Processed locally
Your input stays in this browser.
Recovery works on codewords, not on a painted area
A QR symbol stores data in Reed–Solomon codewords. Levels L, M, Q and H reserve progressively more redundant codewords, so a damaged symbol may still be reconstructed. That redundancy is not a map of safe pixels. A camera must first find the three large finder squares, follow timing patterns, and read format information. Covering one of those structures can stop detection before the decoder reaches the recovery step.
A centred square can hide different things in different versions
Suppose a short URL produces a 29 by 29 module symbol and a contact card produces a 49 by 49 module symbol. A 9 by 9 logo covers 81 modules in either image, but it occupies about 9.6% of the first matrix and 3.4% of the second. More importantly, the locations underneath it can include ordinary data in one version and an alignment pattern in another. Area percentage therefore cannot predict whether either design will scan.
Work the logo test as a controlled comparison
Start with the same payload, dark modules, light background, and correction level H. Export a plain 33 by 33 module control, then add a 7 by 7 module logo panel in the centre without touching finder squares or the outer quiet zone. If the control decodes and the branded export fails, the payload is exonerated: reduce the panel, restore contrast, or remove the logo. If both exports fail, inspect the encoded text or the encoder before changing artwork.
A white halo has a cost as well
Designers often put a pale box behind a mark so that it remains legible. That box erases modules just as the mark does, even when the logo itself has transparent pixels. A coloured halo can also make pale modules merge with the background after compression or printing. Count the complete cleared panel when comparing alternatives; judging only the visible ink understates the damage.
Print changes the failure mechanism
A phone display preserves square modules and sharp edges. On uncoated paper, ink can spread into the light gaps; on metallic packaging, reflected light may wipe out contrast; on a curved bottle, the camera sees a warped grid. Those faults are not repaired by choosing H. Test the finished raster and then a proof on the intended substrate, because a logo design that works at 800 pixels may fail after halftoning.
The boundary is an approval test, not a coverage rule
No fixed logo percentage certifies every payload, encoder, print process, and camera. Keep the original unbranded export beside the exact branded proof, record the payload and correction level, and scan both with ordinary phones under the intended conditions. When a design needs a larger mark than the proof tolerates, the honest answer is a different visual treatment, not a claim that unused recovery capacity will save it.
Separate correction level from artwork clearance
The letters L, M, Q and H select a Reed–Solomon recovery setting in the QR symbol. They do not reserve a central blank canvas. A logo panel removes whichever modules happen to sit beneath it, while the camera still needs intact finder squares, timing structures and format information to identify the symbol. This is why a design can pass with a short URL and fail with the same visible logo after a longer payload produces another version of the matrix.
Measure the cleared panel, not the coloured logo
Imagine a 9 by 9 pale square placed behind a 5 by 5 transparent logo. The code loses all 81 modules in the pale square, not only the 25 modules covered by coloured ink. On a 29 by 29 matrix that is 9.6 percent of matrix positions; on a 49 by 49 matrix it is 3.4 percent. Neither percentage says whether it crosses a functional pattern. Record the panel's module dimensions and location instead of approving a slogan such as 'under ten percent coverage'.
Run a four-file comparison
Create a plain black-on-white control and a branded version at the same payload and correction level. Export each as a source SVG and as the final raster or PDF artwork. Decode all four. If SVGs pass but the branded PDF fails, the logo compositor, PDF export or colour treatment is implicated. If both branded files fail, reduce or remove the cleared panel. If even the plain control fails, inspect the payload and encoder before discussing the logo. This sequence makes the result actionable rather than aesthetic.
Keep clear of structures that make location possible
The three large corner finder squares, timing rows and columns, format information and version-dependent alignment patterns are not ordinary data. A centred logo normally stays away from finder squares, but it can still overlap alignment or data patterns in a larger version. Do not crop the outer quiet zone to make room for branding; the cleared logo panel and the quiet zone are separate losses. GS1's 2D guidance specifically warns that distortion and modification make 2D codes harder for scanners to identify and decode.
A packaging example with numbers
A cosmetics label has an 80 mm allocation. Its 41 by 41 matrix plus eight quiet-zone modules creates a 49-module panel, or 1.63 mm per module. A centred 11 by 11 white panel is 17.9 mm wide and removes 121 module positions before printing. It may look modest on screen yet consume much more recovery margin than a 7 by 7 panel. Print both options on the actual coated label. If the larger panel only reads when the phone is perfectly square to the label, choose the smaller mark or a separate brand area.
Printing errors add to the logo loss
Error correction cannot distinguish a deliberately erased centre from dot gain, blur, glare, scratches or a fold. A metallic substrate can lower the contrast of the entire grid; a varnish may make only one viewing angle fail. The branded proof must therefore be tested after the same finishing process, not merely in the generator. Keep the plain control beside it so that a failure can be classified as a general reproduction problem or a cost introduced by the visual treatment.
Do not turn a passed test into a percentage promise
ISO/IEC 18004 specifies QR Code symbol formats and error-correction rules, but it does not license a universal logo-area percentage across every payload, printer and camera. The right approval record names payload, matrix size, correction setting, panel size, final file and successful physical proof. If a marketing layout needs more blank area than that record supports, redesign the label. A visible logo is optional; a code that a customer cannot locate is not a successful brand treatment.