A few things that can help also is to check by sweeping a multiband compressor in solo at the end of the signal chains whether it's a distortion free signal or whether there is difficult to hear clipping inside of the signal. It is also helpful to use monitors or cans that you can trust in terms of when the mix clips and when the monitors/cans clip. On cans it is the SPL data in the specification that is the figure that needs to be high, many cans break down before the mix clips and when you are clipping the mix during mastering you want to be able to separate the distortion from the content and the distortion from the monitors/cans.
Also pay attention to the effects you are using, many of them could be introducing the distortion on purpose.
When you gain stage it can work to add gain stages that are there purely to protect your signal and to leave you with more gain staging options later. Because late in the process you might have fairly long signal paths with many fx stages and it can be very complex to figure out (besides clipping) where the gain is best added or removed. Another thing is that with these additional gain protection stages, you are forced to work more efficiently with the signal because you have less headroom left, which means that when you then undo some of these at the right places it can sound very good due to the gain staging you've created. So I recommend that you build the mix in such a way that you have good gain staging options later on. If you don't have that, then at least ensure you are well free from clipping. But be aware that a signal that is very far from clipping due to a lot of gain reduction moves on a lot of places, can sum up to a mix that lacks depth at a specific playback volume. Since you are working against certain commercial loudness ranges, then it's going to translate into a depth issue when you decrease the gain too much, you are losing the natural character of the sound. An acoustic guitar for instance can go from stable, clear and comfortable to weak and muddy just because you don't have a good gain staging.
Another thing is that if you mix into processing, which I do, use a processing that not only makes your mix sound good because you mix into it, but also use a processing that guides your mixing moves to your goals because you are mixing into that processing. A typical example would be to mix into 100%L and 100%R, you can do so but because you have no room for additional widening you are more dependent on stereo widening tools and other complex routes at the point when you feel the mix is not wide enough. Similarly, you can create a fake clipping point that can help make you back off much enough. So when you know by ear that the mix sounds sweet because it does not clip, then when you also remove the fake clipping point, now you add that extra bit of distance to the ceiling and all you need to do is to ensure the ceiling on the last limiters are keeping the final signal free from clipping when consumed by a target encoding, as a secondary clip protection. So instead of having only a single clip protection layer, it is smarter to have a few.
