How to Vet Custom Conveyor Belts Suppliers: 12 Checks

Ordering one custom belt is easy. Ordering the tenth copy of that same belt, three years later, from a supplier who has quietly changed its splicing crew, its fabric lot and its press operator, is where most procurement failures actually surface. The distance between a good first article and a good repeat order is the thing a real vendor audit has to close.

That is a different job from vetting a catalogue item. With a standard belt you can argue about grade, cover class and price. With a bespoke one you are also buying somebody's interpretation of your drawing, and interpretations drift. A 1,200 mm wide belt that arrives 8 mm narrow still bolts up, still runs, and still costs you a fortnight of edge damage four months later because nobody measured it at goods-in. Buyers who understand this treat conveyor belt manufacturer selection as a dimensional-control problem first and a materials problem second, and they pick their conveyor belt supplier on the strength of process records rather than on the brochure.

Send us your belt drawing and we will tell you which of these twelve checks your current supplier cannot pass

01Why Custom Belts Fail on the Tenth Order, Not the First

A bespoke belt is a set of drawings, a set of tooling and a set of process settings that exist only because your purchase order created them. Nothing about it is self-correcting. If the press operator on shift two decides the splice needs an extra ten minutes at 145 °C because the day feels humid, you get a joint that passes a thumbnail test and delaminates at month nine. If the edge gum is extruded at the top of its tolerance band on a Monday and at the bottom of it on a Friday, two belts from the same production order will track differently across the same conveyor rollers. On a 1,100 t/h sorting line we audited for a recycling client in early 2024, the buyer had three supposedly identical belts from one conveyor belt factory, built across eleven months, and the width spread between them was 9 mm with a splice-thickness spread of 1.8 mm.

None of those three belts was defective on the day it shipped.

That is the whole problem. A single-unit inspection tells you almost nothing, because the failure mode you are protecting against is variation, and variation only becomes visible when you compare records rather than parts. A supplier who can show you a splice log with press temperature, pressure, dwell time and operator initials for every joint is a fundamentally different risk from one who shows you a belt. The same logic applies to width, length, cover gauge and edge geometry, and it is why the audit below is built around evidence you can file, not around a factory tour.

Commercial structure matters here too. A trading house that buys from three unrelated workshops and resells under one brand cannot hold a tolerance band, because it never controls the press. Ask where the belt is actually built. When a vendor genuinely owns its lines, you can negotiate a documented tolerance and then hold them to it, and you can usually buy trial quantities through the same channel as production volume, which is what wholesale conveyor belts pricing should look like when the factory is also the seller. Split the two and you inherit a blame loop you will never win.

02Checks 1 and 2: Drawing Confirmation and Interface Data

Before anyone cuts fabric, the drawing has to be frozen and signed. We ask for a revision-controlled PDF with a revision letter, a date, a drawing number and a signature block, and we refuse to release tooling against a marked-up sketch or a photograph of a sample. Buyers who skip that step are, in effect, allowing the supplier to interpret the job, and interpretation is exactly what produces a belt that is 40 mm short with no legal recourse, because nobody ever agreed what the length was. The industrial conveyor belt you receive is only as good as the document the factory built it from, and a supplier who cannot produce that document on request is telling you something useful.

Interface data is the second half of the same check, and it is the half buyers forget. A belt does not exist in isolation; it exists over pulley diameters, under a specific trough angle, between skirt rubbers and against a take-up stroke. Give the supplier the drive pulley diameter, the tail diameter, the troughing angle, the minimum take-up travel, the loading height and the maximum lump size, and ask them to confirm back in writing that the carcass and cover grade suit those numbers. If they confirm without asking a single question, treat that as a red flag. On a mining & quarrying job last year the customer had specified a belt with a carcass far too stiff for a 400 mm snub pulley, and no one caught it until the splice started cracking at the pulley lapping line. A two-line confirmation email would have prevented a three-week shutdown.

Two traders sent us drawings last quarter that were identical except for one line. One drew the belt as 24 m of finished length with the splice included; the other drew 24 m of open length and added the splice afterwards. The difference is roughly 400 mm of belt, and on a fixed-centre conveyor that is the difference between a working line and a tensioner that has run out of stroke. This is why the rubber conveyor belt dimension has to be defined as either open or finished length, in words, on the face of the drawing.

03Checks 3 and 4: Tooling, Moulds and Splice Control

Tooling is where a bespoke order stops being a document and starts being a physical commitment, and it is the check most buyers never perform. Ask who owns the mould or the profile die, where it is stored, and what its condition was at the last shot. If the supplier cannot answer the ownership question, you are paying for tooling you will never be able to move, and that erodes your leverage on every subsequent order. For a belt with a moulded profile, the die condition sets the geometry of every cleat, rib or sidewall for the whole campaign, so a worn die quietly changes your belt while the paperwork still says the drawing is unchanged.

Splice control is the single most revealing thing about a custom belt supplier, because a joint is handmade work inside an otherwise industrial product. Ask for the splice procedure as a controlled document with press temperature, pressure, dwell time and cool-down under load, and then ask how the operator records what he actually did. The gap between the procedure and the log is the gap you are auditing. On a 1,400 mm wide chevron job we reviewed, the supplier's procedure called for a 200 mm bias step at 145 °C for 35 minutes, and the log for four joints showed three different dwell times between 28 and 41 minutes with no explanation. Every one of those joints passed the tap test on the day of installation.

Bias angle and step length deserve their own attention when the belt is narrow or the pulley is small, because a splice that is fine on a 1,200 mm head pulley can crack on a 500 mm snub. The supplier should be able to explain why the bias is what it is, not just quote a house default.

04Checks 5 to 8: The Four Dimensions That Decide Whether It Tracks

Width, length, gauge and edge geometry are the four numbers that make a custom belt either invisible or a permanent argument, and they are the four that a supplier is most tempted to leave loose. Take length first, because it is the one that cannot be corrected on site. A finished length tolerance of plus or minus 0.5 percent sounds generous until you apply it to a 120 m belt, where it becomes 600 mm and blows straight through your take-up stroke. The honest answer is that length should be quoted to a fixed figure with a tight tolerance, typically plus or minus 0.2 percent or better on a spliced belt, and the supplier should be measuring it on the floor with the belt under a stated tension rather than laying it out slack.

Width is easier to hold but easier to ignore. A belt that is narrow by 10 mm lets material spill under the skirt rubber on the return side and loses you a clean-up labour cost you will never trace back to the belt. A belt that is wide can foul the structure or the skirt board and will show its problem immediately, which is why the narrow error is the dangerous one.

Gauge and edge geometry are the pair most often treated as cosmetic. Cover gauge below nominal wears through to the carcass early, and a ragged or unsealed edge wicks moisture into the fabric plies, which is the beginning of ply separation. On a two-ply food packaging line running a white belt, we have seen an edge cut clean on the top cover but torn on the bottom, and the resulting fray cost more in rejected product contact than a complete resupply would have. That is the sort of thing you only find if you inspect the edge at goods-in with a light behind it.

Notice how the four checks sit together. Length and camber decide whether the belt wants to run straight at all; width and edge condition decide whether it survives the transfer points once it does. A supplier who can hold all four is almost always a supplier who owns the press, because you cannot hold camber through a splice if the joint is made by a third party on a subcontract basis.

The same discipline carries into the drive side of the order. Plenty of custom-belt projects also need drive and synchronous belts for the feeder and the take-away, and the buyers who interrogate the dimensional paperwork hardest are usually the ones who already source from a V-belt manufacturer who issues a profile chart with every delivery. Once you have seen matched pitch data on a drive belt, an unnumbered splice log on a conveyor belt is much harder to accept.

05Check 9: Sampling, First-Article Approval and Trial Runs

A custom order should never go straight to full production without a first article, and a first article should never be approved on the strength of a photograph. Ask for a physical sample of the belt section with the profile or splice included, measured and tagged against the drawing revision. The sample is your only chance to catch a misread dimension at a cost of a few hundred millimetres rather than a few hundred metres. We insist on a first-article report that carries the actual measured values, not a tick box, because the numbers are the only part a different engineer can check three years later. Buying from a transmission belt manufacturer who also supplies belt-driven equipment teaches the same lesson in miniature: the profile chart either matches the drawing or it does not, and no amount of reassurance substitutes for the chart.

If the belt drives a critical line, push for a trial length. A 10 m trial run on the real conveyor for two shifts will expose splice thickness problems, edge wander and cover scuffing that no bench sample can reveal. Budget it as a line item rather than treating it as a favour. The cost of a trial is trivial next to the cost of a planned shutdown that turns into an unplanned one, and any serious conveyor belt distributor or factory-direct seller will support it because they want the same answer you do.

One practical caution. A supplier who offers a free sample and then refuses to issue a first-article report is not saving you effort, they are removing the baseline you will need when the second order arrives. The report is the product as much as the belt is.

06Check 10: Material Traceability and Compound Documentation

Traceability answers a single question with a clean yes or no. Given a belt serial number, can the maker tell you which fabric lot, which compound batch and which press cycle produced it? If the answer requires two weeks of internal detective work, the supplier does not have traceability, they have a filing cabinet. Ask for the chain on one sample order and then test it six months later on a repeat, because traceability degrades quietly as staff turn over and the person who wrote the traveller sheet leaves.

Compound documentation deserves the same scepticism. A data sheet that quotes a grade without a number is marketing. A heat grade should be tied to a stated continuous service temperature, an abrasion grade to a measured volume loss figure under a named drum test, and an oil grade to the swelling result in the specific medium you are handling. For a timing belts line running through a wash tunnel, the relevant question is not whether the compound is 'washable' but what the supplier measured and how the belt behaved after 200 wash cycles. Numbers travel; adjectives do not.

Where a bespoke belt uses more than one material system, the interaction is where documentation usually breaks down. A cover compound may be perfectly documented while the skim between plies is a generic stock with no lot record, and that skim is the layer that decides whether the belt delaminates in service. Ask separately for the cover compound, the skim or cushion compound and the carcass, and be suspicious if only one of the three has a certificate. This is also where a maker's habits show: a factory that runs heat resistant conveyor belt and abrasion resistant conveyor belt builds on separate line records will have the paperwork discipline already, because they cannot run a heat grade through a press that was last used for general purpose stock without documenting the changeover.

We ran that last test on a supplier last spring. It took them eleven days to find the traveller for a belt delivered in the previous quarter, which is roughly ten days longer than a recall decision allows.

07Check 11: Batch Consistency Across Repeat Orders

Consistency is not a virtue a factory can promise, it is a property that shows up in data. When you place a repeat order, ask for the first-article report of the previous run and a comparison sheet for the new one, side by side, dimension by dimension. Any supplier who tracks their own numbers will produce that in an afternoon. Any supplier who does not will send you a reassurance email instead, and that email is your answer.

The practical reason consistency matters more for bespoke work than for catalogue work is that a bespoke belt is dimensioned to a specific conveyor. A catalogue belt that drifts by a few millimetres is somebody else's problem to absorb. A custom belt that drifts by the same amount becomes your problem, because your structure, your skirt clearance and your take-up position were all set to the first article. We saw this on a port bulk material handling contract where the second batch of belts came in with a 6 mm width reduction and a cover gauge 0.4 mm down. Individually, both numbers were inside the tolerance that the purchase order actually stated, which is precisely why the buyer had no claim and had to buy a second set of skirt rubbers to match.

The lesson is that your purchase order wording decides whether drift is a defect or a variation. Write the tolerance into the order, state the measurement method, and state the reference tension. Then a 6 mm change is a non-conformance rather than a discussion.

08Check 12: Marking, Packing and the Twelve-Point Scorecard

The final check looks trivial and is where the audit either becomes usable or stays theoretical. A belt that arrives without a serial number, without a length tag and without the drawing revision on the label cannot be matched to any record, which means everything you gathered in checks one to eleven evaporates at goods-in. Ask for a marking standard: belt serial stamped or tagged at both edges, finished length and width on the tag, drawing revision on the packing list, and a statement of the roll or fold direction so the crew fits it the right way round.

Packing follows the same logic. A long custom belt that is folded rather than rolled develops creases that never flatten, and a crease in a moulded profile belt shows up as a tracking fault from the first shift. Ask how the profile is protected, whether the belt is coiled on a core, and whether the covers are separated by a liner. Those answers are cheap to give and expensive to retrofit.

Score a supplier on those twelve lines and you will normally find that they cluster. A factory with a good splice log almost always has good length control, and a trader with no drawing control almost never has traceability, because both are symptoms of the same thing. Anything below 75 out of 100 is worth a second audit before you place a repeat order, and anything below 60 should not be given a first one, however attractive the quotation looks. For the wider reliability context around this audit, our quality assurance page sets out the inspection stages we run, and the service team can attend a first installation so that the trial data is recorded by someone who is not selling the belt.

Send your drawing for a free twelve-point supplier audit review

09Frequently Asked Questions

How long should a first-article approval take on a bespoke belt order?

Plan on three to five weeks from drawing release to a signed release note, and treat anything shorter with modest suspicion. The section sample itself usually needs one to two weeks because the sample has to be cured before it can be measured, the splice sample adds three to five working days, and the first-article report takes another two or three days to prepare and countersign. Where a supplier promises approval inside a week, they have either skipped the splice sample or they are measuring against nominal values rather than taking real readings. Both shortcuts remove the baseline you will need when the repeat order arrives.

Which dimension should I write into the purchase order first if I can only tighten one?

Finished length. It is the only dimension that cannot be corrected after delivery without making a second splice, and it is the one that eats your take-up stroke on a fixed-centre conveyor. A short belt slips under load because the tensioner runs out of travel, and a long one leaves you with slack you cannot remove. Quote the length as either open or finished, state the measurement tension, and set a tolerance of plus or minus 0.2 percent or tighter. Width, gauge and camber matter enormously in service, but all three can be argued about on site, while length simply decides whether the belt can be installed at all.

Do I need a different audit if the belt carries a moulded profile or a sidewall?

The twelve checks stay the same, but two of them get heavier. Tooling control moves to the front, because the die cavity decides the geometry of every cleat, rib or sidewall belt skirt for the whole campaign, and a worn cavity can pass a tape measurement while still being out of geometry once the material shrinks. Edge and sealing checks also tighten, since a profiled belt with an unsealed edge under a heavy load will lift its cover at the trough transition. For a steep incline, the same logic applies to a chevron belt, where profile pitch and height drift are the two numbers that turn a working incline into a slipping one.

How do I check a supplier's records when I have already bought from them for years?

Give them one serial number from a belt they delivered at least six months ago and ask for the full record pack: traveller sheet, splice log, cover gauge readings, mill certificate for the carcass and the finished length measurement. Measure the response in days, not in promises. A supplier with genuine traceability will produce that pack within a working day, because the serial resolves to a file. A supplier who takes a week is telling you that the records exist somewhere but are not organised, and the difference matters enormously the first time you need to argue a warranty claim or trace a contamination incident back to a specific production run.

What is the most common reason a second batch behaves differently from the first?

Crew and compound change, in that order. If a different press operator made the joints, the dwell and cool-down will differ even when the procedure is unchanged, and the thickness rise across the joint moves with them. If a different mixing batch was used, hardness can shift by two or three Shore A points, which quietly changes the wear rate and the belt's grip at the drive pulley. Whenever you place a repeat order, ask who will press it and which compound batch is allocated, then compare the answers with the previous order's records. Neither question costs the supplier anything to answer honestly, which is exactly why a vague reply is informative. Buyers running a hot or oil-mist duty should also re-check that the grade still matches, since a heat resistant conveyor belt and an oil resistant conveyor belt cannot be substituted for one another on price alone.

Should the supplier handle the changeover to a new belt as well as the manufacture?

It is worth paying for. An installation crew that already knows the belt's camber, its coil direction and its splice position will record the tracking data you need for the next order, and they will notice an edge fault while it is still a warranty issue rather than a maintenance job. This is where the wider hardware matters as well, because a belt change is the natural moment to inspect conveyor pulleys, return idler sets and impact idler stations against their original signature. If a supplier delivers belts but cannot supply the conveyor components that sit around them, you carry the interface risk between two vendors. For bulk terminals and processing plants where downtime is charged by the hour, that risk usually costs more than the belt itself, and a single source across belt, rollers and structure is the cleaner commercial position.

If your operation sits inside a distribution centre rather than a process plant, the same twelve checks apply but the priorities shift toward cleanliness and tracking, and our notes on logistics & warehousing conveyors set out what to add. In heavy industry the balance moves again, toward impact resistance and carcass strength, which is why cement and mining & quarrying audits read differently from a packing-hall audit even when the form is identical. Whichever duty you are in, the sequence is unchanged: freeze the drawing, control the tooling, log every splice, measure four dimensions, prove the materials, and compare the batches. Do those things and a custom belt stops being a gamble on a supplier's memory.