Being straight with you, as in F110-04: this is not a named incident with a figure attached. No well-documented public case exists, because these failures are private and the people they happen to do not write them up.
What exists is the survey that opens this course. Of 1,000 US crypto holders: 35 percent had lost access to a wallet or account, and 31 percent of those never recovered the assets. The leading causes were a forgotten password, a lost recovery phrase, and lost two-factor access.
In the same sample, only 25 percent stored their recovery phrase on paper.
And 15 percent had ever tested whether their recovery process actually works.
The failures a drill catches are mundane and individually invisible. A word transcribed wrong five years ago. A passphrase you were certain you would remember. A derivation path the old wallet chose silently. A plate in a deposit box at a branch that has closed.
There is no bad decision here to point at. The decision is the one nobody took: a backup was written, filed, and never exercised. Which means the claim that it works stayed untested until the day it was needed, and that day is the one day on which testing it is not an option.
Every previous lesson in this course produced an artefact. This one is the only one that produces evidence.
An untested backup is a hypothesis
You have written down some words. What you actually have is a set of claims:
- that the words are the right words, in the right order
- that your handwriting is legible to you under stress, years from now
- that you will remember the configuration, including a passphrase and which wallet made it
- that you can physically reach the place you put it
- that the process, end to end, works
Not one of those is verified by having done the backup. They are verified by doing the recovery, and there is exactly one moment at which you can afford for the answer to be no: now, deliberately, with almost nothing at stake.
Four failures only a drill finds
A transcription error. One wrong word. F103-02's checksum catches most single substitutions and roughly one in sixteen passes by chance, and it catches nothing at all if you wrote the right words in the wrong order.
A forgotten or wrong passphrase. No error message, ever. F103-02 covered this and it bears repeating because it is the most common cause of a false catastrophe: the wallet opens, and it is empty.
A derivation path mismatch. The old wallet used one default, the new wallet uses another, and you are looking at a real, valid, empty address. F103-L exists to make you produce this on purpose.
An inaccessible location. The safe you cannot open, the branch that closed, the relative who moved, the box whose key is in the box.
Every one of these is silent until you test. Three of the four produce the identical screen.
The wipe drill
This is F104-L2, and it is the second half of this course's practical assessment for good reason.
1. Put real but small funds on the device. Enough that the outcome matters emotionally. Not enough that a failure hurts.
2. Record the addresses you expect to see. Write them down separately, because your memory of an address is not a check.
3. Wipe the device. Deliberately. Factory reset, the real thing. This is the part where people's stomachs drop, and that reaction is diagnostic: if wiping a device with $50 on it feels frightening, you have just learned how much you were relying on the device rather than the backup.
4. Restore from the written backup alone. Only what you would have in a genuine emergency. Not your phone's notes, not a photograph you took "just in case", not the software still logged in on your laptop. If you reach for something that would not exist in the real scenario, the drill has already failed and you have found something worth fixing.
5. Confirm the addresses match and the balance is there.
If you use a passphrase, do the drill both ways: restore without it and confirm you get the empty wallet, then restore with it and confirm you get the real one. Seeing the empty wallet once, in a controlled setting, is the single best inoculation against panicking at it later.
A restored wallet showing nothing has four possible causes and they look identical. Work through them in order of likelihood, and do not skip to the last one.
1. Wrong or missing passphrase. By far the most common. Try without, try with, try variants including capitalisation and spacing.
2. Different derivation path. Change the path in the wallet's advanced settings, or restore into a different wallet application and see whether the funds appear.
3. Transcription error. A word wrong or two words swapped. If the phrase failed the checksum you will already know; if it passed, the wallet is valid and simply is not yours.
4. Actual theft. Check the address on a block explorer. If funds left in a transaction you did not make, this is F106-06 and you are in class 1.
The order matters because the first three are recoverable and the fourth is not, and people who assume the fourth stop looking for the first three. Notice also that this is precisely where the recovery scam waits, per F106-06's autopsy: a person who has just seen an empty wallet is the most receptive audience there is.
Cadence
Annually for the vault. Put it in a calendar. Twenty minutes.
Off schedule, after any change to the configuration. A new passphrase. A plate moved. A new device. A wallet migration. A house move. The drill validates a specific configuration, so changing the configuration invalidates the last result.
Once, immediately, on any new setup, before it holds anything meaningful. F104-03's final step.
Tier the effort with F104-02. The burner needs no drill. The daily deserves one at setup. The vault gets the annual discipline, and it is one seed.
The other drill
The second half of this belongs to F110, and it is worth naming here so you know it is coming.
The wipe drill proves you can recover. It does not prove anybody else can, and F110-01's whole subject is that most crypto dies with its owner.
F110-06 is the written plan, and its optional drill is a named heir recovering a test wallet from your document alone while you sit silently in another room. That fails in exactly the same instructive way this one does: at some small step you assumed was obvious.
Do this one first. It is the prerequisite for the other being meaningful.
I checked the words against the device when I set it up, so the backup is verified.
The setup confirmation and the drill test different things, and the difference is the whole lesson.
The device's confirmation step checks that what you wrote matches what the device currently holds. Genuinely useful, and it catches transcription errors at the one moment they are cheapest to fix.
It does not test the thing that actually kills people. It runs while the device is already initialised and already knows the seed, so it does not exercise the restore path at all. It cannot tell you whether the derivation defaults match, whether you will remember the passphrase in four years, whether you can physically reach the plate, or whether the process is one you can execute alone under stress.
Confirmation tests the transcription. The drill tests the system, which includes the software, the configuration, the location, and you.
Only 15 percent of holders in this course's survey had done the second one, and the survey's other finding is that 31 percent of those who lost access never got it back.
An untested backup is a hypothesis about your handwriting, your memory of the configuration, and your ability to reach a place, and none of those is verified by having written it down. Four failures are silent until you test: a transcription error, a forgotten passphrase, a derivation mismatch, and an unreachable location, and three of them produce the identical empty wallet. So wipe a device holding real but small funds and restore it from the written backup alone, using nothing you would not have in a genuine emergency. Then diagnose an empty wallet in order, passphrase first and theft last, because the first three are recoverable and the people who assume the fourth stop looking.