My Customer Says the Candle Tunnels but My Retained Sample Burns Perfectly — Is It the Wick or the Way They Burned It?

What the retained sample proves — and what it does not

Be exact about the claim your sample supports. You lit a jar from the same batch, it reached the glass, the surface stayed level, and you concluded the batch is sound. What you demonstrated is narrower: this formula, in this vessel, with this wick, is capable of a full melt pool under your burn conditions. Capability — not compliance, not uniformity, and nothing about the room the complaint came from.

Three ordinary gaps sit between your jar and theirs. The unit is not the same unit: a pour run is not homogeneous — first jar and last jar come off the pitcher at different temperatures, a wick can sit a millimetre off centre on one unit and not another, a sticker can lift. The jar is not necessarily the same jar: glass arrives in cartons from different production runs, so unless you measure each one, "the same jar" means "the same SKU". And the burn is definitely not the same burn: yours happened in your workspace, at a length you chose, wick trimmed by reflex. Theirs happened somewhere you have never been.

The symmetry runs both ways. A maker who always burns three or four hours at a stretch will find that an under-wicked candle looks perfectly acceptable, because a long session has time to widen the pool to the wall even from a slow start. Your generosity as a test burner can hide the exact fault you are trying to rule out — which is why the test below has two arms.

Why an adequately wicked candle still tunnels in someone else's room

Tunnelling is not an event, it is a path a candle gets locked onto. The mechanism is worth knowing properly, because it lets you reason about the customer's answers rather than guess.

Two things follow. A candle can be adequately wicked and still tunnel in use, if nobody gives the first session time to reach the wall. And a candle can be under-wicked and still look faultless to its maker, if the maker always burns long. Those are one fact seen from two ends, and together they are why a single perfect retained sample cannot settle a tunnelling complaint.

The five questions, and how to word them

Every one of these can be heard as an accusation, and a customer who feels accused stops describing and starts defending. Two framings fix most of it. Put the burden on your product, not their behaviour — you are gathering information, not building a case. And offer the replacement before you ask anything, so the questions are clearly not a test they must pass to get one. That reordering improves the answers more than any phrasing trick.

Read the two photographs together. The lit overhead shot shows where the flame sits relative to centre, whether the pool is round or crescent-shaped, and whether the flame is down inside a well. The cold shot shows the finished geometry without the flame in the way: the width of the unmelted ring, whether it is even all round, and whether the surface is level or cratered. A narrow concentric column with an even ring reads as a short-session history; a crescent pool reads as airflow; a deep narrow well after two or three lightings is the pattern that should make you take the wick question seriously.

The two-arm reproduction test

The design principle is one sentence: burn the candle their way, not your way. Every maker's instinct is to relight a retained sample and watch it behave — which reproduces the maker's conditions and therefore the maker's result. What you do not know is what it does in forty-minute stretches on a windowsill with the fan on.

Take six retained jars from the complained-of batch, eight if you have them, and split them into two arms of three. Hold the batch and the wick constant — same pour, same wick lot, same jar SKU — and change only the burn pattern. Vary the wick too and you have destroyed the experiment before lighting it: two variables moving together tell you nothing about either.

Weighing matters more than it looks. A cold mass before and after every session, on a ₹354 weighing scale, turns an argument about appearance into numbers you can compare between arms — and it survives you forgetting what the jar looked like three weeks later.

Retained samples, traceability and the two free checks

All of the above assumes you kept samples. If you did not, the first corrective action from this complaint is not a wick change — it is starting to keep them, from the next batch you pour. Without them a complaint is one person's memory against your own, and no reasoning closes that gap.

A batch code is only as good as what it connects to. If your records turn it into nothing but a date, you cannot investigate — only guess more confidently. A batch number must resolve internally to all nine of these.

Now the two checks that cost nothing and should come before you spend a rupee: both are documented constraints and both produce burn complaints.

What the test costs, against what re-pouring costs

The case for gathering evidence is arithmetic. Figures are CSI live pricing, September 2026, on a 150 g soy container candle at a 10% load — 135 g wax, 15 g oil — in a jar that publishes its fill capacity, which most jar families do not. Where a jar publishes none, pour one and weigh it: "200 ml" is a volume, not a 200 g fill, and a 200 ml salsa jar at ₹29.50 each in packs of 50 or 100 takes whatever mass it takes.

So the test costs about 4.6% of a re-pour that reuses the glass, and about 2.6% of one that does not — and that is the expensive version, where you keep no retained samples and must pour a set specially. If you already keep retained samples it costs ₹0 in materials and about a week of evenings. Against a decision that could commit ₹38,186.70 of stock to the bin, there is no version of this arithmetic in which guessing wins.

Now the part against our own interest, because it is true. The most likely outcome of this investigation is that you buy nothing. Most single tunnelling complaints resolve into a short first burn, a draughty position, or a jar from the tail of a pour run, and none of those is fixed by a purchase order. We would rather sell you one roll of care stickers at ₹413 and keep you for five years than sell you 500 wicks you did not need.

Frequently asked questions