What Every Gas Water Heater Should Be Tested for Before It Ships — and How to Verify It in an Audit
Factory audits fail for a boring reason: they are conducted against a checklist the factory has seen before. The equipment is on, the floor is clean, the certificates are laminated, and nothing you are shown connects to the question you actually care about — will the hundred units in this container all work when they arrive?
That question has a narrower answer than most audit templates suggest. For a gas appliance it comes down to five tests that should be performed on each unit, plus a paper trail that proves they were performed. This article lists the five, explains why sampling cannot replace them, and describes how to verify the whole thing in a single visit or a single video call.
The Five Per-Unit Tests
1. Gas tightness, per unit, under pressure. Every unit has gas-carrying joints: the inlet connection, the valve body, the manifold, the burner feed. Each of those is a possible leak path, and a leak in a gas appliance is not a cosmetic defect. The test is a pressure-hold on the gas circuit with the burner off, held long enough to reveal a slow leak, with the result recorded against the unit's serial number. This is the single most important test in the list, and the one most often reduced to a spot check.
2. Ignition and flame stability. The unit is fired, the ignition sequence is observed, and the flame is checked for lift, yellowing, or instability across the modulation range. A unit that ignites on the bench but not on a cold morning at 4 °C has not really been tested; the test should include a cold-start condition if the market requires it.
3. Temperature rise and flow, measured rather than assumed. The unit is run at a known inlet temperature and a known flow, and the outlet temperature is measured. This is what confirms that the rated figures on the spec sheet are the figures the machine actually produces. A factory that tests only for "it gets hot" is not testing the specification.
4. Water-circuit sealing. Cold inlet, hot outlet, drain, and any internal joints are checked under operating pressure. This is a separate test from gas tightness and is frequently skipped because a water leak is merely annoying, while a gas leak is dangerous. Both matter to the buyer who receives a wet carton.
5. Safety interlocks and shutdown behaviour. Flame-failure protection, over-temperature cut-out, and — on units with automatic drain — the freeze-protection trigger. These are tested by inducing the fault condition, not by reading the datasheet. Ask what the factory does to trigger the flame-failure device, and listen for whether the answer describes a procedure or a paragraph.
| Test | What is measured | Record it should leave behind |
|---|---|---|
| Gas tightness | Pressure hold on the gas circuit, burner off | Serial number, pressure, hold time, result |
| Ignition and flame | Ignition sequence, flame stability, cold start | Pass or fail with the observed symptom |
| Temperature rise and flow | Inlet temperature, flow rate, outlet temperature | Three measured values per unit |
| Water-circuit sealing | Operating pressure on inlet, outlet and drain | Leak check result at each joint |
| Safety interlocks | Flame-failure, over-temperature, freeze trigger | Fault induced, response time, reset result |
Why Sampling Cannot Replace a Per-Unit Gas Test
Statistical sampling is the right tool for cosmetic and functional defects in a lot. It is the wrong tool for gas tightness, because the cost of a single escape is not measured in defect rate. A 2.5 % AQL on a lot of five hundred units permits a surprising number of defects, and no distributor wants to explain to an end customer why one of them was a gas leak.
The distinction is worth stating to a supplier in exactly these terms: sampling is for attributes, per-unit testing is for safety functions. A supplier who proposes an AQL plan as a substitute for per-unit leak testing has told you something about how they think.
Where sampling does apply — appearance, labelling, packaging, accessories, and functional checks other than safety — ask for the plan in writing: the standard referenced, the inspection level, the sample size, and the acceptance number. The widely used framework is ISO 2859-1, and the common levels in this trade are 1.0 and 2.5. If a supplier cannot name the standard and the level, the plan is improvised.
A pressure-hold test, with the gauge and the record sheet side by side: the rig shows the test can be run, the filled-in sheet keyed to serial numbers shows it was.
Three Pieces of Evidence That Prove the Tests Happen
Equipment proves capability. Records prove use. Ask for these three.
Test records tied to serial numbers. A sheet per unit, or a digital log keyed to serial, with the measured values rather than tick marks. If the record shows a pressure figure, a temperature figure and a flow figure per unit, the tests are real. If it shows three checkboxes, it is a form.
A quarantine area with actual parts in it. Incoming material that fails inspection and finished units that fail testing both have to go somewhere that is not the production line. An empty, clean, labelled reject area is a warning sign, not a positive one — it means either nothing fails or failures are being moved somewhere less visible.
A rework record. Rework is universal. What matters is whether it is counted. A rework log with reasons and quantities tells you the real first-pass yield; a plant that claims no rework has either perfect processes or no measurement.
Designing a Remote Audit That Is Hard to Stage
If you cannot travel, a video call is workable, but only if you control the route. Four rules make it useful:
- Do not accept a pre-recorded walk. Insist on live, and ask the person holding the phone to turn around and walk somewhere you choose.
- Ask for a specific unit, then follow it. "Show me unit number whatever is on that pallet, and open it" defeats staged display areas.
- Interrupt the route. Mid-walk, ask to see the reject area, the rework bench, then the gas test station in that order. Staging three locations in an unscripted order is expensive.
- Ask a process question the guide cannot anticipate. "What is the hold time on the leak test, and what pressure?" A scripted guide has a scripted answer; a production engineer has a number.
What to Ask Before You Book the Flight
Six questions, sent in writing, that tell you whether the visit is worth making:
- Is gas tightness tested on every unit, or sampled? What is the pressure and hold time?
- Is each unit fired before packing, and is a cold-start condition included?
- Is the outlet temperature and flow measured per unit, or only on a reference unit?
- What is the inspection standard and AQL level for the non-safety attributes?
- Can you provide a redacted test record for a recent lot, keyed to serial numbers?
- What is the rework rate, and how is it recorded?
Practical Pitfalls
- Do not accept equipment as evidence. A leak-test station with no records is furniture.
- Do not let a sampling plan cover gas tightness. Safety functions are tested per unit or they are not tested.
- Do not audit a line that is not running. A stopped line answers every question correctly.
- Do not confuse a certificate with a test. Certification covers a design; per-unit testing covers the unit in your container.
- Do not skip the drain and freeze-protection check on RV units that have it. The feature is only real if the discharge path is clear.
- Combustion appliances produce carbon monoxide, which is colorless, odorless and toxic. Installation and operation must follow the unit's manual, and portable outdoor units must never be used in tents, vehicles or other enclosed spaces.
- SEFFY heaters are designed to international gas-safety standards; we are actively preparing for EU CE certification, expected soon.
