How Many Jars per Wax Should I Pour When Running an Initial Glass-Adhesion Comparison?

The short answer: three jars screens, seven starts a decision, the pilot settles it

The honest answer to "how many jars?" depends on what you want the first round to do. If all you need is to catch a wax that fails badly in your jar — big pull-away patches on most jars by day seven — three jars per wax will usually show it. If you want the round to start telling two decent waxes apart, you need more jars than most makers pour, because wet spots are patchy and jar-to-jar variation inside one wax can be as large as the difference between waxes. This page puts numbers on that, using nothing but the arithmetic of counting jars, and then prices each jar count in real CSI materials so you can see what each extra jar buys you.

The working recommendation for an initial comparison in the series default candle — 150 g in CSI's Clear Glass Jar with Golden Lid (150 gram), 12 g of oil and 138 g of wax at 8% — is seven jars per wax. That is not a magic number. It comes from three facts: the addendum convention of at least three jars per arm for a screen; the arithmetic below, which shows how little three jars can prove; and the pack maths, because four arms of seven is 28 jars, which is three packs of 10 with two left over for a weighed fill check. Six per arm would leave six jars unused in the same three packs, and ten per arm needs a fourth pack.

What seven per arm does well is eliminate. If one wax spots on five of its seven jars and the others spot on none or one, you can drop it and carry two finalists into a second round of ten jars each. What seven per arm does not do is crown a winner between two waxes that both look reasonable. That decision belongs to a 50–100 candle pilot, judged against a threshold you set in advance — the acceptable-percentage post shows how. The first round's job is to make the pilot cheaper by sending fewer waxes into it.

What three jars can and cannot show

Start with the simplest question a round can answer: "does this wax spot in my jar at all?" Suppose that, in your jar with your oil and process, a wax produces a visible wet spot on one jar in five — a 20% per-jar rate. Pour three jars and the chance that at least one of them shows a spot is 1 − 0.8³ = 48.8%. Almost half the time, three jars of a 20%-spotting wax all come out clean. Pour seven and the chance rises to 79.0%; ten, 89.3%. The figure of 20% is a hypothetical rate for the arithmetic, not a measured rate for any wax — CSI publishes no wet-spot rates, and neither does anyone else for your jar.

Turn it round and the problem is clearer. Three clean jars feel like good news. But if none of three jars spots, the true per-jar rate could still be as high as 63.2% at the usual one-in-twenty confidence — three clean jars are consistent with a wax that spots on more than half of everything you will ever pour. Seven clean jars bring that ceiling down to 34.8%, ten to 25.9%, twenty to 13.9% and thirty to 9.5%. This is the exact binomial bound, and it is why a luxury clear-glass line cannot be approved on a handful of jars no matter how beautiful they look.

The table also shows where small arms are useful. A wax that spots on half its jars is caught by three jars 87.5% of the time and by seven jars 99.2% of the time — so if your current problem is gross, as it is for the maker hiding 30% of 2,000 candles a month behind labels, even a small screen will show which candidate repeats it. A wax that spots on one jar in ten is a different matter: three jars miss it most of the time, and only a pilot will find it.

Telling two waxes apart needs more jars than finding a bad one

Comparison is harder than detection, because both arms carry noise. The cleanest way to see this is to ask what split between two arms is strong enough that you would not expect it by chance more than one time in twenty (a one-sided Fisher exact test on spotted versus clean jars). With three jars per arm, even the most extreme result — three of three spotted against none of three — sits exactly on that line (probability 0.05). With six or seven jars per arm, the worse wax has to spot on at least 4 jars while the better one spots on none. With ten per arm the bar is 4 of 10 against 0 of 10; with twenty it is 5 of 20 against 0.

Now ask the practical question: if one wax truly spots on 40% of jars in your setup and another on 5% — a large, commercially obvious gap, used here only as a hypothetical — how often does a round of n jars per arm produce a split strong enough to call? With three jars, only about 5% of the time — and only by the most extreme split possible. With seven, about 21% of the time. With ten, 43%. With twenty, 81%. That is why this page calls the seven-jar round a screen: it will often fail to separate even a real difference, and a non-result at seven jars means "keep both", not "they are the same".

There is one way to get more out of the same jars, and it costs nothing: score how much of each jar is spotted, not only whether it is. A yes/no count throws away the difference between a pin-head spot and a patch covering a third of the wall. If you estimate the percentage of visible glass area showing wet spots on every jar — with a printed grid or a marked acetate sleeve held against the glass — two arms of seven give you fourteen numbers to compare, and a consistent difference in area shows up long before a difference in yes/no counts would. What to record when comparing glass adhesion sets out the full record; the scoring sheet below is the short version for this round.

Seven jars per arm only work if position and handling are equal

More jars do not help if the jars in one arm share an advantage the others do not. The commonest hidden variable in an adhesion screen is position: jars at the edge of a tray, near a window or under an AC vent cool differently from jars in the middle, and cooling is one of the commonly cited contributors to wet spots. Outer jars versus centre jars and spacing jars so each cools at the same rate deal with that in production. In a screen, the fix is simple: rotate the arm letters through the tray, as in the diagram above, so that every wax has jars at the edge and in the centre, and record each jar's position on the sheet.

The scoring sheet for this round

Score every jar on two measures: whether it shows any wet spot, and how much of the visible glass is affected. The grade scale below is an example house scale — use it or write your own, but then keep it for every round that follows, because a score only means something next to other scores taken the same way. Score from the photographs, with the arm letter covered, so the scorer does not know which wax they are looking at.

Two summary lines per arm drive the decision: the count of spotted jars out of seven and the mean area percentage. Write your elimination rule before the round starts. A rule you can adopt: drop a candidate wax if it spots on at least four of seven jars while another candidate spots on none, or if its mean area is clearly higher than the control's on every checkpoint. Everything else goes forward. The full qualification scorecard adds smooth tops, frosting, sinkholes, throw, burn and labour once adhesion has narrowed the field.

The priced first-round kit

Here is the seven-jar round priced with listed, in-stock CSI packs. Grams use the CSI convention (oil = W × L): every jar takes 12 g of British Rose and 138 g of wax. Each CSI arm needs 995.0 g of wax including a 3% pour-loss working allowance, which fits inside a single 1 kg bag. The oil covers all four arms including your control, so the only variable is the wax. Your own wax for the control arm is not in the CSI total.

If the round eliminates one wax, round two pours ten jars each of the two finalists plus ten controls — in the example below, Chunks and CSI 464 survive. That is 1,421.4 g of each finalist wax (1 × 1 kg + 1 × 500 g of Chunks for ₹898.00; 1 × 1 kg + 1 × 500 g of CSI 464 for ₹767.40), 4 × 100 g of British Rose for ₹1,368.80, three more packs of jars and fresh wicks: ₹6,202.20 in all. Leftover wax from round one counts against it if the bag lot is the same — buy round two's wax only after round one is scored.

What one more jar per wax actually costs

Jar counts are usually chosen by feel. It is better to choose them by price, because the price of one more jar is small and visible. One extra jar in a CSI arm is a ₹94.40 jar, 138 g of wax at the 1 kg rate, 12 g of British Rose at the 100 g rate (₹3.42 a gram), a ₹5.90 wick and a ₹2.50 sticker. The glass is most of it.

So moving every CSI arm from seven to ten jars costs about ₹2,021.93 in materials (three more jars in each of three arms), before any extra wax bag the larger arm forces. Weigh that against what the table on comparing arms shows: ten per arm turns a 21% chance of a decisive result into 43% for a large hypothetical gap. Whether that is worth it depends on how many candles ride on the choice. A maker pouring a few hundred clear-glass candles a season can screen at seven and let the pilot decide. A maker like the 2,000-a-month line loses more than that in one week of hidden candles and should go straight to ten.

One more number frames the scale. A 1 kg bag fills 7.25 of these candles' wax requirement, so the seven-jar arm uses most of a bag and wastes little. Ten jars per arm needs 1,421.4 g, which forces a second bag (a 500 g top-up) for each wax — the step where cost jumps. Cost per visually acceptable candle is where these test costs are set against the price of hidden or discounted stock.

The decision rule after round one

Apply from the top and stop at the first line that is true. It is written so that two people looking at the same sheets reach the same next step.

Note the fifth line. Seventeen clean jars across two rounds is a genuinely good result, and it still only bounds the true rate at 16.2%. That is why the pilot exists. For a luxury clear-glass line the acceptance threshold is your choice — there is no industry figure — and the acceptable-percentage post shows how many pilot jars a given threshold demands. One month of storage and the checkpoint schedule decide when a clean jar counts as clean.

The CSI products in this round

Where this round sits in the qualification

This post owns one decision — how many jars to pour in the first comparison. The rest of the qualification is covered by its neighbours, and each one inherits the jars, numbering and scale set up here. Testing several CSI waxes in your exact clear jar explains why the vessel must be the one you sell in. Comparing waxes without changing several variables at once protects the one-variable rule this round depends on. The series capstone lays out the whole path from ugly wet spots to a bulk purchase.

Frequently asked questions