The stencil is on the inner forearm, purple against skin, about four centimeters wide. It looks right. The artist waits while the client points a phone at it. Nothing. Different angle, nothing. A second phone with better light, and it resolves on the fourth try.
That is the moment to stop, not the moment to push through. A stencil that scans reluctantly becomes a tattoo that does not scan at all, because everything that happens after the needle makes the symbol worse.
Skin Is a Difficult Substrate
Four forces work against the grid.
Ink spreads. A line in the dermis is not a printed line. Pigment migrates a little while it heals and keeps softening over years. A module edge that is crisp at week two is fuzzy at year five, and blowout in thin skin turns a sharp corner into a haze. Decoders sample near module centers, so mild softening is survivable. Adjacent modules merging into each other is not.
Small modules bleed together. Two dark modules side by side are normal. The failure is a single light module trapped between dark ones filling in. Once that happens, the timing pattern, which is the row and column of alternating modules a decoder uses to lock onto the grid, stops alternating.
Bodies curve and flex. Decoders tolerate perspective and moderate warp because the three finder patterns and the alignment patterns let them fit a mapping onto the symbol. A forearm at rest and the same forearm flexed are two different surfaces. A code wrapped around a wrist or a calf presents a surface the camera can only ever see part of.
Skin tone eats contrast. Binarization is a luminance decision. Black ink on pale skin gives a wide gap. Black ink on deep skin gives a narrow one, and the narrower it is, the more the outcome depends on lighting. White ink as the light module is not a fix; it heals unevenly and its opacity is unpredictable.
The Payload Decides Everything Downstream
Symbol version, 1 through 40, running from 21x21 modules up to 177x177, is chosen by how much data goes in and how much error correction you ask for. Every extra character pushes toward more modules, and since the tattoo has a fixed physical width, more modules means each one is smaller.
That single relationship, explained in how versions scale, is the whole game on skin. A destination of 20 to 25 characters sits near the bottom of the range with fat, forgiving modules. A 90-character tracking URL with campaign parameters does not.
So strip it. No query strings, no parameters, short host, short path. There is also a quirk worth using: the alphanumeric encoding mode covers digits, uppercase letters, space and a few punctuation marks, and packs them more densely than the general byte mode. An all-caps short URL can land a version lower than the same address in lower case, provided your server treats the path as case-insensitive.
High Error Correction, With a Catch
Four levels exist: L, M, Q and H, recovering roughly 7, 15, 25 and 30 percent of the codewords in a symbol. For skin, H is the obvious pick and also the correct one. The level comparison holds here with unusual force, because damage on a tattoo is not an accident, it is the default state.
The catch is that error correction lives as extra codewords inside the same symbol. Moving from L to H on an identical payload pushes the version up, adds modules, and shrinks each module at fixed physical width. You buy tolerance and pay in module size, and module size is what skin attacks hardest.
The way out is not to compromise on the level. It is to shorten the payload until H fits in a low version. H plus a short URL plus a generous symbol is a good tattoo. H plus a long URL at the same width is a worse tattoo than M would have been.
Size and Placement
Work backwards. Decide the smallest module you trust healed ink to hold, multiply by the module count of the version you landed on, then add four modules of quiet zone on every side. That is your minimum width. The arithmetic matches any printed code, with a stricter floor because the substrate is alive and changes.
Placement rules follow from the physics:
- Flat, low-movement areas: outer forearm, inner forearm, upper back, outer thigh, calf.
- Nothing that forces a wrap: wrist, ankle, ribs, fingers, the side of the neck.
- Avoid areas with heavy stretch or constant sun, since faded ink is low-contrast ink.
- Keep the quiet zone genuinely empty. Four modules of bare skin on all four sides, with no neighboring linework and no plan to fill that space later.
The last one gets broken constantly. A code placed inside a sleeve loses its quiet zone the first time something new is added beside it.
Contrast Is Not Just About Black Ink
The decoder needs a luminance difference between dark and light modules that survives whatever light the phone happens to have. Solid fill, no gray wash, no shading or texture inside modules, no color that flattens to a mid-tone in grayscale. The rules for contrast apply exactly as they do on paper, with less room for error, since one of your two colors is skin and you did not get to choose it.
On deeper skin tones, contrast becomes the binding constraint rather than module size. That usually means going larger. Sometimes it means the honest answer is that a particular design will not read reliably at any size a person wants on their arm. An artist who says so before starting is doing you a favor.
Test the Stencil, Then Test It Again
Before any needle touches anything:
- Generate the final symbol at the exact physical dimensions planned.
- Print it on paper, hold it against the spot, and scan from 20 centimeters, from arm's length, in daylight and in a dim room.
- Apply the real stencil and repeat on the actual body part, in the position it will normally be seen.
- Use three phones, including an older one. Camera and decoder quality vary more than people expect.
- Flex. Scan with the muscle tensed, relaxed, and with the limb rotated.
If any of those five struggles, the design is too dense, too small, or in the wrong place. Fix the file instead of hoping. Expect a step down in readability between the printed test and the healed result, and treat that gap as the margin you are deliberately building in. Anything touching aftercare, skin health or placement over scar tissue belongs to a professional artist and, where relevant, a doctor. None of this is medical advice.
The Ten-Year Problem
A tattoo outlives most links.
Domains lapse. Companies get acquired and retire old paths. Redirect services close, sometimes with notice and sometimes without, and every code pointing through them becomes a symbol that scans perfectly and lands nowhere. Marketing teams live through this regularly when a generator shuts down, and a poster can be reprinted. Skin cannot.
The only defensible choice is a destination you control and can repoint: your own domain, a short path, a DNS record and a redirect rule with your name on them. Then the tattoo becomes a permanent pointer to a target you can change, and in ten years you edit a redirect instead of your arm. If a service manages that redirect for you, find out what happens to the link the day you stop paying, and prefer one where the code itself never expires.
Even then, be honest about the deal. This is permanent art built on an internet that has never kept a promise for a decade. Some people accept that and choose a destination whose meaning survives a dead link. Others tattoo the code and treat the working link as a bonus.
If you are going ahead: pick the destination first, cut the URL below 25 characters, generate at level H, print it at final size, and scan the paper version ten times in five different rooms before booking the appointment. Then hand the artist a vector file, not a screenshot.
