Why Is Changing One Variable at a Time Important When Reformulating a Commercial Candle?

What "one variable at a time" actually buys you

Changing one variable at a time means that between a control candle and a test candle, exactly one thing is different: one material, one percentage, one process setting. Everything else — the other materials, the jar, the fill weight, the pour day, the burn room — is held the same. It sounds like a laboratory habit, but for a commercial candle maker it buys three very practical things: attribution (if the test candle is different, you know why), reversibility (if it is worse, you know what to undo), and a clean record (six months later, the file says what each change did).

The opposite — changing several things at once and testing the result — is not always wrong. It gives you one answer about one new candle, and sometimes that is all you need. What it cannot give you is an answer about any individual change. If the new candle passes, you do not know which change was safe to keep elsewhere. If it fails, you do not know which change to undo. The two-changes-failing post is what that looks like after the fact.

The candle counts come from simple arithmetic. With three changes, each of which is either old or new, there are 2 × 2 × 2 = 8 possible candles. Testing all of them at three candles each is 24; testing only the two ends is 6; walking through them one step at a time is 3 × (3 + 3) = 18. One-at-a-time sits in the middle on candle count and at the far end on calendar time, because each step waits for the previous one. That trade-off is the whole decision.

Why changes hide each other

The deeper reason for one-at-a-time is that candle changes interact. The wick burns a mixture of wax and fragrance oil, so a change in either can alter what the wick needs; a jar change alters the pool the wick must melt; a load change alters how much oil is in that pool. Because the parts are coupled, two changes can cancel each other out — a new wax that would have made the pool too shallow, masked by a load change that makes it deeper — and the all-at-once candle looks fine. Later, when one of the two changes is reversed for some other reason, the hidden problem appears, and nobody knows where it came from.

The worked example: three cost changes, in order

Here is a typical commercial reformulation. A 200 g candle in a White Matte Glass Jar, scented with Sandalwood Fragrance Oil, currently uses Luxury Soy Wax Chunks at a 9% load (the maker's formula) with a Cotton Braided wick. The maker wants to cut material cost with three changes: move the wax to Luxury Soy Wax CSI 464, drop the load to 8%, and move the wick to Eco Thin C1. Both loads sit inside both waxes' stated maximums (Chunks up to 13%, CSI 464 up to 10%). Costs below use CSI's 1 kg prices for wax and oil and exclude labour, packaging, freight and GST.

The wax step saves ₹16.67: 182 g of wax at ₹599.00 against ₹507.40 a kilo. The load step saves ₹9.04: two grams less Sandalwood at ₹5.03/g, partly offset by two grams more wax. The wick step saves ₹6.10, the difference between ₹12.00 and ₹5.90 per wick. Together, ₹31.81 per candle, from ₹282.30 to ₹250.49.

The order is not arbitrary. It follows dependency: the wick is sized to the fuel it burns, so it comes last, after the wax and the load are settled. The load is a formula choice that only makes sense in the wax you will actually use, so it comes after the wax. The wax comes first because everything else depends on it. Doing the wick first would mean sizing a wick for a wax and load you are about to abandon — and then redoing it.

The protocol, step by step

The decision rule is the same at every step: pass when no safety line is worse than that step's control and the performance and storage lines are the same or better; escalate when a performance line is worse but safety is fine (for example, weaker throw at 8% — you may decide the saving is not worth it); fail when any safety line is worse. CSI publishes no numeric flame-height or jar-temperature limit; acceptance is no worse than the control, within your jar supplier's guidance and the recognised fire-safety standard you work to (ASTM F2417 is commonly cited). A cost saving never pays for a hotter jar.

The priced kit and the payback

Across the three steps the sequence pours 18 candles. The current materials come from your stock: Chunks for the Step 1 controls (546 g), Cotton Braided wicks for 15 candles, and Sandalwood for all 18 (306 g, about ₹1,538.20 at the 100 g rate). The kit buys what is new: CSI 464 for 2,748 g of candles plus a 5% pour-loss allowance (a labelled working assumption), Eco Thin C1 for Step 3, and jars.

₹2,997.20 in purchases. Add an example ₹2,000.00 for your time across three burn rounds — replace it with your own figure — and the one-off cost of doing this properly is ₹4,997.20. At ₹31.81 saved per candle, payback is ₹4,997.20 ÷ ₹31.81 = 158 candles. For a bestseller that is a short payback. For a slow SKU it may not be worth reformulating at all; the ₹15-per-candle post and the 10% cost-reduction post work that decision in more depth.

When one-at-a-time is the wrong tool

One-at-a-time is a default, not a law. It is the wrong tool in two situations. The first is when the changes cannot be separated in time — both the wax and the fragrance supplier are out of stock at once, and there is no old material to hold constant for long. Then a small factorial design, where all combinations are poured and burned together, gets the same attribution in one round. The simultaneous wax-and-fragrance post designs exactly that, with a control and candle counts.

The second is when you are not trying to learn about each change at all, only to approve one specific new candle quickly — a one-off corporate order in a different jar, say, that will never be repeated. Then an all-at-once test of that exact candle against your approved control is honest, as long as the record says so: "this candle approved as a whole; individual changes not evaluated". What is never honest is testing all at once and recording it as if each change had been approved separately.

The habit also protects your purchasing. When purchasing is free to substitute "equivalent" materials, several changes can arrive in the same month without anyone deciding to make them. The procurement post explains how uncontrolled substitutions break one-at-a-time from the outside, and the approved-supplier list is the tool that stops it. For how this fits in a full change-control system, see the system post.

Frequently asked questions