No single test proves that a label will stay on a glass bottle. A sound programme pairs each real exposure — shelf handling, an ice bucket, a chilled chain, oily or sanitized hands, a humid warehouse — with the method that imitates it, runs that method on production bottles carrying production labels, and judges the outcome against a control sample using failure definitions agreed in writing. A peel value taken from an adhesive data sheet does none of this, because it describes a bond to a standard panel, not to your bottle.
Nothing below is a specification. Adhesion belongs to one adhesive, one face stock, one glass surface and one set of service conditions, so every setting has to be confirmed by measurement on your own pack.
Five things a programme must state
A supplier either runs an adhesion programme or owns a purchased document. You can tell which by asking for five statements.
- The service condition being imitated. A dry ambient shelf and an ice bucket call for different tests, even when the bottle is identical.
- The surface state under the label. Flint glass, a frosted band and a sprayed lacquer are separate substrates, each with its own surface energy.
- The measurement and the person taking it. Peel force from an instrument and edge lift read off a steel rule are two kinds of observation. They cannot be averaged into one figure.
- The pass or fail rule, written as a judgement and not lifted from a data sheet.
- Monitoring after first approval. Adhesive, glass surface and labelling line all drift across a production season.
Labels do not fail in a laboratory. They fail on a retail shelf, in a cooler box, in someone's hand twenty minutes after the bottle came out of ice water, or on a pallet in damp storage. The programme exists to recreate those moments while a change is still easy to make.
Name one owner for the assembled-pack test
Four parties shape the bond, and only one of them ever handles the finished, filled, chilled unit. That is why adhesion so easily ends up belonging to nobody.
| Party | What it controls |
|---|---|
| Bottle maker | The glass surface leaving the annealing lehr, including any mould release agent or surface treatment still on the outside wall, plus waviness at the heel where a wrap label must conform |
| Decorator (if used) | Application and cure of the frosting or coating beneath the label; a part-cured coating is a different substrate from a fully cured one |
| Label converter | Face stock, adhesive and release liner, and the die-cutting that sets stiffness, cut quality and liner release — the things that decide conformity to a curve at line speed |
| Filler | Label application, often onto bottles that are cold, dusty or wet with condensation, before the pack enters distribution |
If each party tests only its own output, a field failure turns into a long exchange of clean reports. Appoint one owner for the assembled pack and require tests on production bottles with production labels under a stated condition. The labelling house is usually the natural choice, since it runs the application step and can reproduce its variables. It still needs the bottle maker's surface record and the converter's adhesive grade in writing.
Put the split in the purchase documents, not in meeting notes. One clause should name the test owner, the condition imitated, the reading method and the acceptance rule, and all four parties should agree to it before the first production run. Auditing then becomes simple: each party shows the clause and the record that satisfies it.
Identify the glass surface before blaming the adhesive
A lifting label draws attention to the adhesive, yet the surface is more often the root cause. On glass it varies more than most buyers assume.
Plain flint glass
Untreated flint is smooth, high in surface energy and slightly alkaline. Most pressure-sensitive adhesives wet it well when it is clean and dry. Trouble comes from contamination: mould release agents, silicone carried over from a nearby process, dust settled on the shoulder or base, hand cream or machine oil picked up at the labelling station. Each lowers the effective surface energy, giving a bond that looks fine on application day and releases weeks afterwards.
Silicone cannot be seen. The defence is a written cleaning and handling rule and a surface check performed at the labelling station, not in a laboratory.
Frosted and matte bands
A rough texture cuts both ways. It gives adhesive something to key into, but a stiff adhesive or low application pressure will bridge the valleys and reduce true contact area. Frosted zones also trap dust and moisture, so the drying step counts as much as the wipe.
Sprayed and lacquered bottles
Here two bonds are stacked: coating to glass, and label to coating. When a label peels from a coloured or soft-touch bottle with a thin skin of coating attached, the label is not at fault, and changing adhesive will not help. Cure state, film thickness and surface treatment of the coating decide the result, which makes the spray coating process for glass bottles the relevant reference.
Moisture and storage age
Two surface effects should be tested, not assumed. A bottle labelled only a few degrees above cold-room temperature may carry an unseen condensate film that stops the adhesive touching glass at all. Glass kept for months in humid storage can build up a film that a quick wipe leaves behind. Both tend to appear in the first real production week, so both belong in the test conditions.
Match each method to the exposure it imitates
Borrowing one method to answer another method's question is the usual source of false confidence. A standard explains how to run a test; it never says what your pack must achieve. The family is often called label durability testing, a useful name because it points at service exposure and away from whichever bench rig happens to be free.
| Method | Exposure it imitates | Settings that change the outcome | How to judge it | Where it applies most | Question for the label supplier |
|---|---|---|---|---|---|
| 180 degree peel, production samples | Everyday shelf handling; a consumer removing the label | Angle, speed, dwell, temperature and humidity when read, and the label width behind the force-per-width figure | As retention against a reference-conditioned control, stated as a proportion | Dry ambient packs in plain flint; wide wraps for spirits and oils | Adhesive grade, dwell applied, and whether the quoted value came from a standard panel or your bottle |
| 90 degree peel at heel or shoulder | Pulling at a curved label on a taper | Local curvature, applicator pressure, warm or cold application, where the peel line falls against the mould parting line | Consistency around the curve, not one peak; note lifting near the shoulder transition | Boston rounds, tapered oil bottles, strongly shouldered shapes | How the label was laid at that curvature, and whether pressure is controlled or left to an operator |
| Cross-cut / cross-hatch | Mechanical soundness of a sprayed, frosted or printed layer | Spacing, count and direction of cuts, blade condition, compressibility of the film | Detached proportion graded against photographs, not described in words | Glass whose coating or print is the surface under the label | Confirmation the coating was tested, not the label, and that production cure matches the tested sample |
| Water immersion, then peel or lift | Ice bucket, cooler box, sink, wet bar top, melt water on pallets | Soak time, water temperature, wet reading versus reading after stated recovery | Insist on the recovery reading: wet alone exaggerates damage, dry alone conceals it | Beverages, chilled sauces, hospitality packs, anything sold from ice | Whether the adhesive is rated for intermittent water contact, and how end users will store the pack |
| Chill / freeze cycle with warm recovery | Chilled distribution, cold stores, winter transport | Lowest temperature, hold time, warm-up rate, dwell given before chilling | Edge lift and curl after recovery; lift appearing only after cycling rules that adhesive out for that use | Chilled food and drink, cold pressed juices, refrigerated display | Temperature range the grade was chosen for, and whether the buyer stated it beforehand |
| Alcohol, oil and sanitizer wipe | Sanitized hands, product residue, bar and kitchen cloths | Wipe count, direction, force, contact time of the agent, print versus edges assessed | Corners and edges as well as print; log softening that recovers on drying as a separate finding | Food oils, spirits, hand sanitiser bottles, hospitality | Whether face stock or varnish protects the label, and whether an overlaminate is specified |
| Humidity and heat ageing | Warehousing, container transit, humid retail | Agreed per project | Rank one grade against another; never read it as predicted field life | Long shelf life, humid export markets, slow-moving stock | Whether ageing used assembled packs or loose label stock — the two disagree |
| Rub / scuff | Bottles touching inside a carton; contact with shelf edges | Cycles, load, abrasive medium, curved or flat test area | Legibility and edge integrity together; readable print with a frayed edge is rejected | Multipacks, cartons of small bottles, returns and reverse logistics | Availability of an overlaminate, and whether the pack must survive re-cartoning |
Peel
Peel tests record the force needed to strip a bonded label at a fixed angle and speed after a fixed dwell. The pressure-sensitive label trade usually works to the FINAT peel test methods, ASTM D903 and ASTM D3330; ISO 29862 plays a similar role for adhesive tapes. The result is force per unit width, and a figure quoted without angle, speed and dwell cannot be compared with any other figure.
Peel suits comparing adhesives and measuring change after ageing. It does not tell you whether a shopper will accept the pack, since a respectable peel value and visible edge lift can exist on the same bottle.
Cross-cut and cross-hatch
ISO 2409 and ASTM D3359 describe scoring a lattice into a surface and assessing what comes away. This is the standard route for coating adhesion on rigid substrates and works well on sprayed, painted or lacquered glass. On a soft pressure-sensitive label the blade cuts a compressible film unpredictably, so the result says more about face stock than bond. Keep cross-cut for the coating and assess the label by peel or lift.
Immersion and soak
A defined soak is read twice: straight out of the water, then again once the pack has recovered. Comparing the pair shows whether water merely softened the adhesive or displaced it from the glass. The second reading matters because people judge a bottle after lifting it from the cooler, not while it drips.
Wipes, ageing and rub
Wiping imitates handling, and it abrades as well as exposing the label chemically. A wipe test that inspects print alone will under-report. It matters most for food oils, where a thin product film outside the bottle is normal.
Accelerated ageing in a humidity chamber or controlled oven squeezes months of storage and transit into days. Provided the condition is stated and repeated consistently, it ranks two adhesive grades reliably. It predicts absolute field life poorly.
A Sutherland-type rub test, built on methods like ASTM D5264, measures print durability and surface integrity under carton and shelf abrasion. It sits alongside adhesion testing without replacing it: a label may come through rubbing with sound edges and still fail a chilled soak.
Treat the chilled chain as its own stress case
Labels qualified at ambient fail most often in chilled distribution. Three mechanisms combine in a cooler.
- Adhesive stiffening. Adhesive is organic and stiffens sharply when cold. A formulation that flows into the micro-roughness of glass at room temperature may be too rigid to wet out in the brief contact a labelling line allows.
- Face stock movement. Paper especially swells with warm humidity and shrinks in the cold. Glass stays dimensionally stable, so the mismatch appears as edge curl.
- Water at the perimeter. Condensate present during labelling, or melt water later, attacks the bond edge and creeps beneath the adhesive.
A realistic test is therefore a sequence. Apply under production-like conditions, give the adhesive its dwell, chill, hold, return to ambient, then read after a defined recovery. Freezing followed by controlled warm-up reveals the most, because it loads both the adhesive and the dimensional mismatch. A room-temperature peel cannot show what this sequence shows.
Two process consequences follow. First, tell the converter the pack is chilled before the adhesive is specified; a standard permanent acrylic selected for a dry ambient pack is not automatically right for hours in ice. Second, bottle temperature during labelling is a controlled parameter and should be checked periodically. Lines that label cold, damp bottles produce high and erratic failure rates whatever adhesive they use.
Write the acceptance standard as judgement logic
An acceptance standard built from borrowed figures will not hold up. Build it in five steps, the same structure an auditor should find in a supplier's own procedure.
- Describe failure in observable terms. Examples: edge lift past a stated small dimension, a flagging corner, a wrinkle that collects dirt, adhesive transferred to the glass on removal, a cohesive split leaving adhesive behind. Photograph each as a reference so two inspectors reach one verdict.
- Fix the reading conditions. Rest time after labelling, temperature, humidity, and whether the reading follows the stress directly or after recovery. Specifications omit the interval more often than anything else, and different intervals give different verdicts.
- Set a baseline. A control held at reference conditions turns readings into a comparison. Retention as a proportion of that control outlasts a change of adhesive or test rig; an absolute figure does not.
- Record the sample plan. ISO 2859-1 with an agreed AQL and inspection level gives a defensible unit count and tolerable number of failures. Which AQL matters less than writing it down, stating whether the plan is by attribute or variable, and tying it to a defined lot size.
- Define the response to a fail. Who is told, how the batch is contained, whether it is held, and what evidence the corrective action requires. Without this, the first failure becomes a negotiation between two quality departments.
Investigate a lifted label in this order
Sequence saves more time than any individual test. Work from the obvious toward the exotic.
Begin with a magnifier and the failed surface. It takes minutes and points one of three ways:
- A clean adhesive layer left on the glass is a cohesive failure and implicates the adhesive.
- Perfectly clean glass is an adhesive failure and implicates the surface or the wetting step.
- Coating film attached to the label places the problem below the label, in the coating process.
Next check surface and application: cleaning chemical and wipe direction, dust control, bottle temperature during labelling, dwell before handling, and applicator pressure over a curved heel. Reopen adhesive selection only once these are cleared. Face stock stiffness and label geometry, including overhang around a shoulder, come after that.
Handling and storage sit last on the list yet often settle the matter. A pack that leaves the line clean and lifts after a spell in a humid warehouse has an acceptable bond and inadequate durability. Those are different problems with different fixes, and keeping them apart is what the ranking is for.
Evidence to request when auditing a labelling line
The audit is brief because the evidence either exists or it doesn't. Request:
- the written procedure, read for its reading conditions and not its length;
- instrument records with the calibration trace and the repeatability check on the holding fixture — a drifting instrument gives confident wrong answers;
- retained samples from the last two or three production runs, kept at reference conditions;
- the surface record for the bottles in those runs, with coating batch where relevant, and the adhesive grade and lot actually used;
- the temperature and humidity log from the labelling station, the record that most often explains a first-season failure;
- the shelf life retention sample plan;
- the corrective action file from the last out-of-window reading.
Some answers signal a gap. An appeal to the adhesive supplier's assurance means nobody tested the assembled pack. "We have never had a complaint" means nothing was measured, since a programme that has logged no failures is not really a programme. "We test to the standard", offered without condition, dwell or interval, has no content. In each case ask for the last three sets of raw readings together with their retained samples.
What to settle before and alongside testing
Testing cannot rescue a label that is the wrong size or in the wrong place, so design decisions come first. Size, position, orientation and mandatory content on a food oil pack are covered under labels for olive oil bottles. Front and back label conventions in the wine trade, and the packaging formats around them, are under wine bottle labels. If the pack itself is still open, base formats and compatible surface options are listed with our glass bottle collections.
When we help plan a test programme, two inputs matter most. One is the use condition: hours chilled in a cooler, ambient storage in a humid market, or contact with oil and sanitizer in normal use. The other is the surface: plain flint, frosted across the label zone, or sprayed and lacquered, and whether the label lies wholly on one surface or straddles two.
Label stock and adhesive grade, pack format and fill volume, and the destination market where compliance applies are useful additions when already fixed. From these we can set out the exposures to imitate and their order, how to condition and read the pack, which causes to check first after a failure, and an acceptance rule that labelling house, filler and brand will all interpret alike.
Questions buyers ask about label adhesion testing
How does an adhesion test differ from a peel test?
Adhesion testing is the full programme: exposure, sample plan, reading method and acceptance rule. Peel is a single method within it. Request the programme and not just a peel number.
Bottle or standard panel?
The production bottle, every time. A standard panel shares neither the geometry nor the surface energy of your glass, and a flat plate behaves unlike a curved heel. Panel results are only good for comparing two adhesives with each other.
How long should testing run?
Long enough to cover the shortest realistic route to failure. For chilled products that usually means a few cycles, not a few months. Repeat the apply–dwell–chill–hold–recover–read sequence until one odd result cannot decide the verdict, and keep the readings as a baseline for later batches.
Why would one bottle colour hold a label and another not?
The colour is incidental; the surface treatment producing it is what differs. Two bottles that look alike from a distance may present coatings of different cure, thickness and texture. Review the cure and thickness records before altering the label.
Can condensation alone cause failure?
Yes, and it is among the commonest first-season causes. The label ends up bonded to a water film, then lets go days later with no apparent trigger. The remedy is procedural: a stated bottle temperature before labelling, humidity control at the station, and a record of that temperature.
How many samples are enough?
Whatever a recognised sampling scheme gives for your lot size, inspection level and AQL. A small set of well-conditioned samples under a written plan is worth more than a large unplanned one.
What does a clean, residue-free release indicate?
An adhesive failure, as opposed to a cohesive one. Typically the surface was contaminated, the bottle too cold, or wet-out incomplete. Look at the labelling station and the cleaning procedure before the adhesive specification.
Is a supplier's laboratory report sufficient qualification?
It shows capability and is a place to start. Verify it on retained production samples, under conditions you select, using your own reading interval and recovery period. What counts is whether the supplier can repeat the result on request and produce the raw readings.