Transliteration does indeed vary widely. There's a de facto standard used rigorously in academic writing known as IAST (http://en.wikipedia.org/wiki/International_Alphabet_of_Sanskrit_Transliteration), but it's largely unknown to Hindi speakers unless they happen to have been exposed to it in academic works written in Western languages.
For informal use you'll indeed see e.g. थोड़ा written as thoda and thora, because that sound is somewhere between an r and a retroflex d (the technical term is retroflex flap (http://en.wikipedia.org/wiki/Retroflex_flap)). A fluent speaker will already know the word, so there's usually no ambiguity. Similarly a fluent speaker will not be confused when no distinction is made between dental and retroflex consonants, so for example both त and ट will be rendered as t. This makes things rather challenging for us non-native/non-fluent speakers, to say the least.
You might want to look into ITRANS (http://en.wikipedia.org/wiki/ITRANS), a scheme that uses regular Latin characters to represent all sounds accurately. It started to catch on for a while for online correspondence and for things like the the spectacular Hindi Songs Archive (http://smriti.com/hindi-songs/), but appears to be slipping into obsolescence now that it's much easier to type Devanagari directly as Unicode.
Transliteration
Transliteration does indeed vary widely. There's a de facto standard used rigorously in academic writing known as IAST (http://en.wikipedia.org/wiki/International_Alphabet_of_Sanskrit_Transliteration), but it's largely unknown to Hindi speakers unless they happen to have been exposed to it in academic works written in Western languages.
For informal use you'll indeed see e.g. थोड़ा written as thoda and thora, because that sound is somewhere between an r and a retroflex d (the technical term is retroflex flap (http://en.wikipedia.org/wiki/Retroflex_flap)). A fluent speaker will already know the word, so there's usually no ambiguity. Similarly a fluent speaker will not be confused when no distinction is made between dental and retroflex consonants, so for example both त and ट will be rendered as t. This makes things rather challenging for us non-native/non-fluent speakers, to say the least.
You might want to look into ITRANS (http://en.wikipedia.org/wiki/ITRANS), a scheme that uses regular Latin characters to represent all sounds accurately. It started to catch on for a while for online correspondence and for things like the the spectacular Hindi Songs Archive (http://smriti.com/hindi-songs/), but appears to be slipping into obsolescence now that it's much easier to type Devanagari directly as Unicode.