My own personal 5 eggs based on much trial and error on my part over 15 years or more:
1.) As a beginner useful input has to be audio and it has to be comprehensible - no ifs no buts. (And I’ve tried almost every approach one can imagine - including some pretty funny experimental sh_t!)
2.) Any reading or writing seems to be of very limited benefit before one is able to understand the spoken language to some significant degree and to manipulate thoughts in the language.
3.) Learning lists of vocabulary is also exponentially more effective if one uses bilingual audio - either bought or self-recorded. Generally lists work best for nouns - especially if they are physical objects. It is very hard to learn a verb without an example sentence.
4.) At a more advanced level there is a kind of paradigm shift, and reading and writing become more beneficial - but that’s only after one is already pretty advanced by normal standards. (Speaking and literacy are almost two entirely separate things. As a student I knew two people who had grown up in a bilingual Anglo-German environment. In both cases, their spoken German was not far very sub-native level; their written German was actually weaker than the average foreign learner of the language!)
The first point is the real bummer. How can input ever be comprehensible - especially in the case of more exotic languages?
If one is learning a language relatively close to English (Spanish or Italian maybe) one could perhaps try massively listening to content where one has some existing grasp of the context, and the fog might eventually start to clear? In the case of more difficult or complex languages that will hardly ever be the case, I think.
What works for me is this: I take audio content (in my case from Assimil, Linguaphone or Buske - but it could be from any other course, or indeed from LingQ) and edit it using Wavepad Pro. I double up each sentence, I use the digital text-to-speech function to add in word-for-word English translations before each sentence, and then I have the whole segment repeated over without translations. If necessary I use the trim function to take out silences longer than about 1 second. If I’m feeling especially funky I might add a backtrack of jazz guitar at reduced volume too.
An example from an early lesson of Assimil Farsi:
Boy in home was. Pésar dar manzél boud. (Farsi repeated)
Mother came and said. Mâdar âmad va goft (Farsi repeated)
You where were? To kodjâ boudi? (Farsi repeated)
I in market was. Man dar bâzâr boudam. (Farsi repeated)
Pésar dar manzél boud.
Mâdar âmad va goft.
To kodjâ boudi?
Man dar bâzâr boudam.
This is listened to when walking, driving, sketching, working out, doing random chores, etc. It’s a fair amount of work to edit the files. But it’s for sure more effective than anything else I’ve ever done!
(It’s considerably easier to work with Glossika this way, because all one has to do is trim silences and add music as required - but the rascal wont get around to releasing Farsi, Urdu and Hebrew.)