Every data recovery company advertises a success rate. You’ll see 95%, 97%, “over 98%.” None of them tell you how it’s calculated, and that’s the problem — a lab that declines every hard case before accepting it has a spectacular success rate and tells you nothing useful.
The number you actually want isn’t a company average. It’s the odds for your specific failure. Those vary enormously, and here’s my honest read on them.
Logical failures — very high, if you stopped in time
Deleted files, formatted volumes, corrupted file systems, lost partitions.
These recover most of the time, and usually completely — filenames, folders, timestamps intact. The data was never destroyed; the map to it was.
The variable isn’t the failure. It’s what happened after. Deletion removes the pointer, not the file, and that stays true right up until something writes over the space. A drive powered down within minutes of the mistake is in excellent shape. The same drive after two weeks of continued use, or after a recovery tool got installed onto it, is a different case entirely.
The single biggest predictor here is how fast the drive got turned off.
Degraded drives — good, and it depends on where the damage is
Bad sectors, slow reads, CRC errors, accumulating S.M.A.R.T. failures.
Most of these recover the great majority of the data. A drive with a few thousand bad sectors out of two billion is still 99.99% readable — the question is only whether the unreadable fraction landed somewhere important.
Bad sectors in the middle of a video file cost you a few seconds of that video. The same number of bad sectors sitting on the MFT can obscure thousands of files at once, though those are often rebuildable from NTFS’s mirror copies.
The thing that hurts these odds most is prior repair attempts. A drive that’s had chkdsk /r run across it has been through hours of maximum-stress reads, and I see meaningfully worse outcomes on those. Why that happens.
PCB and controller failures — high, and often complete
Burned circuit board, dead USB bridge, failed power regulation.
When the electronics fail and the mechanics never did, the platters are pristine. Match the board, transplant the adaptive ROM where the model requires it, and the drive images at full speed like nothing happened.
The main risk isn’t recovery — it’s the failure taking the drive’s own PCB with it, or a well-meaning shuck of a hardware-encrypted enclosure putting the key material out of reach.
Free evaluation. No data, no charge.
Clicking and head failures — moderate to low, and I mostly refer these out
Clicking, beeping, grinding, heads that can’t read at all.
This is where the honest numbers drop and where advertised success rates get slippery.
Whether a drive with failed heads recovers comes down to platter condition. Heads that failed cleanly, without touching the surface, are replaceable in a cleanroom and often produce complete recoveries. Heads that scraped the platters have removed magnetic material — and no lab recovers what physically isn’t there anymore.
Both of those look and sound identical from outside the drive.
What decides which one you have is largely how many times it was powered on after the noise started. Every spin-up is another chance to convert the first case into the second. That’s why the advice on every clicking-drive page I write is the same one sentence: turn it off.
These are cleanroom cases and I refer them out rather than attempting work I’m not equipped for.
Near zero, and I’ll say so upfront
- Encrypted drives with no key. Not a difficulty problem. Working as designed.
- SSDs where the controller has failed and TRIM has run. Deleted data on a TRIM-enabled SSD is often genuinely gone, not merely unlinked. More on SSD specifics.
- Physically destroyed platters. Fire damage, deep scoring, shattered glass substrates.
- Overwritten data. One full pass over the top and it’s gone. The “you can recover overwritten data with an electron microscope” idea is a 1996 paper about drive densities that stopped existing two decades ago.
What moves the number more than anything
Not equipment. Not the lab’s marketing. These:
- How quickly it was powered down.
- Whether repair tools were run on it. CHKDSK, First Aid, format, initialize, “scan and fix.”
- Whether recovery software was installed onto the affected drive.
- How many times it was power-cycled after symptoms started.
Every one of those is a decision made by the drive’s owner in the first hour, usually before they’ve spoken to anybody. That’s why the first-hour checklist is the most useful thing on this site.
The bottom line
Logical failures usually come back whole. Degraded drives usually come back nearly whole. Electronics failures usually come back whole. Head failures are a genuine coin flip decided mostly by how many times the drive was powered on after it started clicking. And a handful of categories are honestly unrecoverable, which I’ll tell you at evaluation rather than after you’ve paid.
If a lab quotes you odds before they’ve seen the drive, they’re quoting you their marketing average, not your case.
Want a real read on your odds? Start a case — describe the symptom and everything that’s been tried so far. The second half matters as much as the first, and evaluation costs nothing.