Here's an example. If you read 'Learning Principles', you might remember that 'die Gießkanne' means 'watering can'. (It wouldn't be at all surprising if that fact had been forgotten, because the optimal time to review it was probably about 60 seconds after you first read it.) The 'kanne' (for 'can') doesn't seem too hard to remember - you could just pick the closest-sounding English word, and you'd be right. The German word for 'to pour' is 'gießen', so the word is quite logical if you already know 'gießen' but it is pretty arbitrary if you are meeting 'Gieß' for the first time. 'Gieß' just doesn't sound anything like its English equivalent. In fact, it sounds like 'geese' (once you know that ß is pronounced as 'ss', that is). On being asked to translate 'Gießkanne', many English speakers will guess that some sort of can is involved... but at that point they are likely to stall.
That's where mnemonics step in. After forgetting 'Gießkanne' for the second time around, my son opened up Microsoft Paint and, in about 60 seconds, drew the following picture:
It won't win any art prizes but it looks enough like a watering can made out of a goose that it links 'watering can', via 'geese-can' to 'Gießkanne'.
He saved it as a GIF file, and added it to the item in the edit window, like this:
Now, if he ever gets it wrong again, the image will appear. More importantly, the mental effort of making the association has made it very unlikely he will ever forget it again.
If mnemonic images are loaded into the vocab file before a student tries to learn it, the mnemonic image is displayed when the item is first encountered: