How Much Glass-Adhesion Variation Is Normal in Soy Candles Before It Becomes a Commercial Quality Problem?

How much variation is normal? The honest answer

No published standard says how many soy candles in a batch may show wet spots, or how large a spot may be, before the batch is a quality problem. CSI publishes no adhesion data for any wax or jar, and there is no recognised industry threshold to quote. That is not a gap in our research; it is how cosmetic quality works for a handmade product. "Normal" for your candle is the variation your current process produces when nothing has gone wrong — and the only way to know it is to measure it once, carefully, so that you can recognise when something has changed.

So the answer comes in two parts. First, measure your baseline: pour 20 jars in one session, exactly as you normally do, number them, photograph each from four sides at fixed checkpoints and grade them on a house scale. The share of jars at each grade is your normal. Second, decide your tiers: which grades sell at full price, which become seconds, and which mean the process has drifted. That is a commercial decision about your brand and price point, not a technical fact about wax. Variation becomes a commercial quality problem when either the share of graded-out jars costs you more than fixing the cause would, or a batch moves outside your own baseline by more than chance can explain. Both tests are arithmetic, and both are on this page.

If you have not yet established whether your spotted jars burn the same as clean ones, the paired burn comes first: a spot that affects burning is a fault at any frequency. This page assumes the spots are cosmetic and asks how many of them your business can carry. The section hub routes every other wet-spot question.

Measuring your baseline: the 20-jar batch

A baseline is only useful if it is your normal process, unimproved. Do not clean the jars more carefully than usual, do not move the cooling board, do not pick a cool evening. Pour exactly as a production day pours. The example uses CSI's Clear 150 g jar — the one clear CSI jar with a published fill, approximately 150 g, "safe for hot pour up to 85°C" on its page — with Natural Soy Pearl Wax. The Soy Pearl page publishes no pour temperature, melt point or fragrance ceiling, so follow the wax maker's instructions (ask CSI on WhatsApp), keep the pour at or below the jar's stated 85°C, and record the figure you used. At 8% a candle is 12 g of Japanese Cherry Blossom and 138 g of wax. Eight per cent is a labelled working load; with no ceiling published for Soy Pearl, check the page or ask before going higher.

How many jars before a share means anything

This is the part most cosmetic QC skips, and it is pure arithmetic — no wax data involved. If you grade a small number of jars and none of them shows a spot, you have not shown the true share is zero; you have shown it is probably below some limit that depends only on how many jars you looked at. The table gives the 95% upper bound for zero spotted jars in a sample of n, worked from 1 − 0.05^(1/n). The quick approximation 3 ÷ n (the "rule of three") is shown beside it.

Read the first and last rows together. Six clean jars still allow a true share up to 39.3%. A hundred clean jars bring it down to 3.0%. At ₹217.03 of material per candle (production rates, below), getting from a rough idea to a tight one costs real money — which is why the baseline batch is 20 jars, a compromise that bounds a clean result at 13.9%, and why production batches, which you are pouring anyway, are the cheapest place to keep counting. How many jars per wax in a first comparison applies the same arithmetic to choosing between waxes.

The reverse question matters just as much: if spotting is real, how likely is a small check to catch it? The chance that a sample of n jars contains at least one spotted jar, when the true share is p, is 1 − (1 − p)^n.

A five-jar spot check misses a 5% problem more often than it catches it (77.4% of the time it shows nothing). That is the arithmetic behind a familiar complaint — "it was fine when I checked". If your tiers matter to you, check enough jars per batch to see the shares you care about, or grade every jar; at a few seconds per jar through clear glass, grading every jar is often the cheaper option.

When a batch has drifted: an alarm limit from your baseline

Once you have a baseline share, you need a way to tell a genuinely worse batch from ordinary luck. A long-established method from general quality control is the p-chart: plot each batch's share of jars at grade 2 and above, and set an upper alarm limit at the baseline share plus three standard errors, p + 3 × √(p(1 − p) ÷ n), where n is the number of jars graded. It is a general statistical tool, not a candle standard, and it only works if every batch is graded the same way.

The table shows why small batches are noisy. With a 10% baseline, a 20-jar batch has to reach 30.1% before it is clearly worse than normal; a 100-jar batch only 19.0%. Below the limit, a bad-looking batch is probably your normal variation and the right action is to keep grading. Above it, something changed — the season, the jar lot, the wax lot, who poured, where the jars cooled — and it is time to look. Monitoring future wax and jar lots builds this into incoming-material checks, and seasonal versus raw-material variation helps separate the workshop from the materials.

When variation becomes a commercial problem: the cost

The second test is money. A graded-out jar costs you either the price difference between full price and a second, or its whole material cost if it is not sold at all. Here the production-rate material cost of one candle is ₹217.03: Natural Soy Pearl Wax at the 5 kg rate (₹424.80/kg — Soy Pearl is the one CSI soy wax with a real bulk break, from ₹450.00/kg at 1 kg), Japanese Cherry Blossom at the 1 kg rate (₹4.84/g), an Eco Thin C1 wick at ₹5.90 as the example wick line, and the Clear 150 g jar at ₹94.40. It excludes labour, labels, packaging, freight and GST.

The second column is cost per visually acceptable candle — total material divided by the candles that sell at full price, ₹217.03 ÷ (1 − share). Notice how gently it climbs: even at 20% graded out the material per full-price candle is ₹271.28. Material is rarely where wet spots hurt. The third column is: at the example prices, every 10 percentage points of seconds gives away ₹1,500.00 per 100 candles. That is the number to set against the cost of fixing the cause. If a process change — cleaning jars, moving the cooling board — costs less than that, the variation is a commercial problem worth solving. If the fix costs more, your current share may be the sensible normal for your price point. The cost of cosmetic rejects and comparing waxes by cost per visually acceptable candle take this further.

The priced baseline batch

Twenty jars at 150 g, Soy Pearl, Japanese Cherry Blossom at 8%, one Eco Thin C1 wick per jar plus two spare, and a 5% pour-loss working allowance on wax and oil. Wax per jar 138 g; oil 12 g. The Clear 150 g jar's pack of 20 is in limited stock, so the kit uses two packs of 10 — the same price per jar.

Leftovers: 102 g of wax and 13 g of oil. The wick is an example line — CSI does not publish a wick size for this jar, so use the wick that already passes in your candle. Every jar in a baseline batch is a saleable candle if it grades well; the batch is not a cost so much as a production run you happened to measure. Non-CSI items are examples, not prices: a printed grid or marked acetate sleeve, and a plain light background for photos.

The decision rule

Stop at the first line that is true.

The tiers themselves deserve a yearly look. A brand moving up in price — from an example ₹499 to ₹1,199 — usually tightens what counts as full price, and the same baseline then produces more seconds. That is a business decision, not a change in the wax. What share is acceptable in a luxury pilot covers setting tiers for a premium line, and customers who think spots are oil covers what a grade-2 jar looks like from the other side of the counter.

The CSI products in this baseline

Frequently asked questions