I’m Launching Three Jar Sizes in Eight Fragrances — How Do I Organise Wick Testing so I Can Identify Jar Effects Separately from Fragrance Effects?

Why a grid of convenience can't separate them

Most launch tests grow by convenience: the oils you have most of go in the jars you have most of. Then a result arrives that you can't read. Vanilla Bean needed a bigger wick in the 10 cm jar than Lavender did in the 7 cm jar. Was that the jar, the vanilla, or both? With those two cells alone, the question has no answer, because the jar and the oil changed together. Statisticians call it confounding; candle makers call it "the wick chart we made that nobody trusts".

The full grid avoids confounding by brute force: bracket every oil in every jar. With three jars and eight oils, at CSI's brackets (C1 v C2 at 6 and 7 cm; Cotton Braided v 2 × C1 v 2 × C2 at 10 cm) and two candles a cell, that is 112 candles and 32,000 g of candle, three-quarters of it in the 10 cm jar. It tells you everything, including a lot you could have predicted.

The anchor design gets the separation for far less, because it only needs each effect measured once, cleanly, and then checks the combinations. It is the same idea as the five-jar matrix in the wick-matrix hub, turned ninety degrees: the matrix varies jars with the oil fixed; the anchor column adds the other axis. With twenty oils rather than eight, sort them into groups first (the grouping post) and anchor on one oil per group.

The three jars, from their pages

The worked example uses three CSI listings, because CSI genuinely lists jars at about 6, 7 and 10 cm. None of CSI's 72 container listings states an internal diameter, so each figure below is read as the outside, with a 0.35 cm wall assumed. Measure your own jars at the rim, middle and near the base before you trust any of it.

Two things follow. First, the jars are not evenly spaced: the 7 cm jar has 41% more pool than the 6 cm, but the 10 cm has 2.2 times the 7 cm's, so the jar effect between 7 and 10 cm is a change of configuration (single to pair), not just a step in size. Each wick of a pair at 10 cm serves 33.96 cm², close to the whole 7 cm pool, which is why the pair arms use the same wicks as the 7 cm bracket (the 10 cm pair-or-single hub). Second, the 85°C wording on two of the pages is a pour limit, not a burn rating: pour CSI 464 at 80–85°C in those, and at the low end of its range with pre-warmed jars in the 6 cm, which states nothing (our judgement). Never judge glass heat by touch.

One row, one column, one control

Hold wax (CSI 464), load (8% of total weight), cure, pour temperature and test room constant for every candle. Then pour three kinds of cell:

The order matters for the calendar too. The anchors are 44 candles and can all be poured on day one; the confirmations wait for the anchors' screening, so plan on two pours a couple of weeks apart. Identical test candles (same wax lot, same weighing, same cure) are the foundation; the identical-test-candle post covers making them.

Reading the anchors: two effects and a prediction

Express every result as a step on the jar's own ladder. At 6 and 7 cm the ladder is C1 → C2 → Cotton Braided. At 10 cm, for the pair arms, it is 2 × C1 → 2 × C2, with the Cotton Braided single read alongside. A step is a move of one rung; it is a bookkeeping device, not a claim that rungs are evenly spaced in heat.

The jar effect (the Lavender row)

Illustrative result: Lavender passes on C1 and is over on C2 in both the 6 cm and the 7 cm jars, and passes on 2 × C1 in the 10 cm jar (Cotton Braided under, 2 × C2 over). So between 6 and 7 cm the jar effect is zero steps, and between 7 and 10 cm it is a change of configuration, C1 to 2 × C1.

The fragrance effect (the 7 cm column)

Four oils sit with Lavender (step 0) and four sit one rung up (step +1). The +1 group is unsurprising for Roasted Coffee, Vanilla Bean ("Vanillin Content: 2.5%" on its page) and Woody Oud, because heavier, vanillin-rich and resinous oils are commonly linked to a need for a larger wick. Bergamot Amber is the one to notice: its family is "citrus amber", but its stated notes end in "woody musk and smooth vanilla base". The column caught that; a grouping by name would not have.

Confirmations, and catching interactions

Each of the fourteen off-anchor cells gets two candles on its predicted arm, screened at burn 3 and burned to about 1 cm left. Three outcomes:

In the worked example, thirteen of fourteen confirm. The fourteenth, Bergamot Amber in the 6 cm jar, runs over on its predicted C2: a tall flame and a pool past 1 cm. On C1 (two more candles) it passes. Its +1 step was real in a 7 cm pool and not in a 6 cm one. That is an interaction, and it costs 2 candles and a little more oil: Bergamot Amber's 150 g is 10 g short after the extra cell, so it needs another 50 g bottle. Every confirmation that holds, on the other hand, is a cell you did not have to bracket. Keep the anchors on file: if a jar later changes by a few millimetres, the anchor row is what you re-run first (the 5 mm jar-change post).

Candles, grams and rupees against the full grid

Cash outlay, including jars (because every test candle uses one), at CSI live prices on 29 September 2026. Costs exclude labour, freight, packaging and GST.

The anchor design saves 38 candles (34%), 47% of the wax and oil, and ₹24,842.68 of cash in the worked example. Most of the saving is in the 10 cm column, where the full grid would bracket every oil with three arms of 500 g candles in ₹212.40 jars. Each oil other than Lavender needs only 144 g, so a 100 g plus a 50 g (6 g spare) replaces the 500 g bottle the full grid needs.

What to buy, and what to skip

Five items, chosen on page statements, price and stock on 29 September 2026; CSI publishes no wick-performance data, so nothing is ranked on burn results. Skip for now: the Apothecary 700 gm variant (₹250.00, the same stated dimensions, not needed at a 500 g fill), 500 g bottles of the seven non-reference oils, and any wick pack above 50.

Frequently asked questions