Print one page. Not the job. One page.
Load the real CSV, pick the real template, set the real field toggles and offsets. Export PDF. Print page 1 only, at 100% scale, on the real stock, on the real printer. A proof on copy paper tells you the encoding is plausible. It does not tell you the liner lines up or the toner contrast is enough.
Hold the proof against a blank liner from the same pack. Check the bottom-right label, not just the top-left. Alignment failures stack toward the last row. If the last row is on the liner, stop and open the Avery 5160 alignment guide before you print page 2.
Three scans, three corners, two devices if you have them
Scan top-left, center, and bottom-right. A phone camera app is a start. Many warehouse guns are less forgiving: they want higher contrast and a true quiet zone. If you have both, the gun wins. A pass on a phone and a fail on a gun is a fail.
Read the decoded value out loud against the spreadsheet row. “It beeped” is not a test. You are checking that the value is the FNSKU or GTIN you meant, not that some barcode existed. Mixed sheets are where a UPC in an FNSKU column still “scans” as a different symbology and still wrecks inbound.
Lighting, angle, and the quiet zone you can see
Scan under the light you will use on the bench, not under a desk lamp you will never take to the warehouse. Gloss and laminate glare are lighting problems. If you have to tilt the phone to get a read, a gun on a belt will not tilt for you. Matte stock, no pouch, bars not against the die-cut.
Look at the left and right of the bars. You should see paper, not title text, not a logo, not the next label’s edge. That pad is the quiet zone. LabelStack insets the symbol; if you then opened the PDF in a design tool and stretched the bars, you may have stolen it back.
What a desk test cannot promise
Amazon fulfillment centers and grocery receiving lanes have their own guns, their own firmware, and their own patience. A desk test catches encoding, check digits, quiet zone, contrast, and alignment. It cannot catch a policy problem (you printed a UPC when they wanted an FNSKU) or a damaged poly bag in transit. For those, the symbology chooser and the FBA guide are the pre-flight, not the scanner.
When the proof passes, print the rest of the job on the same driver settings. Do not “improve” quality, paper type, or scale between proof and run. Change one variable and the proof is void. Keep the CSV, the PDF, and a photo of the print-dialog settings if the job is large enough that a reprint would hurt.