My First 50 Candles Smell Different From the Last 50 Even Though They Came From One Fragrance Batch — How Should I Investigate Mixing Uniformity?

One fragrance batch does not guarantee one fragrance level

A single fragrance lot tells you the oil itself was the same throughout. It says nothing about whether every candle received the same share of it. Your first 50 and last 50 candles came out of one vessel, one pour session and one oil lot, yet they smell different. That points the investigation at what happened between the bottle and the jar: how the oil went in, how it was blended, how long the wax waited and how it was drawn off.

Mixing uniformity is the obvious suspect, and it is often the right one. But an investigation that starts by assuming the answer tends to find it. The approach below confirms the difference, describes it precisely, reconstructs the batch, and only then profiles uniformity on a controlled re-run.

The batch behind this investigation

Take a batch of 100 candles at 150 g each, which is 15 kg of scented wax. At 8% of total weight that is 13.8 kg of wax and 1.2 kg of oil. Say the wax was Luxury Soy Wax Chunks, whose product page gives a maximum fragrance load of 13% of wax weight; 1.2 kg on 13.8 kg of wax is 1.2 ÷ 13.8 = 8.7% of wax weight, comfortably inside that limit.

Buying 15 kg of Chunks as one 10 kg pack at ₹5,999.00 and one 5 kg pack at ₹2,995.00 costs ₹8,994.00, leaving 1.2 kg of wax spare. The oil was Bergamot Amber: one 1 kg bottle at ₹5,437.00 and two 100 g bottles at ₹560.00 each, a total of ₹5,437.00 + ₹1,120.00 = ₹6,557.00 for exactly 1.2 kg. Before jars, wicks and labour, the batch cost ₹8,994.00 + ₹6,557.00 = ₹15,551.00.

Notice the three containers. Even when every bottle carries the same lot number, the order in which they were emptied into the vessel matters. If the 1 kg bottle went in first and stirred well, and the two 100 g bottles were added a little later as the wax was already being poured, you have a sequence problem, not a lot problem.

A five-stage investigation

Running the smell comparison fairly

Blind comparisons fail for boring reasons. Use jars from the same carton with identical lids, let them all sit at room temperature together for a day before the test, and test in a room that does not already smell of wax or oil. Give the assessor a break between rounds and something neutral to reset on, such as the back of their own hand or fresh air at a window. Keep each lid off for the same few seconds, and change which position the odd jar sits in from round to round so nobody learns a pattern.

Write the result down immediately, including wrong picks. A notebook that records only the rounds that agreed with your suspicion is not evidence, it is a story.

What reconstructions commonly turn up

When makers sit down and rebuild the timeline of a batch like this one, a handful of events come up again and again. The oil was added while pouring had already started because the wax looked ready. One bottle was emptied on the near side of the vessel and stirred only there. Stirring stopped entirely once the first pitcher was filled. A helper took over halfway through and filled the pitcher from a different depth. The jars for the second half came from a new carton with slightly different dimensions. Any one of these can split a batch into two halves without anything being wrong with the oil or the wax.

Profiling uniformity on the next batch

Make the next batch with the same wax lot, the same oil lot and the same method, but record it candle by candle. Number every jar as it is poured. Take a retained candle at fixed intervals through the run, for example every tenth candle, as a suggested starting point. Ten retained candles across 100 give you a profile instead of a before-and-after snapshot.

Read the profile as a shape. A flat line means mixing was uniform and the old difference came from something else in the reconstruction. A steady climb from first to last points to oil concentrating low in the vessel. A step change at one point, say between candles 40 and 50, points to an event: a late oil addition, a pause, a new jar carton or a change of operator.

A step is the most useful result of all, because it has a timestamp. Match the clock time against your notes and you usually find the cause within minutes.

Deciding on the 100 candles and on the process

For the existing stock, keep the halves separate. Compare each half blind against your approved reference candle. The half that matches can be released after a normal burn check. The other half needs its own decision: an honest description as a lighter or deeper version, use as samples, or rework if the product allows it. If the amber-forward half carries more oil, burn-check it before release and stop any test that smokes or flares. Never leave a test candle burning unattended.

For the process, the fix depends on the profile shape. A climb calls for full-depth stirring and a re-stir before each fill. A step calls for a rule that all oil goes in before pouring starts, weighed together into one container first. Combining the three bottles into one weighed container before addition is a small habit that removes a whole category of error.

Add oil below its published flashpoint and never with the melter at maximum. Because the CSI melter's display carries a stated ±10°C deviation, read the wax with a separate thermometer before adding the oil.

Resist the urge to change two things at once on the re-run. If you combine the bottles, deepen the stirring and shorten the hold all in one go, a flat profile will tell you the problem is solved without telling you which change solved it. That matters later, when a helper skips one step to save time and nobody knows whether it was the important one. Change the most likely cause first, profile, then add the next control only if the curve still climbs or steps.

Record the result against the oil and wax lot numbers. If a future batch drifts, you will know whether this same combination has done it before.

Where this page fits

CSI already has guides that cover parts of this question. This page covers the failure investigation for the question above; these go deeper on the rest:

Three things to buy for this job

Frequently asked questions