Quick answer: Click Cancel and stop writing anything to the drive. The message „You need to format the disk before you can use it" and the RAW status don’t mean your data is gone — Windows simply couldn’t find a file system structure it knows how to work with. The safe route runs through a sector-by-sector copy onto another medium and recovery performed on that copy, not through formatting, CHKDSK, or scanning the original.
This text isn’t a guide to getting the drive working at any cost. It’s a decision procedure: what to do in the first minute, what can irreversibly overwrite your data, and when it makes sense to stop trying and hand the medium over for diagnostics. The difference between „Windows can’t see the file system„ and „the data is irreversibly gone" is enormous — and a single click turns the former into the latter.
The drive wants to be formatted: what to do right away
Windows offers to format the drive the moment it can see the partition but can’t mount it. That’s a fundamental difference from a drive that doesn’t show up at all. The message alone still doesn’t prove the data has disappeared; only the structure the system uses to find it may be damaged.
First steps:
- Cancel the offer to format, and every prompt to initialize the disk.
- Don’t save anything to the drive — not even one file „just temporarily."
- Don’t run
chkdsk, a volume repair, or a partition recovery tool. - Don’t reconnect the drive over and over hoping it’ll come up next time.
- If it clicks, scratches, disconnects intermittently, or reading visibly stalls, power it off.
- Note down what preceded the failure: a crash, a power outage, unplugging without ejecting, an update, working with partitions, pulling the drive out of a NAS.
That last point deserves no less attention than the others. „The drive dropped out of a Synology array„ and „the laptop fell off the desk" lead to two completely different procedures, and without that information the first hour of diagnostics is wasted groping in the dark.
What RAW means
RAW isn’t a file system like NTFS, exFAT, or FAT32. It’s the way Windows says „I can’t see a structure on this partition that I know how to read." Nothing more, nothing less — it says nothing about the contents of the platters or the NAND chips.
And only part of the metadata may in fact be damaged. NTFS keeps its key information in the volume boot sector and in the file table, the MFT; the partition layout is stored in the partition table (MBR or GPT). When the boot sector or an important part of the MFT gets damaged, Windows can lose its entry point and report RAW even though part of the file contents remains readable on the medium. NTFS also uses a backup boot sector and a $MFTMirr file, but their presence alone doesn’t guarantee a full recovery of names and directories.
The trouble is that very different causes trigger the same message — and each one tolerates a different amount of further powering on:
| What happened | Typical symptoms | Risk of powering it on again |
|---|---|---|
| Damaged file system | The drive reports the correct model and capacity but is RAW | Every write can overwrite metadata needed for recovery |
| Bad sectors on an HDD | Reads slow down, the drive disconnects, the system freezes | The defect spreads, reading further damages the surface |
| Faulty USB cable or enclosure | The drive comes and goes | Lower, as long as the drive makes no unusual sounds |
| Read-head failure | Clicking, no spin-up, repeated restart attempts | High — the heads can scratch the platters |
| SSD or controller failure | The drive vanishes, shows the wrong capacity or RAW | The state changes without warning; the drive can „die" between starts |
| Drive pulled from a NAS | Windows offers to format even though the drive may be healthy | Formatting damages a member of the RAID or SHR |
The last row is the most treacherous. With a perfectly functional drive from a Synology, Windows behaves exactly the same as with a dying HDD — and from the message the user can’t tell the difference.
Why not to click Format
Formatting isn’t a repair tool. It’s a write. It creates a new, empty file system structure and stores it exactly where the original one sat — that is, over the records that would let you trace file names, folders, and their locations.
Even a quick format has consequences. True, it doesn’t overwrite the whole volume sector by sector, but it creates new file system structures and removes the original file table and root directory. Depending on the type of the original and the new file system, this can overwrite the metadata needed for file names, folders, and locations. Recovery can then end up as a signature-based file search — with names like file0001.jpg and no original directory tree.
With an SSD, recovery can additionally be severely limited by the TRIM command. With it, the operating system tells the storage which areas it no longer needs; the device can then release them. Whether and when that happens depends on the system, the controller, the firmware, and the specific operation, so after a quick format no recovery time window can be promised. The safe approach is to stop using the drive immediately.
If you’ve already clicked Format, that’s not automatically the end. But stop using the drive immediately — every installation, copy, and further recovery attempt run directly on this medium shrinks what can still be saved.
When you can check the cable and when to stop powering the drive on
One controlled test — a different port, a different known-good data cable, or a different computer — makes sense only if the drive:
- spins up normally and stays connected,
- makes no clicking or scratching sounds,
- doesn’t disconnect during reading,
- doesn’t cause the system to freeze,
- wasn’t dropped or spilled on.
Connect straight to the computer, not through a monitor, a dock, or a passive hub. With USB-C, make sure the cable actually carries data — some cables in the drawer only charge. Why this connector can backfire like that is something we go into in the article why we don’t like USB-C.
If the behavior doesn’t change after swapping the cable, further testing at home won’t yield any new information — just extra strain. And with an HDD that has bad sectors, that strain isn’t neutral: every further attempt to read the damaged spot means more passes of the heads over the problem track.
If the drive clicks, restarts repeatedly, or changed its sound after a fall, power it off and don’t try anything else. The procedure for this situation is described in what to do when a drive clicks. When the drive doesn’t show up at all, an external drive won’t load or isn’t visible picks up from there.
Why not to run CHKDSK, TestDisk, or a recovery program on the original
chkdsk is a tool for repairing file system consistency, not for recovering data. On a volume reported as RAW it usually ends with a message that „CHKDSK is not available for RAW drives," and does nothing. But when it does start running — typically on a volume that’s still recognized as NTFS, just damaged — it works by writing: it repairs structures, frees records it judges to be invalid, and dumps the fragments it finds into a found.000 folder as file0000.chk files. From a recovery standpoint that’s the worst possible outcome: the data is still on the drive, but chopped up and without names.
The same logic applies to macOS First Aid and to fsck on Linux. On a healthy medium with a logical inconsistency they help. On a drive with bad sectors they add strain and writes exactly where you want them least.
TestDisk is a good tool, and the analysis itself doesn’t have to write anything. It writes changes to the partition table or boot sector only after you explicitly confirm the Write option; a wrongly confirmed write, however, can complicate any further recovery. Recuva, Disk Drill, R-Studio, or DMDE make sense for a purely logical loss on a healthy medium. But none of them should scan a physically failing drive directly: a deep scan means hours of reading the entire surface, which is exactly what an unstable drive may not survive.
The safe procedure therefore has two separate phases:
- First, create a sector-by-sector copy (an image) onto another storage device.
- Only then search the copy for partitions and files — with any tool, repeatedly, with no risk.
Unstable media are read with a tool that can skip a bad spot, come back to it later, read from the other direction, and keep a map of what’s already been read. A typical example is GNU ddrescue — and that map is the crucial part: it lets you interrupt reading and resume without the drive dwelling again on the spots where it already failed once. Even the best tool, though, won’t fix a drive with failed heads; that’s then a job for a cleanroom, not for software. What can and can’t be done with a damaged surface is shown in bad sectors on an HDD.
A drive pulled from a Synology or QNAP: a message that doesn’t mean a failure
This is the most common false alarm. Both Synology DSM and QNAP QTS build their volumes on Linux software RAID (md), with LVM on top and an ext4 or btrfs file system. Windows doesn’t read any of these layers natively. As soon as you plug such a drive into a computer, the system sees a partition it doesn’t understand — and offers to format it. Yet the drive may be perfectly healthy.
The brand on the label says nothing about it. WD Red, Seagate IronWolf, and Toshiba N300 all behave the same in Windows, because what matters is the data format, not the manufacturer. Which of them makes sense in a NAS is something we compare in the article WD Red vs. Seagate IronWolf vs. Toshiba N300.
With an array, rescuing a single drive in isolation isn’t enough. Reconstruction requires knowing the order of the members, the array geometry, and the state of all the media — and with Synology SHR the data usually isn’t in one contiguous RAID but split across several RAID groups on top of LVM, so a single pulled drive can’t yield complete files on its own, not even in theory. Don’t click initialize or any „Recover"-type prompt, don’t put the drives back in random order, and above all don’t start a rebuild on an array that’s already reporting a problem — why that’s a bad idea is covered in why a rebuild kills a RAID array. For specific situations, Synology: storage pool degraded and QNAP: RAID group degraded follow on.
What professional recovery of a RAW drive looks like
The first thing determined is what type of fault it even is: a logical error, a communication problem, surface degradation, or the mechanics. Diagnostics should answer at least this:
- Does the drive report the correct model and capacity?
- Can it be read stably, or does it disconnect during reading?
- Are unreadable sectors increasing, or is their number stable?
- Is only the partition structure damaged, or the hardware itself?
- Is the volume encrypted with BitLocker?
- Which data is a priority for the company?
That question about BitLocker isn’t a formality. An encrypted volume without the key looks like random data when read, and without the recovery key no one will get anything out of it — not even with healthy platters. Where to look for the key is described in BitLocker.
For a readable drive, an image is taken and the file system is reconstructed on it. For a physical fault, work on the electronics or the heads comes into play. Recovered data is always saved to another medium — never back onto the original drive.
It pays to set priorities in advance. An accounting database, in-progress project documentation, or current orders typically carry a different weight than an archive of installation files. A priority list matters most where the whole drive can’t be read — and with bad sectors that happens routinely.
What it costs and what to check beforehand
The price is set by the type of fault, not by capacity. A logically damaged, stably readable drive is a different league from an HDD with failed heads or an extensively damaged surface. With an SSD, the type of controller matters, along with encryption and whether a method for reading the memory outside the controller is even available for that platform.
Before you place an order, clarify:
- whether the diagnostics are paid and what exactly they include,
- whether you’ll get a price or a price range before recovery begins,
- what counts as a successful recovery,
- whether you’ll receive a list of the recovered files to check,
- what medium the data will be handed over on,
- how the provider handles confidential company data,
- what happens to the original drive and the working copies after handover.
No one can seriously set a reliable price from the RAW message alone — because at that moment no one yet knows whether the problem is in a single damaged structure or in the heads. What actually makes up the price is broken down in how much data recovery from a drive costs.
You’ll recognize a sensible approach by the questions, not by the promises. The provider should ask about the symptoms of the fault, about previous interventions, about encryption, and about the importance of the data. They should separate work on the original from file recovery, and they shouldn’t ask you to approve unlimited costs before diagnostics. A warning sign is a vague price with no explanation, a request for the password to an ordinary company account instead of a specific BitLocker recovery key — and above all a suggestion to „first try formatting it and then we’ll see."
What to prepare before handing over the drive
Write down how the problem showed itself and what has already been done with the drive — including attempts that didn’t work. A CHKDSK run you keep quiet about isn’t a time saver, just a dead end in diagnostics.
For an externally powered drive, include the original adapter too. For an encrypted device, track down the BitLocker recovery key, but don’t send it by unencrypted email. For a NAS, label the drives by their positions in the bays and don’t change the order — and don’t hand over only the drive the system flagged as faulty. Reconstructing the array usually requires the whole set.
Frequently asked questions
Should I click Format when Windows insists the drive can’t be used otherwise?
No, not if you need the data on the drive. Formatting writes a new file system over the structures used to trace the original names and folders. And the message says nothing about the drive’s health — even a perfectly functional drive pulled from a NAS triggers it.
Can data be recovered from a drive that reports as RAW?
Often yes, but it depends on the cause. With a damaged file system on an otherwise healthy medium, most of the data is usually readable. Bad sectors, head failure, SSD failure, or previous writes to the drive lower the chances.
Will CHKDSK fix a RAW drive?
It isn’t a data recovery tool. On a volume reported as RAW it usually doesn’t run at all, and when it does, it works by writing — it can dump the fragments it finds into a found.000 folder as files without their original names. Working on a sector-by-sector copy is always safer than repairing the original.
What if I’ve already quick-formatted the drive?
Stop using it and don’t save anything to it. On an HDD part of the data remains recoverable, but the original names and folder structure are usually damaged. On an SSD, recovery can be severely limited by TRIM — the released blocks are emptied and reads return zeros.
Why does Windows offer to format a drive from a Synology or QNAP?
Because Windows doesn’t read Linux partitions, ext4, btrfs, or the software RAID and LVM layer that NAS volumes are built on. So the prompt to format need not mean any failure at all — but if you confirmed it, you’d damage a member of the original array.