I Have Five Jar Diameters and 20 Fragrances — Do I Really Need 100 Separate Wick Formulations or Can I Intelligently Group Similar Candle Systems for Initial Testing?

Why 100 formulations is the wrong unit of work

Take the five-jar matrix from the wick-matrix hub: eleven bracket cells across 6, 7, 8, 9 and 10 cm jars, two candles a cell, at worked-example fills of 100, 150, 200, 250 and 500 g. One oil through it is 22 candles and 5,800 g of candle. Run all twenty oils through it and you pour 440 candles, 116,000 g: 9,280 g of oil and 106,720 g of wax, plus 240 C1, 200 C2 and 120 Cotton Braided wicks and 440 jars. Before a single candle is sold.

Most of that work repeats itself. If Lavender, Lemongrass and Lime, and Verbena all burn cleanly on C1 in a 7 cm jar, the fourth light herbal is unlikely to need a three-arm bracket to find out it does too. It still needs to be tested, because "unlikely" isn't "won't". The question is whether it needs to be bracketed. That difference is the whole of this page.

Collapse the fragrance axis, not the jar axis

Grouping diameters looks tempting and rarely pays. On a 0.35 cm wall (our assumption), each step from 6 to 10 cm adds 41%, 34%, 29% and 26% to the pool area, and the bracket itself changes from small singles to pairs. Two jars belong in one column only when their inside diameters are within a few millimetres: a 3.3 mm difference around 7 cm already moves the pool area by about 10%. How close is close enough is the one-wick-across-jars post. With five sizes a centimetre apart, the five columns stay.

The fragrance axis is where the collapse happens. Oils share behaviour more often than jars share geometry, and CSI's pages give you enough to make a first guess: the stated family, the stated notes (especially the base), any vanillin line, and the stated load range. The groups below are provisional, a sorting for test planning, not a claim about how any oil burns. How to score pages for likely outliers in more detail is the paper pre-screen post.

How the groups were drawn. Light: citrus, herbal and aquatic families with nothing heavier than a soft musk, cedar or driftwood base. Middle: florals and herbal woods whose bases add amber, sandalwood or creaminess; Bergamot Amber sits here, not with the citrus, because its page lists "warm amber" and a "smooth vanilla base". Heavy: vanilla, coffee, oud and spiced gourmands, including the three with a vanillin statement. All twenty state a candle load or range; two state a maximum above CSI 464's 10% (Japanese Cherry Blossom "6-13%", Cinnamon Vanilla "Up to 13%"), and on CSI 464 the lower ceiling governs. The matrix runs every oil at 8%.

Stage A: two edge oils per group, every jar

Each of the six edge oils goes through the full matrix: the bracket for every jar (6–7 cm C1 v C2; 8 cm C2 v Cotton Braided; 9 cm Cotton Braided v 2 × C1; 10 cm Cotton Braided v 2 × C1 v 2 × C2), two candles per cell, screened at burn 3 and carried to full life. That is 132 candles, 34,800 g of candle and 464 g of each edge oil, which a 500 g bottle covers with 36 g to spare.

The output of stage A is a group wick for each jar: the arm both edges passed on. Three outcomes per group and jar:

The third row is not a failure of the method; it is the method working. A group that splits in the 10 cm jar but holds in the 6 cm jar is plausible (a small pool may forgive differences a large one exposes; judgement, to be tested), and it tells you exactly where the extra testing has to go. If two of your three groups split in most jars, your provisional groups were wrong; re-sort before stage B.

Stage B: every grouped cell verified before sale

This is the rule grouping cannot remove: every oil, in every jar it will be sold in, passes two candles on its group's wick, burned to about 1 cm left. Two candles because one can be an accident of pouring; full life because the top of a candle cannot approve the bottom (how far down to burn a test candle). For the fourteen non-edge oils in five jars that is 140 candles and 33,600 g of candle, 192 g of each oil, which two 100 g bottles cover with only 8 g spare.

Our worked example has six failed cells out of seventy (illustrative, not a prediction): Vanilla Bean and Cinnamon Vanilla under in the 7 and 8 cm jars, Japanese Cherry Blossom over in the 7 cm jar, and Lemongrass and Lime under in the 10 cm jar. The two 2.5% vanillin oils failing together in two jars is exactly the pattern step 5 is for: in a real run, it would split the heavy group in those jars. Stage C costs 12 candles and 2,700 g of candle, and each failing oil needs another 100 g bottle, because verification left only 8 g in its first two.

Candles, grams and rupees before and after

All figures are the worked example: the five-jar matrix, twenty oils, 8% on total weight, CSI 464 at the 10 kg price, oils at the per-gram price of the bottle bought, wicks flat per wick. They exclude jars, labour, freight, packaging and GST.

Per oil, the change is clearest in the bottles. Before grouping every oil needs 464 g, so you buy a 500 g of each. After, only the six edges need that; each verified oil needs 192 g, two 100 g bottles. A 500 g bought "to be safe" for a verified oil would leave 308 g on the shelf before you know whether it sells. One oddity cuts the other way: Blue Ocean's 500 g is fractionally dearer per gram than its 100 g (₹0.01 a gram more), so for that oil small bottles are not even a premium.

When grouping saves less than you think

Grouping removes bracketing, not testing. In the worked example the six edges still account for 34,800 g and verification for 33,600 g: verification is nearly half of what remains, and the rule says it cannot be skipped. Three things move the number honestly.

Sell fewer oils in the big jar. The 10 cm column is half the grams of every row. If only six of the fourteen verified oils will ever go into the 10 cm jar, you verify those six there and no others: 16 fewer candles and 8,000 g less candle. The rule is "every jar it will be sold in", not "every jar you own".

Get the groups right first. Every failure in stage B costs two more candles and another bottle. A group sorted carelessly fails more often; one sorted on stated bases, vanillin lines and families fails less. The pages are free to read.

Don't add groups to feel safe. Four or five groups mean more edges in stage A (two per group in every jar) and fewer members per group to share the saving. Three is our starting suggestion for twenty oils; two can work for a narrow range. If it is the jars, not the oils, you are unsure about, the anchor row-and-column design reads jar effects and fragrance effects separately; once groups are settled, fewer wick SKUs is the SKU audit.

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; the oils are picked as group edges, on what their pages state. Skip for now: 1 kg of any oil (no cell here needs more than 464 g), and 500 g bottles of oils that are only being verified.

Frequently asked questions