There is nothing to listen for. A solid state drive that has failed presents exactly as one that was never fitted, which is why owners spend a fortnight reading forums before they ring anybody. Flash work happens in the same building as disk work and has almost nothing else in common with it: no platters, no head stacks, no clean air. It runs here every week for households across the city and for the design studios, engineering shops and small practices around Warwickshire that quietly put everything they own onto storage with nothing moving inside it.
Looking at an SSD costs nothing. What comes out of that examination is one written figure, agreed between us before anybody picks up a screwdriver: £300 + VAT for one hard disk or SSD, an NVMe blade counting as a single drive.
Logical work is no fix, no fee. Four things sit outside that: electronic and mechanical failures, chip-level work, DVR jobs and forensic work, and anything physical is half up front. All five bands are laid out on the data recovery cost page, while data recovery services follows a postal job from the parcel to the answer.
Tracing a symptom back to whatever caused it is where the job genuinely starts, and these thirty cover very nearly every box opened on the Oxford bench. A fault that is not on the list is not an unfamiliar one: describe it over the telephone and you will get a straight reading of the odds before you have spent a penny on postage.
This is the limit that has to be stated before anything else on the page. When a file is deleted from a solid state drive, the operating system normally tells the drive those blocks are no longer wanted, and the controller erases them in the background within minutes without asking anyone. Once that has happened the contents are gone, not hidden. Nobody reverses it, here or anywhere. The exceptions are real but narrow, and the deleted file page sets out which ones they are.
A solid state drive is flash packages plus a controller that decides which physical page every logical address currently occupies. When the controller dies, that mapping goes with it, and memory in perfect condition becomes unreachable because nothing outside the drive knows the arrangement. Reaching it means addressing the packages directly and reconstructing the mapping from what comes off them, which is bench work rather than software.
A Phison controller has dropped into its boot loader because it could not load its own firmware, and the identity string it presents in that state is the same on every affected drive. It looks like a catastrophe and it is one of the more routine outcomes on this bench. Do not attempt the manufacturer reflash utilities circulating online: several of them rebuild the mapping tables, which finishes the job properly.
Another safe-mode identity. Several controller families fall back to a tiny reported size when they cannot validate their internal tables, and a 1TB drive appearing as a few megabytes is a drive that has protected itself rather than one that has lost anything. It needs the vendor diagnostic path, which is not the same thing as the vendor's consumer toolbox.
Some drives make this transition deliberately when the spare block pool is exhausted or an internal check fails, on the reasoning that a drive that cannot be written to at least still gives up what it holds. If yours is in that state, copy everything off it right now, in one go, before anything else. Read-only is not a stable condition; it is the last stage before silence.
A firmware flash that lost power partway through leaves a controller with an incomplete image and no way to start. The drive is usually invisible or reports a size of zero. This is recoverable in most cases through the production programming path, and it is one of the situations where the original drive matters, so do not throw it away in favour of the replacement.
Nothing in the firmware setup, nothing in Disk Management, nothing in Disk Utility. Try a second cable and a second port before assuming anything, because SATA cables fail more often than people credit. Beyond that, a genuinely invisible drive is a controller problem, and the memory underneath it is quite likely intact.
Documents that will not open, images with bands of colour through them, archives failing their checks, and no obvious grouping by date or folder. That usually means blocks are being retired faster than the controller can compensate, or that the error correction is failing on marginal cells. It gets worse with use and the useful response is to copy everything off today rather than to investigate.
Consumer drives generally have no capacitor bank to finish an in-flight write, so an outage can catch the mapping tables halfway through an update. The result is a drive that will not initialise. The stored data is normally still present; what has gone is the drive's own account of where it put it, and that account can often be rebuilt.
Nothing was changed, nothing was dropped, and the machine now looks past the drive entirely. Solid state failure gives no warning of this kind, which is why people spend a fortnight on forums before ringing anybody. Almost all of these are the controller. Take the drive out, stop using the machine, and let a diagnostic tell you which side of the line you are on.
Minutes to open a folder, a spinning cursor over ordinary work, and copying that stalls and resumes. A drive spending its time on error correction and retries behaves exactly like this. It is still readable while it is doing it, which makes today the right moment to take a full copy, and next month the wrong one.
Flash cells tolerate a finite number of erase cycles, and a drive that has been used as a scratch disk for video editing or as a database volume can genuinely reach that ceiling. It is rarer than the internet suggests, but on a drive out of a busy workstation around the Ansty or Prologis estates it is a real diagnosis rather than a theoretical one.
Some of the very cheap high-capacity drives are small drives with the firmware told to report a size they do not possess. They behave normally until the writing wraps around, at which point the earliest files are destroyed rather than merely lost. Nothing recovers what has been written over. A capacity test on a new drive costs nothing and settles the question before anything important goes on it.
Data is striped across the packages, so a single failure does not lose a neat fraction of the files: it damages a share of nearly every file on the drive. Some controllers carry a parity scheme across the packages and can rebuild what the failed one held, and where that scheme exists the outcome is far better. Establishing whether yours has one is part of the diagnostic.
Overtightened at the standoff, or flexed while the machine was being reassembled. A hairline crack through the substrate breaks tracks that are too fine to see, and the drive goes dead. Repaired at magnification it will often come back up for long enough to image, which is all that is required of it.
A drink into a laptop, or a machine that spent a night in a flooded unit. Flash survives water better than most components provided nothing is powered up while it is damp and nobody applies heat to dry it. The damage that finishes these drives is almost always corrosion under the packages afterwards, so the sooner it is opened and cleaned the better the odds.
Cold, it enumerates and reads. Warm, it disappears. That is a marginal solder joint or a component drifting out of tolerance with temperature. It is imaged under active cooling in short passes, and it is one of the situations where an owner trying repeatedly at home is genuinely making things worse, since each failed session ends in an unclean shutdown.
The volume mounts, has no usable size, and Windows offers to format it before use. Refuse. On flash this prompt normally means the file system header could not be parsed, and the allocated data behind it is untouched. Accepting is worse here than on a spinning disk, because a format on a modern drive can trigger the very background erase described at the top of this list.
The partition is still declared and its contents can no longer be identified. On solid state that is usually a mapping table that has been partly written, or a boot sector damaged during an unclean shutdown. It is worked from an image. Running a repair tool against the drive itself is the one action that can convert this into a genuinely unrecoverable job.
Once a sanitize or secure erase command has been accepted by the controller, the blocks are electrically cleared and there is no undo, no shadow copy and no laboratory technique that reaches them. If the command was interrupted early, some regions may not have been reached yet. That is worth checking, and it is worth being told plainly that it is a slim hope rather than a plan.
Most modern drives encrypt everything at rest as a matter of course, with the key kept in the controller, and this is invisible while the drive works. When the controller fails, reading the memory directly returns ciphertext. Sometimes the key can be recovered from the controller and sometimes it cannot, and which of the two applies is established during the free diagnostic rather than promised beforehand.
A migration that copied a sector range rather than a volume, or that was pointed at the wrong disk. The original is frequently untouched and simply no longer bootable. Send both drives together, label them so the bench knows which is which, and stop using the machine, because the usual next step is somebody reinstalling Windows on top of the failed clone.
A format on flash is a much worse event than a format on a hard disk, for the reason set out at the top of this list, and a week of ordinary use afterwards compounds it. It is still worth a look, because allocated data is not always what is lost, but this is the category where the free diagnostic most often returns an honest no.
Two drives striped together for speed in an editing machine or a small server, one of which has stopped. A stripe has no redundancy, so losing one member takes the volume down entirely. Both drives are needed and both are imaged before anything is assembled. Sets of this kind are the array band, from 500 pounds + VAT, and there is a page devoted to how striped sets fail.
Flash holds charge in cells, and cells leak, slowly, especially at higher temperatures and on drives that were already well used when they were put away. A solid state drive is not archival storage and should never be treated as a long-term copy. Where the charge has drifted, a low-level read with adjusted thresholds sometimes still gets it out.
A handful of families have shipped with counters that overflow at a specific number of powered hours and take the drive offline in an instant. If yours went dark at a suspiciously round figure, that is worth mentioning on the booking form, because the fix is a firmware path rather than a recovery at all and it is one of the quickest jobs on the bench.
Apple used proprietary blade modules between roughly 2010 and 2017, with pin arrangements that changed several times and no relation to the M.2 standard. Any of them that unclip can be worked here with the right adaptor. Later Macs, where the flash is soldered to the logic board, cannot be helped, and that is a firm no rather than a maybe.
An encrypted volume can be recovered where the key, the recovery password or the credentials can be produced, and it is charged at 400 pounds + VAT rather than the single-drive figure because of the extra work involved. Without any of those, an encrypted image is mathematics rather than engineering. Find the recovery key in your Microsoft account or your organisation's records before sending the drive, not afterwards.
A 2.5-inch drive that has been crushed or dropped hard enough to break the board. Whether anything comes back depends on the packages: if they are intact, they can be lifted and read directly, which is chip-level work at 50% up front and outside no fix, no fee. If the packages themselves are cracked, there is nothing left to read and you will be told so at no charge.
SMART is clean, the capacity is right, the drive performs normally, and a folder that held four years of work now holds nothing. That points at the file system rather than the hardware, and it is frequently one of the better outcomes on the page, because allocated data that has merely lost its index is exactly what a rebuild from an image is for.
On a healthy SSD with TRIM enabled, a deleted file is usually gone beyond anybody's reach. The operating system passes the drive a list of blocks it has finished with, the controller wipes them in the background as housekeeping, and reads of those addresses afterwards return nothing but zeros. No laboratory restores an erased flash page, and any firm implying otherwise is selling a story. That leaves a question of timing rather than technique. Delete something and shut the machine down inside a few minutes and there is a real chance; leave it running overnight and expect a realistic answer instead of an encouraging one, given during the free examination rather than after an invoice. None of this bears on a drive that has broken rather than mislaid a file. A controller that has stopped responding, a mapping layer that has fallen apart, a board with a cracked track — in every one of those the memory is untouched and the prospects are good. The deleted file page on this site covers the same ground for spinning disks, where the position is a great deal more forgiving.
Take the label off and there is a controller, a firmware layer and a row of memory packages. Failures concentrate in the first two. A controller that stops answering takes the drive off the bus completely, so the BIOS reports nothing at all. A mapping layer that has become inconsistent produces the odd capacities people ring up about: 8MB, 0MB, or a model name like SATAFIRM S11 that no manufacturer has ever printed on a box. In each case the memory behind it holds everything it held last week. The first route back is the diagnostic mode most controller families keep for use at the end of the production line, and it goes first because the drive stays in one piece. Where nothing answers at all, the packages are lifted off the board under controlled heat and read one at a time, after which the interleave, the page order and the error correction the original controller applied have to be reconstructed in software before a single file resolves. That second route is chip-level work, half up front and outside no fix, no fee, and which of the two you are looking at is established before you commit to anything.
Flash wears, though far less dramatically than the early scare stories suggested, and a consumer drive in a machine that is never switched off will reach its endurance figure long before anyone has planned for it. The pattern is recognisable. Slow first, as the internal error correction works harder. Then writes refused and a switch into read-only. Then one morning it does not appear. That read-only stage is a piece of luck rather than a disaster, because everything can still be copied off in a single pass, so copy it and retire the drive instead of hunting for a utility that claims to clear the flag. The other half of the story is retention, and it gets nothing like the attention it deserves. Cells lose charge whether the drive is working or not, and they lose it faster after heavy use. A drive that has lain unpowered in a drawer since 2019 is not an archive. Some arrive with material the error correction on the bench can still lift and some do not, and nothing about that improves with another year of waiting. Either way the figure is £300 + VAT, the same as a spinning disk.
Disk work and flash work happen in the same building and share almost none of the same equipment. Nothing here has to be opened to the air, so the filtered bench stands idle and the rework station, the package reader and the service terminal do the work instead. Everything below is read-only towards your drive.
Phison, Silicon Motion, Marvell, SandForce and Samsung controllers each have a diagnostic path used in production, and getting a stalled drive back onto that path is the first thing tried. It is also the safest, because it deals with the drive's own mapping rather than trying to reconstruct one from the outside.
Where the controller cannot be revived, the flash packages are lifted off the board under controlled heat, cleaned and read one at a time on a programmer. This is chip-level work: it takes 50% of the agreed figure up front and sits outside no fix, no fee, and you are told you are in that category before you commit to anything.
Raw flash is not files. It is pages, spare areas and scattered metadata, and turning it back into a file system means reproducing the exact page order, the interleave across packages, the XOR scheme and the error correction the original controller used. This is the slow half of every chip-level job and the part that decides whether it succeeds.
Cracked substrates, lifted pads and burnt rails repaired under a microscope, so that a board which is electrically dead can be brought back up for as long as it takes to produce one image. Nothing is repaired to be used again; it is repaired to be read once.
The same equipment that images mechanical disks earns its place here for a different reason. A drive on the edge of failure will stall for tens of seconds on a page it cannot correct, and hardware that can cap that, log the address and come back at the end is what turns an aborted copy into a complete one.
SATA, USB and M.2 all go through hardware write blocking. On flash that matters twice over, because an operating system given write access to a drive can trigger the background erase this page opens with, and a mount that lasts four seconds is long enough to do it.
Take that as a record of what comes in rather than a definition of what is accepted. Two names dominate, the Samsung EVO line and the Crucial MX500, because between about 2017 and 2020 they were the obvious way to buy a tired laptop another few years, and thousands of machines around Coventry and the Warwickshire towns were treated exactly that way. The A400 and the cheaper badges appear roughly in the proportions they sold in. A drive introducing itself as SATAFIRM S11 has a Phison controller sitting in its boot loader and is a routine morning rather than a catastrophe. The price is 300 pounds + VAT, identical to a spinning disk, and the assessment ahead of it is free and closes on the second working day after the drive is logged in. Taking the packages off the board is chip-level work: half in advance, and named among the published exclusions. Where the background erase has already been through the thing you are looking for, you get told so in plain words and you are charged nothing for the asking.
Almost everything worked on here arrived in the post. A drive that is already in trouble has an easier time boxed, padded and insured than it does being carried round in a bag for an afternoon, and a parcel handed over in Coventry today is normally booked in at Oxford tomorrow morning.
As a rule the storage comes out and the machine stays where it is. That applies to a laptop, a tower, an iMac and to the recorder sitting under a counter. Taking equipment apart is not something this bench does, and a repair shop will free a drive in a few minutes. Two things go the other way: an external drive stays sealed inside its own case, and a NAS travels as a complete unit with its disks still in their bays. A Fusion Mac is a third case — both of its drives come out and travel together, each one labelled. The single situation nobody can work around is memory soldered flat onto a mainboard, which is how Apple Silicon Macs and a good many slim laptops are built: if the storage will not unbolt, there is no parcel to send.
↓ Print the shipping & booking-in form (PDF)
Put Oxford Data Recovery on the label. From Coventry it is roughly fifty-five miles straight down the M40, about an hour if you would rather drive it in than post it. Either way you are told the moment it is logged, and the free diagnostic finishes two working days later.
Unsure what ought to go in the box? Ring 0800 689 0668 before you seal it, or work through the free online diagnostic and let it do the asking.
M.2 storage runs hot, then drops off the list altogether
TRIM settles most of this before anybody rings, though not all of it
Every repair tool writes, and on flash that is the worst move going
Platters give some warning first. Controllers rarely bother to.
Looking at it costs nothing, one written figure follows, and the band governing this page is £300 + VAT for one hard disk or SSD, an NVMe blade counting as a single drive.