I'm Manufacturing 1,000 Candles a Month — How Should I Test a New Fragrance Before Using It Across a Full Production Batch?

What a thousand candles actually exposes

A thousand 200 g candles at an 8% load is 184 kg of wax and 16 kg of fragrance oil, before any allowance for spills and top-ups. Both of those numbers sit above the largest packs CSI lists, which is itself worth knowing: wax is listed to 10 kg (Natural Soy Pearl Wax also at 50 kg) and fragrance oil to 1 kg. There is no 5 kg or 10 kg fragrance pack listed, so a month at this volume means sixteen 1 kg bottles of one scent at the listed per-kilo price, and that listed price is the ceiling — recurring volumes at this level are quoted on WhatsApp, and we will not guess what that quote is.

Now the sentence that reframes the whole question. ₹3,06,602.60 is the cheap version of getting it wrong. If the fault is cosmetic and you catch it before dispatch, you have lost a month of materials. If it ships — a wick that drowns at the 10% load you guessed, a vanillin-present oil that yellows a white wax over three weeks on a shelf, a scent that reads beautifully at day 2 and flat at day 21 — you lose the batch, you make it again, and you have spent ₹6,13,205.20 of materials to sell one month's candles. Labour, freight, packaging and the stockist's opinion of you are all extra.

The four gates, priced

The design principle is simple: each gate should cost roughly ten times the previous one and prove roughly ten times as much. That way the expensive proof is only ever bought for candidates that have already survived the cheap proof. It also means a failure is never a disaster — it is the system working, at the price the system was designed to fail at.

Read the cumulative column from the diagram alongside that table and the economics settle themselves. ₹2,206.09 is the total amount of money you can actually lose in this system, because from Gate 3 onward you are making sellable candles. About seven-tenths of one percent of a month's materials, to find out whether a month's materials should be spent.

Gate 1: the page and the bottle

Before any oil is bought, read the candidate's product page and fill in five fields. What matters is not whether every field has an answer — it is whether you know which fields are blank, because a blank is a number your own test will have to supply. Across CSI's fragrance range the pages genuinely differ: some publish a candle load, a measured flash point, a vanillin figure and a cure time; 56 of 107 pages state no flash point at all, and a further group state only the shared perfume-grade line "Flash point > 93°C (suitable for alcohol-based products)", which is a threshold for alcohol-based formats and not a measured value for that oil. Treat both cases the same way: the figure is not published — ask for the SDS.

That table is the whole point of Gate 1. Saffron Cedarwood publishes a full note pyramid and no percentage, which does not disqualify it — it means the load has to come from Gate 2, capped at the wax's own stated figure, and that the flash point is a question for the SDS rather than a number you may assume. Japanese Cherry Blossom publishes a wider range than most, up to 13%, which is only usable if your wax also publishes that much. Bergamot Amber publishes a conventional 6–10%. Three candidates, three different amounts of homework, all knowable before you spend ₹104.

Gate 2: the six-jar load bracket

Gate 2 answers one question: what load does this oil want in the wax I already use? Everything else is held constant — same wax, same wick, same jar, same add temperature, same pour temperature, same carton, same cure. Two loads, three jars each, because the third jar in each arm is what lets you tell a result from a fluke. The bracket for a candidate with no stated percentage is 8% and 10%: 8% because it is a conservative working default, 10% because it is where most CSI oil pages that do publish a figure put their ceiling for soy.

Two arms is the minimum; three is better if the candidate's page publishes a wide range. Japanese Cherry Blossom's page states 6–13% for candles and describes 9–11% as the balanced band, so on that oil the honest bracket is 8%, 10% and a third arm at 12% — and only if the wax also publishes that much headroom. Chunks state 8–13%, so 12% is inside both figures; CSI 464 states up to 10%, so on that wax the third arm does not exist. The arithmetic behind choosing between loads is in 6 vs 8 vs 10 fragrance load, and the distinction between a ceiling and a best-performing load in maximum vs best-performing load.

Gate 3: the forty-jar pilot

This is the gate most makers have never run, and it is the one that catches the faults that actually reach customers. A six-jar bracket is poured carefully by whoever designed it. A production batch is poured by a team, at speed, from a bigger melt, into jars from a different carton, on a day when something else went wrong. Gate 3 makes the pilot look like that on purpose: forty jars, at the chosen load, poured at production settings, by the people who will pour the batch.

Forty is not arbitrary. It is large enough for fill-weight variation to show up, for one bad jar to be one in forty rather than a catastrophe, and for you to pull burn-test units and a retained control and still have saleable stock. It consumes 640 g of oil and 7,360 g of wax — ₹12,120.13 of materials — and the sensible purchase at this gate is the 1 kg bottle at ₹6,867.60, because 640 g goes into the pilot and the remaining 360 g starts the production run if the gate passes.

Gate 4: first batch, with a hold-back

Release the scent into production, and hold back fifty candles — five percent of the month, about ₹15,330.13 of materials — from the first batch. They are not stock; they are evidence. Inspect them at day 14 and day 28, burn two of them, and only let the last cartons of that batch ship once the hold-back is clean. Then keep one retained sample per batch, labelled and lidded, for the life of the SKU. The day a customer or a stockist raises something, the retained sample is the difference between an investigation and an argument.

The pipeline and the festive calendar

Add the gates up and a single scent takes roughly five to six weeks from first sniff to a released recipe — two days of desk review and blind cards, sixteen days for the load bracket including its cure, and twenty-one days for the pilot, with a little overlap. At a thousand candles a month you cannot stop production for that, so qualification has to run as a rolling pipeline: one candidate at Gate 1 every month, one at Gate 2, one at Gate 3, and a release roughly every quarter. Run it that way and you always have a qualified scent available when the market asks for one.

Then work backwards from the dates that actually matter. Diwali 2026 falls on 8 November. A festive scent that has to be on a shelf in the third week of October needs its production pour in late September, which means Gate 3 finished in early September, Gate 2 started in early August, and Gate 1 done in July. If you are reading this in the middle of September 2026, a brand-new scent is not a Diwali scent — it is a Christmas or a wedding-season scent, and the honest commercial move is to run this Diwali on a recipe that has already passed and to start the new candidate's Gate 1 now.

One more planning note, and it is a stock note rather than a test note. Because fragrance oil is listed only to 1 kg, a sixteen-kilo month is sixteen bottles, and a scent you have just qualified is a scent you will be reordering monthly. Tell us the volume and the cadence on WhatsApp before the release, not after the pilot passes — that is where bulk pricing, GST invoicing and the SDS or IFRA documentation for your file are handled. Planning a festive run end to end is covered in how early should candle brands order fragrance oils for Diwali.

Frequently asked questions