My First Candles from a Production Batch Smell Weaker than the Last Ones — Could the Fragrance Oil be Settling or Mixing Unevenly?

Three causes, three fingerprints

Take the batch this page works through. It is 5 kg of finished candle at an 8% load, poured into 25 White Matte jars at 200 g each. By CSI's convention the load is a share of the finished weight, so the batch holds 400 g of oil and 4,600 g of wax. If the oil is spread evenly, every candle carries 16 g. When the first candles smell weaker than the last, something moved the oil between the first jug fill and the last. In general candle-making practice there are three usual suspects, and they leave different patterns across the pour order.

The last two rows matter as much as the first three. Makers often notice the problem because the first candles poured were also the first ones lit, a day or two earlier than the rest. A candle tested at 5 days against one tested at 12 days is not a fair comparison, whatever happened in the pot. Same cure age for a fair comparison covers that trap in full. Smell fatigue is the other trap. Sniff the candles in the same order every time and the first one will usually seem strongest. Fragrance fatigue versus weak throw explains how to tell the two apart.

Temperature deserves a note of its own. The only CSI temperature on record is from CSI's earlier guide to Luxury Soy Wax CSI 464: melt and pour at 80–90°C. That is CSI's CSI 464 guidance, so confirm it on the product page. For Luxury Soy Wax Chunks and every other wax, follow the wax maker's instructions for the fragrance-addition and pouring temperatures. This page does not set a temperature. It shows how to find out whether your temperature stayed where you set it for the whole pour. What happens at the wrong mixing temperature covers the temperature question itself.

What a gradient means in grams

It helps to see the problem as oil per candle rather than as a smell. The table below spreads the same 400 g across the 25 candles as an illustrative gradient from 6.5% in the first five candles to 9.5% in the last five. These figures are invented for the example, not measured. They show how a batch can be correct in total and still wrong in every jar. Oil cost uses Mango Coconut at its 500 g price of ₹1,982.00, which is ₹3.964 a gram.

The totals check out: 65 + 72.5 + 80 + 87.5 + 95 = 400 g. Your weighing was right, and the finished batch still holds candles from 13 g to 19 g of oil. A customer who buys two candles from the same batch may get one from each end. The 6.5% candle is not only weaker. It has less oil than the recipe you approved, so it is a different product from your test candle. The fix is in how the oil is spread, not in adding more of it.

The sampled-pour test

The sampled-pour test turns the complaint into data you can act on. You pour a normal batch, but you record the pour order of every candle and set aside a sample at even intervals. Those samples are then burned together, at the same cure age, under the same conditions, with the tester unaware of which is which. Run it on the batch that showed the problem, or on a fresh batch made exactly the way you usually make it.

Keep the scale of the rules in proportion. "Within about a point" on a 10-point scale is a working threshold for a small bench test, not a statistical standard. If two runs disagree, run the test again before changing the recipe. The aim is to catch a consistent pattern, and a single odd candle is not one.

Fixing the batch process

Once the pattern points to a cause, the fixes are procedural, and none of them needs a different oil. The most useful is also the simplest: make the batch small enough to mix and pour before it has time to separate or cool. Splitting the 5 kg batch into two 2.5 kg sub-batches means each gets its own 200 g of oil, its own full stir and a pour roughly half as long. It adds one more weighing and one more stir. In return, fewer candles are filled from a pot that has been standing.

For stirring time, this page will not give you a single number, and nobody honestly can. How long a batch needs depends on its size, the vessel, the tool and how you stir. The right figure is the one your sampled-pour test shows is enough, and you then write it into your process. Why a 20-second stir fails and mixing time for large batches cover the stirring side. A repeatable mixing and curing protocol shows how to write the result into a standard procedure your team follows every time.

A note on equipment. An electric melter holds temperature better than a pot on a stove, and it can help with drift. It does not replace stirring. A melter with a bottom tap draws from the base of the tank, which is the first place any separated oil could collect, so a pre-fill stir matters even more. If you pour from a tapped melter, add a stir before each draw to the procedure.

What to buy for the re-run

The re-run is one full 5 kg batch made the fixed way, sampled at the same five positions. Everything below is priced from CSI's live listings. The oil quantity follows the batch maths, 400 g, plus the usual working allowance of 3–5% for pour loss and test burns (412–420 g). The 500 g pack covers that with room to spare.

Of the ₹8,027.00, ₹1,103.00 is tools you keep. The consumables (₹6,924.00) make 25 sellable candles if the re-run passes, so the test costs you almost nothing extra when it works. Material cost per candle at these prices is ₹250.34: ₹63.42 of oil plus the ₹186.92 base of 184 g of wax, one wick and one jar. That figure excludes labour, packaging, freight and GST.

When the batch was never the problem

Flat scores are common, and they are worth taking seriously. The usual story runs like this. The first candles poured were the first ones lit, at 3 or 4 days, and they seemed weak. The last candles were lit a week later and seemed strong. The candles had not changed. Their age had. Cure is covered in why curing affects fragrance performance. For this page the practical rule is simple: never compare candles of different cure ages and call the difference a batch problem. If your production runs across several days, standardising cure time before QC shows how to label batches so every burn test happens at the same age.

The second confounder is the test itself. Different rooms, doors open or shut, a fan on for one burn and off for the next, or a tester who knows which candle is which will all produce differences that have nothing to do with the pot. If your blind scores are level but you can still "tell" when you know the labels, that is expectation, not fragrance. Keep the blind codes and the fixed room for every future comparison.

If the five scores are level and every candle still seems weak, the batch is even and the formula needs work instead. That is a different question. What to test before rejecting an oil sets out the variables, and when to stop waiting and reformulate gives a decision rule based on cure checkpoints.

Frequently asked questions