The $22,000 UV Ink Drying Failure: Lessons from a Mark Andy Flexo Press

The call came at 2:40 on a Tuesday afternoon in March 2024. Not that I'd remember the time if it weren't so precisely etched into my memory. It was our production floor supervisor, and his voice had that specific flatness that means something's been going wrong for a while before anyone calls quality.

"We need you down here. The labels are smearing."

Smearing. In the label business, that word sits somewhere between "problem" and "catastrophe." Smearing means the ink isn't dry—or isn't cured—and if it's happening across the run, you've got a pallet of questionable material and a stopped press.

Background: the press, the order, the pressure

For context, I've been the quality manager at a label converting plant in the Midwest for about five years. Before that, eight on the press floor. So I've been on both sides of the quality argument—the operator who thinks QC is just creating obstacles, and the QC guy who watches a preventable issue turn into a crisis.

We run two Mark Andy flexographic presses. The newer one is a ProSeries model with Mark Andy's integrated flexographic UV ink drying system—their Mercury unit. That press handles our prime label work: the stuff that goes on bottles and jars and has to survive a supply chain without looking beat up. The Mercury system is a serious piece of equipment. Water-cooled, variable output, designed to cure at production speed. It's a good system. But "good" doesn't mean "maintenance-free." Nothing in this industry is maintenance-free.

In Q1 2024, we landed a contract that felt like a milestone. A regional beverage brand wanted 50,000 labels for a seasonal run. Not the biggest order we've ever had, but the kind of client we wanted more of. Timeframe: four weeks, contract to delivery. We put it on the ProSeries because the job spec called for UV-cured inks—the client needed rub resistance, and UV gives you that.

The investigation: eliminating variables

When I got to the press, the operator had already stopped the line. Next to the rewinder sat a pallet of labels—maybe 8,000 of them—all with the same problem. Drag a fingertip across the trailing edge of the image and the ink moved. Not the whole image. Just the section that passed last under the UV lamps.

My first suspect was the ink. Two weeks earlier, we'd switched formulations because our regular supplier had a raw material shortage. The new ink was supposed to have the same cure response. "Supposed to" is doing a lot of work in that sentence.

"Nope, formulation's the same as our standard," the ink rep told me. "Should be fine."

Should be. Two words that have launched a thousand quality investigations.

Next suspect: press speed. The operator had been running the job at 400 feet per minute to hit the deadline. That's within the Mercury system's rated envelope, but I started to wonder if we were pushing against its real-world limit. So we ran a controlled test at 250 fpm on the same film.

The smear got better. Not gone, but better.

And that's when my brain started to itch. Because "better at lower speed" doesn't automatically mean "speed was the problem." UV ink curing is about energy density: the total UV energy striking the ink. Slower press means more exposure time per inch of web. If the lamps were outputting at rated power, 400 fpm might be just fine. But if the lamps were degraded, a slower speed would partially compensate.

I grabbed the radiometer from the QC lab—the device you use to measure UV lamp output. The recommended practice, from Mark Andy and our bulb supplier, is to check it at the start of every shift. Most of us treat that recommendation the way people treat their car's check-engine light: we know it's there, we know it's probably important, but the press is running fine and the day is short.

The reading stopped me cold. The lamps were putting out 60% of their rated peak irradiance. Sixty percent.

I checked the PM schedule in our CMMS. Bulb replacement was on there, right on schedule. But when the beverage contract came through, our maintenance tech had deferred the replacement by two weeks to "get through the run." He didn't flag it to anyone. He just moved the work order to a later date.

That's the thing about preventive maintenance: when nothing goes wrong during the extra two weeks, you start telling yourself the schedule was overly cautious anyway. Then a 50,000-label job comes along and you're staring at a pallet of smeared labels, a supplier saying the ink is fine, and a work order that was moved instead of done.

Two side stories that were the same lesson

Two other things happened that week, and at the time they felt like distractions. Looking back, they were the same lesson wearing different clothes.

First, our design team was mid-debate about moving proofing in-house. The options came down to a Xerox laser printer versus a high-end inkjet proofing device. Someone asked the eternal question: is a laser or inkjet printer better?

My answer, for our context: for proofing flexo work, inkjet wins, full stop. A Xerox laser printer fuses dry toner onto the substrate—fast, crisp, and immediately dry, which is great for office documents. But inkjet lays down liquid ink with a wider color gamut that gets you closer to what the flexo press will actually produce. According to Pantone's color matching guidelines, keeping brand-critical colors within a Delta E of 2 or less is the industry standard—and in our experience, inkjet proofs hit that target more often than toner-based output. For matching a job to a press? Inkjet, not particularly close.

Second, our maintenance supervisor had started experimenting with a large-format 3D printer with a 1000x1000x1000mm build volume. He was making brackets, guards, and jigs for the pressroom. Some of it was genuinely clever—he printed a light shield for a UV lamp housing that worked great. Then he floated the idea of printing replacement press parts for the Mark Andy itself.

I shut that down fast. 3D printing is a fantastic tool for non-structural fixtures and prototyping. But a flexographic press part—something with torque, alignment, or thermal requirements—needs engineering verification. The line between "prototype" and "production part" is exactly the line a quality manager is paid to guard. Cross it casually and you've replaced a defined specification with a hope and a prayer.

Both situations were, at bottom, the same problem as the smeared labels: people making reasonable-sounding decisions without a defined checkpoint to stop and verify.

The fix and the bill

The actual fix was embarrassingly simple. We replaced the UV bulbs the next morning. About $1,800 in parts. Then we reprinted the 8,000 affected labels and let the rest of the run continue. The remaining 42,000 labels were fine. The rework cost us roughly $22,000 when we counted materials, extra press time, and crew overtime.

And we lost three days on the client's timeline. They were professional about it—the reprint quality was good, and they paid on time. But goodwill gets spent in situations like that, and it doesn't refill quickly.

Even after I made the call to pull the press offline, I kept second-guessing myself. What if the radiometer was faulty? What if I was grounding a 50,000-label job over a bad reading? The hours until the new bulbs arrived were genuinely stressful. Turned out the radiometer was fine. I wasn't. But I didn't know that going in.

What we changed

The post-mortem came down to three root causes:

• The bulb replacement was deferred without approval or visibility.
• The press speed was increased without re-verifying cure response.
• No documented checkpoint between "UV system past its service interval" and "run a paid customer job."

None of those were malicious. Each one was a reasonable judgment call in the moment. That's what makes preventive systems hard: they aren't fighting laziness or incompetence. They're fighting a series of individually reasonable decisions that compound into a failure.

So we built a system that respects that reality. Every UV-ink job now starts with a 12-point pre-flight checklist. First item: a radiometer reading, logged into the system. If the lamps are below 80% of rated peak irradiance, the press doesn't run until that's sorted. We also wired the CMMS so any PM task deferred past its due date automatically emails me and the plant manager. Not to point fingers—just so the consequences of a judgment call are visible to someone besides the person who made it.

I also ran a training session on how UV curing actually works. Not "the lamps dry the ink," but: here's what irradiance is, here's why bulbs age, here's how speed interacts with energy density, here's how to read a radiometer. Several operators told me afterward it was the first time anyone had explained the why instead of just the what.

Honestly, I'm not sure why it took a $22,000 failure to make that training happen. My best guess: the day-to-day urgency of production work crowds out curiosity until the cost of ignoring it becomes impossible to ignore.

The bottom line

Five minutes of verification beats five days of correction. I can't prove it in a spreadsheet, but I can show you 8,000 labels that made a believer out of me.

One caveat: this approach worked for us because we're a mid-size label converter with two presses and a dedicated quality role. If you're running a smaller shop with a two-person crew, a 12-point checklist might be more overhead than value. You might be better off with a simpler system—replace the bulbs on a calendar schedule, every twelve months, no exceptions. And if you're not running UV at all—water-based inks, mostly digital work—the failure modes are different and so are the prevention points. I can only speak to my context: flexographic UV ink drying on Mark Andy equipment.

The principle holds regardless. Whatever your process, there's a version of "the radiometer reading" that you've been meaning to check. It costs five minutes. The alternative costs more.