I Changed a Candle Raw Material After Customer Complaints — How Do I Confirm the New Version Fixes the Original Problem Without Creating Another One?

Why a complaint fix needs two tests

Complaint-driven changes are the most dangerous kind, for a simple reason: they are made under pressure and judged by one symptom. Customers say the candle tunnels, you change something, the next candle you light does not tunnel, and the change goes into production. What nobody checked is whether the new version now runs a taller flame, soots the glass, heats the jar, burns through faster or fails in its last few centimetres. The original complaint goes quiet; a different complaint turns up three weeks later from a different customer, and it looks unrelated.

So the test for a complaint fix has two lanes. Lane 1 proves the fix: the old version must reproduce the complaint under controlled conditions, and the new version, under the same conditions, must not. Lane 2 screens for new failures: every other attribute of the new version must be no worse than the old one, which acts as the control. Only a change that passes both lanes is a fix. A change that passes Lane 1 alone is a trade.

Two notes on the table. First, the complaint wording is your customers', not a diagnosis. “Can't smell it” could be a throw problem, a tunnelling problem, or a customer who has got used to the scent; the failure mode column is your hypothesis, and Lane 1 tests it. Second, the Lane 2 column is general candle-making practice about what a wick change commonly affects, phrased as risks, not predictions. CSI does not publish burn data for Eco Thin C1 or Thick C2 and never describes one wick as equivalent to, or a size step from, another — the names are listing names. The one-size-up post covers that move in detail, and the better-throw-hotter-jar post covers the trade you must refuse.

Lane 1: reproduce the complaint before you fix it

Most makers test the new version only. The better order is the reverse. Burn the old version first, under the conditions your customers use, and confirm it fails in the way they described. If it does, you have a reproducible failure and a fair comparison. If it does not, stop: the cause may be a bad lot, a jar that measures differently from your record, a customer burning in a draught, or burn habits the candle was never designed for — and changing the wick will not fix any of those.

That last possibility is common enough to design for. In general candle-making practice, short burns on a container candle are a well-known cause of tunnelling, because the melt pool never reaches the edge and each later burn follows the narrower pool. So Lane 1 runs two burn patterns: the commonly used working cycle of about 3–4 hours with full cool-down, and a customer pattern that copies how the complainants actually burned — for example, one-hour burns if that is what your complaint log suggests. Both patterns are labelled working practices; set the customer pattern from your own log.

Lane 2: screen for the failures you just created

Lane 2 runs on the same candles as Lane 1, so it costs no extra material — it only costs attention. On every burn, record the lines below for both versions. The pass rule is comparative on every safety line because CSI publishes no numeric flame-height or jar-temperature limit: the new version must be no worse than the old on flame, soot and jar heat, and within your jar supplier's guidance and the recognised candle fire-safety standard you work to (ASTM F2417 is a commonly cited one). On the non-safety lines — consumption, throw, appearance — a difference is not automatically a failure, but it must be recorded and decided on.

The consumption line matters more than it looks. If the new version burns more grams per hour, the same 200 g candle lasts fewer hours, and a customer who never complained about tunnelling may now complain that the candle “finished too fast”. Burn life in hours is roughly the burnable wax divided by grams per hour, so measure it on both versions and put the two figures on the product decision. If you state a burn time on your label, this is the test that decides whether it changes. The wick retest checklist covers consumption rate as a test in its own right, and the wick-only change post explains which storage results a wick change leaves intact.

What the wick change leaves valid

Not everything needs retesting. The retest matrix below is for this example — a wick change from Eco Thin C1 to Eco Thick C2 on an unchanged wax, oil and jar — in the series vocabulary. It tells you which Lane 2 checks can be light and which must be full.

Two NEW lines appear, and both come from the complaint. The approval of the old version never tested burn life because nothing had changed it, and never tested the customer pattern because nobody knew it mattered. Complaints are how a test file learns what it was missing — which is why the change-control form asks for the complaint reference as the reason for the change.

The priced two-lane kit

Twelve candles — two arms of three old and three new — at 200 g and 8% need 192 g of oil and 2,208 g of wax, 201.6 g and 2,318.4 g with a 5% pour-loss allowance (a labelled working assumption). If you still have unburned candles from the complaint batch, use them as the old-version arm and the oil and wax drop by half.

Compare that with what the complaints are already costing. As an example only — replace every figure with yours — 14 complaints on 600 candles sold is 2.33%, or 2.33 per 100. If each is settled with a replacement candle (₹259.79 of materials at listed prices) and a courier charge of ₹80.00 (example figure; use your courier's quote), that is ₹339.79 per complaint and ₹4,757.10 across the 14, before your time and the customers who never came back. The wick change itself costs ₹3.54 per candle, or ₹2,124.00 over the next 600 candles.

The decision rule, and what happens after release

Release is not the end. A test on twelve candles proves the mechanism; it cannot prove the complaint rate has fallen, because twelve candles are far too few to measure a rate of 2.33 per 100. That proof comes from the market. Log complaints on the new version against units sold, and compare over the same number of units as the original window — here, the next 600. If the tunnelling complaints stop and no new category appears, close the change. If a new category appears, treat it as a new complaint and start Lane 1 again. The full change-control system calls this post-release monitoring and gives it a named owner.

For complaints that follow a change you already made — rather than a change you are making in response — the investigation runs the other way: the two-changes-at-once post shows how to separate causes, and the discontinuation file shows how to record the outcome so the next person does not repeat it. If the complaint is about scent rather than burn, the specification post is the place to check that the oil on the shelf is the one the candle was approved with.

Frequently asked questions