part marking on electronics and for marking medical devices, where the code goes onto the item itself rather than a label, and the item may be a few millimeters across.

It scans well on difficult surfaces: etched metal, dot-peened castings, molded plastic, curved bodies. The reader is normally a fixed or handheld imager set up for the job, with lighting chosen to suit the marking method. Phone support exists in dedicated apps but not in the native camera, which places Data Matrix firmly on the controlled side of the line.

PDF417: a stack of linear barcodes

PDF417 is structurally unlike the other three. Rather than a matrix, it is a stack of rows, each built from bar and space patterns in a linear format, with start and stop patterns at both ends of every row and Reed-Solomon correction at selectable levels.

That construction has a specific consequence: a sweeping laser scanner can read it, because each row presents as a linear barcode. Matrix symbologies including QR, Data Matrix, and Aztec need an area imager that captures a two-dimensional image. Anywhere a laser scanner is already installed and the payload is too large for a one-dimensional barcode, PDF417 is the format that fits the existing hardware.

It is used on driver's licenses in several countries and on printed boarding passes, both cases where a fairly large text payload has to sit on a document and be read by equipment the issuer specifies. The shape is wide and short by nature, which suits a strip along the bottom of a card and suits almost nothing on a poster.

Aztec: designed to survive a ticket barrier

Aztec puts a bullseye of concentric squares at the center of the symbol and grows outward from it in rings. Its defining practical feature is that it requires no quiet zone at all, so it can sit directly against surrounding print with nothing set aside around it.

That property makes it a natural fit for transport tickets, where it appears widely on rail tickets and on some airline boarding passes. Ticket stock is crowded, the code often shares space with dense text, and the reader is a gate or a handheld unit chosen by the operator. The central finder also tends to be easy to locate on a screen held at an angle, which matters when the ticket is a phone rather than paper.

Reading the decision in reverse

Work backward from the constraint that will not move.

The public is scanning it with their own phone: QR, without qualification. Retail signage, packaging, table tents, receipts, window displays. Codes in a retail setting live or die on whether a shopper can scan without installing anything.

The mark goes directly onto a small or awkward part, read by equipment you own: Data Matrix, with the marking process validated on the real substrate before the design is fixed.

Laser scanners are already deployed and the payload is a few hundred characters of text on a card or document: PDF417, assuming you also control what gets issued.

The symbol goes on a crowded ticket read by gates you operate: Aztec, and the absent quiet zone is doing real work for you there.

One test cuts through most of the debate. Take the phone in your pocket, open the camera, and point it at a candidate symbol. If nothing happens, that format is not viable for anyone outside your organization, no matter what its density or error correction figures look like on paper. Checking a code against real devices before committing is worth more than any specification comparison, and the same reasoning applies when weighing a printed code against a contactless alternative like NFC.