We printed two thousand little squares, stuck one beside every till, and the scan numbers looked great. Orders barely moved. When we actually stood in the shop for an afternoon, the reason was embarrassingly simple: people saw the code on their way out, already holding a receipt, in a hurry to leave. We'd put a landing page in the one spot where nobody was deciding anything.

Look, a QR code isn't a graphic you generate. It's a tiny piece of real estate with a person standing in front of it, in a particular mood, with a particular amount of time and a particular signal strength. Treat it like a landing page and the usual questions come straight back: who's here, what do they want right now, and what's the one thing we should ask them to do.

Placement is the copy

The same code, in different spots, meets completely different people. Beside the till: already paid, leaving. In the queue: waiting, bored, roughly thirty seconds. On the table: seated, settled, willing to look. On a cup or a box: carried away, looked at later. On a poster or in the metro: standing, weak signal, a few loose seconds. The line next to it has to answer what do I get if I scan this now, not scan to learn more about us — that second one doesn't work in any of those spots.

An online code and a printed one are different objects. The online one can be swapped instantly, carries parameters, splits by channel. The printed one lives in a batch, so changing one sentence means reprinting all of it. That's why you test a small run first instead of committing a hundred thousand stickers to copy nobody has read yet. Then there are the two details nobody enjoys: print quality and weak networks. Size, contrast, quiet zone, glare from lamination, worn corners — all of it decides whether the thing opens at all. And if the first screen pulls a huge image, a person on a bad connection is gone before it loads. Keep that first screen light. Don't show off there.

One code, one scene

One scene value and one parameter set per placement. Don't share a single code across every channel, because all you'll learn is that somebody scanned, never from where — and then you can't decide which spot deserves another run of stickers. Give the material a number, a placement date, a batch id. Test two lines of copy on a small scale, leave one spot with no code as a control, and you'll actually know whether the lift came from the sticker.

One trick saves a lot of grief: don't burn the parameters into the printed artwork. Point the fixed pattern at a short link you can edit, so the drawing stays the same while the destination changes. Online it's easier still — swap it, split it by channel, pull it whenever you want.

PlacementWhat the person is doingWhat to measureThe usual misread
Beside the tillAlready paid, receipt in hand, leavingOrders actually started after the scanCounting scans as intent to buy
In the queueWaiting, bored, roughly thirty secondsOpens that reach the useful first screenAssuming a scan means they read the copy
On the tableSeated, settled, willing to lookRepeat use an hour laterCrediting the sticker for a return visit
On the cup or boxCarrying it away, will look laterScans that happen off-premise, hours laterCalling same-day silence a failure
On a poster or transit adStanding, weak signal, a few secondsSuccessful opens on a slow connectionBlaming the creative when the payload never loads

The numbers lie, and the sticker can be swapped

Now the costs. A scan is not a customer. Scan, open, leave — that counts. Scan and nothing loads — that usually counts too. It's a lovely number for a slide and it says nothing about the business. Offline iteration is painfully slow as well: a hundred thousand stickers out there means waiting for the next batch to fix one sentence, so don't ship unvalidated copy all at once. That isn't a campaign, it's a bet.

Channel attribution gets muddy too. Someone photographs the code and drops it in a group chat, people from the shop next door scan it, another campaign reuses the artwork, and the scenes blur together. Don't use it for precise channel ROI; knowing which kind of placement is clearly better is already enough to act on.

And there's a risk almost nobody guards against: the code gets covered or replaced. Someone peels off your sticker and puts up one that points somewhere else, the user ends up in the wrong place, and the reputational bill lands on you. Check physical spots on a schedule, keep a branded frame around the code and a short visible domain people can sanity-check at a glance. Anything that arrives through a scan is untrusted input, the same way a link parameter is.

Next time, put one code in one place, with one scene id and one line that says what the person gets right now. Print a small batch, stand there and watch five people walk past it, fix the sentence, then print the rest. If the scans are healthy and the orders aren't, don't reach for a new design — go look at where the sticker actually is.

Sources