My Custom Fragrance Blend Separates from the Wax Even Though the Individual Fragrance Oils Work Properly on Their Own — Why?

Why a blend can misbehave when its parts don't

A fragrance oil is already a mixture of many materials. When you blend three of them, the wax is being asked to hold a new mixture that none of your single-oil candles ever tested. Sometimes the new mixture behaves like an average of its parts. Sometimes it doesn't: two oils can make a premix that clouds or layers, or a combination can come out of the wax more readily than either oil does alone. Those are tendencies makers report, not predictions anyone can make for a given pair, because CSI does not publish oil composition and we won't pretend to know what is inside a bottle.

There is a useful fact hiding in the arithmetic, though. In a blend at 8%, no component is at 8%. In the worked 50/25/25 blend, Bergamot Amber is 4.00% of the candle and Sandalwood and Ylang Ylang Flower are 2.00% each. So if each oil is dry alone at 8%, "too much of one oil" is the wrong explanation for a blend at the same total: every component is at a fraction of the load it already passed at. That is why the useful question is which combination, and why the test below drops oils rather than lowering them.

If the blend is also a problem in the bottle, before it goes near wax, that is a different and earlier test. The three-supplier compatibility protocol starts with a neat premix clarity check: mixed oils left to stand in clear glass, looked at for cloudiness, haze or a layer. A premix that separates neat will not sit better in wax.

The paper check: whose load is 8%?

The load applies to the combined blend. At 8% of a 150 g candle the whole blend is 12 g of oil, and the wax is 138 g; the ratio splits the 12 g, it doesn't add to it. The full blending method, from 150 g to 10 kg, lives in the blending hub. Here are the ways a blend sheet commonly goes wrong, cheapest to spot first.

Then check the ceiling. A blend's planning maximum is, conservatively, the lowest stated maximum among its components (our judgement; the blend itself has no page). Bergamot Amber states "6-10% for soy wax candles", Sandalwood "up to 10% for soy", Ylang Ylang Flower "Up to 10% for soy wax", and CSI 464 "Up to 10%", so this blend's planning ceiling is 10%. Swap in an oil whose page stops lower and the whole blend inherits that lower figure. At 8% the worked blend sits inside every one of those figures on either reading of their basis (8% of total is 8.70% of wax).

Blend separation: what you see and what it points to

Ranked in the order we would suspect them, most likely first. The ranking is editorial judgement; CSI publishes no failure statistics. The right-hand columns are the checks.

Rows 4 and 6 are why the dropout test pours everything in one session and stores it on one shelf: if the singles and the blend were made on different days, the difference you saw may belong to the day, not the blend. If you are not sure your singles were made the same way, add them to the test (below).

The component-dropout test: the blend and its three minus-ones

Call the oils A, B and C. The test pours four versions, two candles each: the full blend (A+B+C, the control) and three minus-one versions (A+B without C, A+C without B, B+C without A). Every candle carries the same 12 g of oil in the same 138 g of wax. When an oil is dropped, its share goes to the remaining oils in their original proportion, so A+B keeps A and B at 2:1 as they were in the blend. That design is the point: the oil-to-wax ratio never changes, so a version that stays dry has lost an ingredient, not lost load.

Notice two things in the shares. A rises from 4.00% in the blend to 5.33% in the two versions that keep it (66.7% of their oil), still well below the 8% it passed at alone. And in B+C the two minor oils each rise to 4.00%. So a failing minus-one version still can't be explained by one oil being "too strong"; each is below its tested single load. If your ratio is different, use the same rule: the dropped share is shared out in the remaining oils' proportion.

Costs exclude labour, freight, packaging and GST. The cash outlay if you owned none of it would be ₹2,652.20; in practice you own the blend's oils and probably the wax, so the real spend is jars and wicks, or nothing.

Reading the triangle and what to change

The same logic works for other one-side and two-side patterns: the oil that appears on every failing side is the suspect, and a single failing side names a pair without saying which member. When the suspect is a pair, change the oil whose loss hurts the scent least; in the worked blend that is usually a quarter-share component, but it is your blend and your nose. Then re-run just the new blend and the version you changed against a fresh control, not the whole triangle.

Once a version of the blend stays dry, it still needs its own load and wick. Set the load with a blend ladder rather than borrowing a single oil's number, since a blend has its own practical maximum just as each oil does in the per-oil card post. And judge the new ratio in wax, not in the bottle.

What not to buy for a separating blend

For a separation problem with a single oil, not a blend, the first-level decision on whether to replace the oil is whether to change an oil that keeps separating; this page assumes the single oils have already passed.

The three oils and what holds the test steady

The worked example's three components, the wax held constant, and the scale the ratios depend on. Chosen on page statements, price per gram and stock on 26 September 2026; CSI has no measured compatibility or throw data, so nothing is ranked on performance. Oil tables show every size, 150 g candles at 12 g, and whether each size covers that oil's share of the test.

Frequently asked questions