I’ve been using smart.fm for a while now, mostly working my way through the core 2000 series of vocabulary goals. As good as smart.fm is, it has some failings which have led me to start using it a little differently.
There are two main problems. The first is that there is no way to undo an answer. Since you don’t grade yourself, the first answer you give is the only answer you’ll have a chance to give. Therefore, if you make a typo whilst entering the text for the last phase of questions, smart.fm will assume you actually did not know the word, and set back your progress a week or more from where it should be. The next problem compounds this, as well as being a serious nuisance on its own. Normally, to “master” a word, if you get the answers right every time the question will come up four times. If you have some trouble with the word, of course you’ll need to answer it more often. The problem is with the timing. It seems smart.fm’s SRS system does not properly take into account the need for more frequent reviews of problem items. As the goal progresses toward the end, naturally the reviews come further and further apart; but problem items also come further and further apart. That is why a goal can sit at “99%” for ages. There are items you haven’t mastered, but instead of asking you at the appropriate intervals it just lets them sit there at timings commensurate with those for items you already know.
Then, once the goal is at last complete and you’re in the long-term review mode, the frequency the cards come up is set in stone. A true SRS will give you grading options so that items that are still a little difficult can be marked “hard” or whatever scale the system uses, and easy items can be marked “easy” or “5” or whatever, to optimize the efficiency of reviewing. But with smart.fm the question is answered either correctly or not, and therefore the spacing cannot accommodate your real requirements. (In fact, if you do answer a “mastered” item incorrectly, it remains at “mastered” status, so I don’t even know if it takes your answers into account at all once in long-term mode.)
The core 2000 goals have their own unique drawbacks once you want to really master (not “master”) the words; namely, they’re too easy! With every review, you get the audio and a related image – and sometimes the meaning of the word is actually in the image. This is brilliant for initial acquisition, but it becomes a limit to the depth of your memory.
Hence my new and slightly different approach to this. I believe this will take best advantage of the real strong point of smart.fm, namely, initial vocabulary acquisition. That, it is really brilliant at. So now, once the goal is at 98% or 99%, instead of waiting for the incomplete words to leisurely make their way to the front, and then continuing with the very inadequate long-term reviewing, I have made a deck in Anki (just modified the shared core 2000 deck, actually), that I will be using for the final phase. I’ll be reviewing from kana to kanji, in order to best memorize the kanji readings. This will of course require me to write the sentence for each review. As I complete each goal on smart.fm, I’ll stop reviewing it there, unsuspend the cards in my Anki deck, and carry on from there. The one thing I still need to do for this deck is to optimize the initial timing; once I’ve passed a card for the first time, I don’t want to be seeing it again nearly as soon as the default time, because I already mostly know it from my smart.fm reviewing.
edit: a couple points that people reminded me of. You can actually turn off the audio and images if you want – I’d forgotten that because I’d never actually done it 🙂 Also, I should mention that this is all based on using the iKnow! app; I have done very little with Drill Beta. However, as far as I know, the background timing control is the same between them. Please correct me if I’m wrong on that.