Specify wall thickness as a target plus an allowed band at named points on a drawing, add a sampling plan and a rejection rule, and verify it with ultrasonic readings before the goods ship. A bare "3 mm wall" on a purchase order is not enforceable, because bottles from one mold can differ by a millimeter or more at the same spot. It is the thinnest point, not the nominal figure, that decides whether a bottle survives the filling line, the drop test and the sea voyage.
Why thickness varies from bottle to bottle
Glass containers are formed, not machined. On an IS (individual section) machine, a gob of molten glass is spread inside the mold by air pressure or a plunger, in either the blow-and-blow or the press-and-blow process. How evenly it spreads depends on gob temperature, variation in gob weight, mold temperature, machine timing and equipment wear.
The outcome is variation in three directions: around the circumference of one bottle, along the body from heel to shoulder, and between bottles over a run. Tolerance is the agreed band around the nominal value that covers this spread. "2.5 mm ±0.5 mm at the body" means every measured body point must lie between 2.0 mm and 3.0 mm.
Leave the band out and a bottle with a 1.6 mm side panel can still be described as "2.5 mm average", with no contractual basis for rejecting it. How nominal thickness relates to strength, weight and cost is covered separately in our guide to choosing glass bottle wall thickness.
Realistic tolerance bands on IS machines
For standard soda-lime containers from modern IS machines, the following bands are reasonable starting points for negotiation. Bottle size, shape complexity and the age of the equipment all move them.
| Container | Nominal body wall | Common tolerance |
|---|---|---|
| Small bottles, under 250 ml | 1.8–2.5 mm | ±0.3–0.5 mm |
| Mid-size bottles, 250–750 ml | 2.2–3.5 mm | ±0.4–0.6 mm |
| Large bottles and jars, 750 ml and up | 3.0–5.0 mm | ±0.5–0.8 mm, widest at heel and shoulder |
Non-round shapes need separate treatment. Square, oval and flask profiles gather glass in the corners and run thin on the flat panels, which can measure 20–30% below the corners. Write an explicit panel minimum for these shapes.
Newer machines with closed-loop gob control hold narrower bands than older lines. Aim for the tightest band the plant can hold consistently, not the tightest number that looks good on paper. Ask for process-capability data, meaning the actual measured spread from recent comparable runs, before you agree the figure.
What goes wrong when the band is loose
- Weak spots. Impact resistance follows the thinnest point. A bottle that averages 3.0 mm but has a 1.8 mm panel breaks at that panel in a drop test. Vertical load strength, which matters when cases are stacked ten-high in a warehouse, behaves the same way.
- Weight drift. Glass is sold and shipped by weight. A run that drifts +0.4 mm on the body adds grams to every bottle and tons to every container, raising both the invoice and the freight bill.
- Filling-line trouble. Thin shoulders flex under capping pressure and thin heels crack on conveyor impacts. A line set up for one thickness profile begins rejecting or breaking bottles when that profile shifts.
Four elements of an enforceable thickness clause
A tolerance clause only works if all four of the following are present.
- Measurement points. Name them and mark them on a drawing: usually mid-panel on the body, the shoulder transition, the heel and the finish ring. On non-round bottles, list "flat panel center" apart from the corners, so nobody can argue about where the gauge was placed.
- Target and band for each point. When strength is the real concern, a minimum is stronger than a symmetric band, for example "Body: 2.8 mm minimum 2.2 mm". Minimum-only specifications are common for spirits and export containers.
- Sampling plan. State how many bottles are measured per lot, for example 32 pieces per 100,000-bottle lot, taken across the run and across mold cavities. Each cavity of a multi-cavity mold produces slightly different bottles, and cavity-level data is what exposes a worn mold half.
- Acceptance rule. Say which readings trigger rejection, whether a failed lot may be sorted at the supplier's expense, and who pays for re-inspection.
Include this in the technical package that goes out with the RFQ. Glass distribution is built into the mold design and machine setup, so asking for a tighter band once tooling is finished is where projects stall.
How thickness is checked before shipment
Three methods are used, and they answer different questions.
| Method | What it tells you | When to use it |
|---|---|---|
| Ultrasonic thickness gauge | Wall at a chosen point, readable to 0.1 mm, without damaging the bottle | The core of any inspection, at the plant or in your own warehouse |
| Sectioning | Direct wall measurement on a cut bottle with calipers or an optical comparator; destructive but definitive | First-article samples, and whenever ultrasonic readings do not agree with bottle weight |
| Weight check | Total glass per bottle on a 0.1 g scale, compared with the approved golden-sample weight band | A fast screen for a whole lot |
Weight has a blind spot: it detects a change in total glass but not in distribution, so a bottle can be on weight and still have a thin panel. Use it to screen lots, then confirm with gauge readings at the specified points.
On larger export orders, a third-party pre-shipment inspection by SGS, TÜV or a regional equivalent is worth considering as an independent check on a drifted run. Our guide to glass packaging quality inspection explains how to structure that inspection from start to finish.
Checkpoints from sample approval to repeat orders
Thickness control is a sequence, not a single test. When we coordinate a project, the checks fall at these stages.
- Golden sample. Approve it together with measured thickness at every specified point, and keep the data on file alongside the bottle.
- Purchase order. Put the sampling plan, cavity identification and rejection rule into the PO itself.
- Production. Measure first articles at the mold trial, repeat the checks mid-run on long runs, and inspect export orders before shipment.
- Repeat lots. Track the lot data over time. A slow rise in average thickness is a cost leak, and a slow fall is a breakage claim in the making.
Sending the drawing and target band with the first inquiry also makes the quotation more accurate, because the plant knows from the start what it has to hold.