A Tibetan syllable is written as an assembly and spoken as a unit, and the relationship between the two is regular without being transparent.
Several of the written parts are not pronounced at all. A prefix is silent. A superscript is silent. A subjoined wa is generally inaudible. What those silent parts do instead is determine the tone of the syllable, so they are not decorative and they are not merely historical: they are still carrying information, in a different medium from the one they were built for.
Other parts change something other than themselves. A subjoined ya palatalises the letter above it. A subjoined ra alters the consonant above it rather than sounding as an r. A suffix reshapes the vowel in front of it. In each case the written element and the spoken effect are in different places.
The result is a script that cannot be read by substitution but can be read by rule, and the rules are neither numerous nor arbitrary. That is the honest description, and it is worth setting against two easy misreadings. Tibetan spelling is not chaotic, and Tibetan spelling is not phonetic. It is a system in which the written syllable predicts the spoken one, given the rules, and does not resemble it.