does the order of color grading matter?

there is a lot of confident advice about the "correct" order for colour grading, and not much explanation of why. the honest answer is that order matters enormously in some places and not at all in others — and knowing which is which saves a lot of time.

this post covers what actually changes when you reorder your grade, with the reasoning rather than the folklore.

why order changes anything at all

each grading step is a function applied to every pixel. running two of them in sequence means applying one function to the output of another. whether swapping them matters comes down to whether those two functions commute — whether a(b(x)) equals b(a(x)).

for a lot of pairs it doesn't, and the difference is visible. for a few it genuinely does, and swapping them is a waste of time.

the pairs that genuinely do not matter

here is one that surprises people. saturation and contrast produce the identical image in either order.

that is not an approximation. saturation works by pushing each channel away from the pixel's luma: result = luma + (channel − luma) × amount. contrast multiplies every channel by the same factor and adds the same offset to each. because the luma weights sum to exactly 1, that shared offset survives the saturation maths untouched, and the multiplier factors straight out of both. the two operations slide past each other.

so if you are wondering whether to saturate before or after adding contrast: it makes no difference at all. spend the attention elsewhere.

the same holds for any pair of purely linear operations that treat all three channels alike — a global exposure lift and a global saturation change, for instance.

the pairs that matter a great deal

as soon as one of the steps is nonlinear, order becomes significant.

curves. a curve is a lookup, and lookups do not commute with anything. pulling the shadows down and then adding contrast is a different image from adding contrast and then pulling the shadows down — the second version is operating on values the first version had already moved.

a lut. the same argument, more strongly. a creative lut is a fixed colour-to-colour mapping built by someone who assumed a particular input. feeding it a heavily contrasted image is feeding it something it was not designed for.

anything that clips. once a value hits pure white or pure black it stops carrying information. a step that pushes highlights past 1.0 destroys detail that a later step can no longer recover — so the same two operations in the other order can preserve a highlight that would otherwise have been lost.

masks and qualifiers. an hsl qualifier selects a range of colour. if you shift hue before the qualifier runs, you are selecting different pixels than if you shift it afterwards. this is the one that catches people out most often: the mask moves when the colour under it moves.

the order that works for most grades

with that reasoning in place, the conventional order is conventional for good reasons:

1. white balance first. it decides what counts as neutral. every later judgement about contrast, saturation and hue is made relative to that decision, so fixing it afterwards means re-judging everything you have already done.
2. exposure and contrast next. get the image sitting in a sensible range before anything creative touches it.
3. primary colour — wheels, tone. shaping the image while it is still predictable.
4. the lut. creative looks generally expect a normalised image, which is what the three steps above produce.
5. secondaries — hsl qualifiers. now that the overall colour is settled, isolate the parts that still need work. running these before the lut means qualifying colours the lut is about to change.
6. curves for final shaping.
7. vignette, then film effects. these are finishing texture. grain especially belongs last — it is applied on top of a finished image, and anything running after it amplifies the grain rather than the picture.

when to break it

this order is a default, not a rule. two cases where breaking it is the right call:

• a heavily stylised lut. if the look you want depends on the lut receiving a flat log image, put it early and grade around its output instead.
• rescuing a difficult highlight. if a step is clipping something you need, moving it later in the chain sometimes preserves the detail — because the values reaching it have already been brought back into range.

the useful skill is not memorising an order. it is recognising which of your steps are nonlinear, because those are the only ones where moving things around will change what you see.

trying it yourself

in luttie, the grading chain is a sequence of stages you can rearrange. turning on advanced mode shows the chain as a node graph, and each node previews the image as it leaves that stage — so you can watch the result change as you move a step earlier or later, rather than reasoning about it in the abstract.

the fastest way to build intuition here is to take one image, add a strong curve and strong contrast, and swap them. then try the same experiment with saturation and contrast, and watch nothing happen.