A computer that no longer lists its disk is nearly always sitting beside a disk that still holds every file ever put on it. That is the ordinary ending rather than the lucky one. Drives reach the bench from the city itself, from the units at Ansty and Whitley, and from Rugby, Bedworth, Kenilworth and the market towns past them. Finding out which fault you have costs nothing, and the figure that follows is written down before anybody picks up a screwdriver.
Looking at a hard drive 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.
Sweep, pause, reset, sweep again. What you are hearing is the actuator hunting for the servo marks that tell it where on the surface it currently sits, failing to lock onto any of them, and starting the search from scratch. Every repetition is another pass of a head across a recording surface it can no longer read. There is no version of this where the tenth attempt succeeds.
Something solid is now touching a surface with an outer edge moving at roughly the speed of a car on a motorway. The coating comes away as powder, the powder circulates inside the sealed chamber, and from then on it works like grinding paste on every remaining head. Minutes matter here in a way they do not anywhere else on this list. Pull the plug.
No spin, no vibration through the casing, no warmth after a minute. That single description covers at least three unrelated faults: a spindle bearing that has seized solid, a head stack resting on the platters and welded there by time, and a board that has stopped delivering current. They need three different repairs, and nothing you can do at a desk will tell you which one you have.
The motor reaches full revolutions, the controller goes looking for its own firmware modules in the reserved area of the platters, cannot read a clean copy of one of them, and drops everything to avoid making the situation worse. That behaviour is often reported by disks whose recording surfaces are in fine condition. It is answered with a service terminal, not a screwdriver.
The disk takes power and answers when the machine speaks to it, but the name it gives back is a run of symbols, a string of the same character repeated, or an empty field. That is the firmware area talking rather than your files. No consumer utility has ever been able to reach a firmware area and none is going to be written that can.
The translator has collapsed. That is the internal map converting the addresses your computer asks for into physical positions on the platters, and without it the disk genuinely does not know it holds anything, so reporting zero is an honest answer. Rebuilt on the bench, the real figure returns and everything behind it returns with it.
The surface is degrading and the disk is quietly shifting what it can onto its spare tracks. Each week the counters go up. Kept in service, it carries on until the spare pool runs out. Here it comes off power immediately, gets one imaging pass with the difficult zones deferred to the end, and every subsequent step happens against the copy rather than the original.
The heads have reached a band they cannot read and the disk is retrying it, several seconds a sector, indefinitely. Ordinary file copying has no mechanism for giving up gracefully and no way to come back later. Imaging hardware has both, and that single difference accounts for a large share of what a laboratory gets back that an owner could not.
Heads fly a few nanometres above a surface running at 5,400 or 7,200 revolutions a minute. A knock closes that gap. What decides the outcome is what happened next: a disk switched off immediately usually needs a matched donor stack, while one that was powered up three more times to see whether it had survived usually needs that and a scored platter explained.
A surge through the mains, or a failing power supply that let a rail run high. The damage is almost always confined to the electronics, and the recording surfaces underneath are untouched. It is repaired at magnification, component by component, with the disk's own tuning data carried over onto whatever board ends up fitted.
Most drive boards carry small clamping diodes designed to fail closed and short a rail to ground rather than let a spike travel further in. When one has done that, the disk is dead and the diode has done exactly what it was built to do. Removing it and testing what lies behind is a five-minute job. Guessing instead, and putting mains-borne damage through a second board, is not.
This is the commonest self-inflicted delay on the whole page. Every modern disk stores calibration figures for its own particular heads in a chip on its own particular board. Fit a board from another example of the same model and the disk spins, answers, and reads meaningless data, because it is being handed the wrong numbers for its own mechanism. The chip has to be transferred.
Stiction. A disk left standing for long enough can end up with the head sliders chemically stuck to the surface they came to rest on, and the motor no longer has the torque to break them free. Some of these arrive from house clearances around Bedworth and Atherstone with two decades on them. Freed under filtered air with the correct tooling they very often read in full.
Sudden power loss can catch the heads over the data area instead of retracted onto their ramp at the edge. They settle where they are, and the next attempt to spin the disk drags them across the coating. Any disk that has been dropped or lost power abruptly and then refused to start should be treated as this until somebody has looked, because one more attempt is what turns it into the fault above.
The preamplifier sits on the actuator arm itself, inside the sealed chamber, and it serves all the heads together. When it fails, the disk loses every head simultaneously rather than one at a time, which is why a mechanism that was perfectly healthy a second earlier can go completely blind. The repair is the same as a head replacement, because the part is carried on the head assembly.
The high-capacity units, generally eight terabytes and above, are sealed with helium rather than filtered air because the thinner gas allows more platters and less turbulence. If the seal goes, the aerodynamics the heads depend on change, and the disk degrades in a way that ordinary drives do not. These are worked in a controlled environment and they are one of the strongest arguments for not waiting to see whether it settles down.
The spindle bearing is wearing. It starts as a note slightly deeper than it used to be, becomes a resonance through the desk, and finishes as a disk that cannot hold a steady rotational speed for long enough to read a track. The window between the first symptom and the last is measured in weeks, and it is the one mechanical fault that does give some warning.
The disk mounts, gets a letter, reports no usable size and asks to be formatted before use. Decline it. That prompt appears when the file system header cannot be parsed, which is a different matter entirely from the contents having gone. Accepting writes a fresh empty structure across the index describing where every file actually lives.
Windows offers to check and repair a volume it cannot read, and on a disk that is failing physically the repair spends the drive's remaining hours writing corrections onto a surface that cannot hold them, discarding directory entries as it goes. What that costs, and what is recoverable afterwards, is set out on the CHKDSK page rather than repeated here.
Somebody started copying an old disk onto a new one, the software hit a bad region, and the job was cancelled. The destination now holds a partial file system and the source has been read hard for several hours. Both are worth sending together, because the partial copy frequently supplies structure that the original can no longer produce.
The partition still exists and the operating system can no longer identify what file system is inside it. Behind that are three ordinary causes: a boot sector damaged by an interrupted write, a partition table edited by an installer, or a surface fault sitting exactly where the file system header happens to live. All three are worked from an image, never from the original.
Say no. Initialising writes a new partition scheme onto a disk whose existing scheme is merely unreadable, which converts a recoverable job into a much slower one. The pattern usually means the first sectors are unreadable, and the disk itself may be in perfectly good order everywhere else.
A disk written while it sat behind a USB bridge doing sector translation has to be read back the same way, or every offset lands in the wrong place and the volume looks like nonsense. It is a five-minute question on the bench and an unanswerable one at home. If the caddy still exists, send it in the box with the disk.
It mounts, copying starts, and somewhere around the ten-minute mark the disk disappears and the machine has to be restarted. Thermal drift in a marginal board, a weakening motor, or a head that only degrades once it is warm. It is imaged in short cooled passes, which is slow and which recovers a great deal more than a single long attempt that fails at the same point every time.
A machine that survived an outage but has been slow since is often carrying a disk that took a hit to its firmware or lost a chunk of its defect map. It works, after a fashion, but it is spending its time recovering from errors rather than reading. Back it up today and have it looked at, because this is a condition that does not stabilise.
The identity comes back correctly, the capacity looks right, and then every command times out. On the Seagate families that is generally the service area rather than the mechanism, and it is routine bench work with the right terminal. There is a fuller account of the Seagate ranges and the faults specific to each on the Seagate page.
Some WD families accumulate relocation records in the service area until the firmware spends nearly all its time consulting the list. The platters can be in excellent condition while the disk reads at a fraction of its rated speed. Cleared on the bench, it images normally. There is nothing an owner can do to it from a running operating system.
Shingled recording overlaps the write tracks like roof tiles to gain capacity, and rewriting anything in the middle means rewriting everything after it. Several very common desktop and NAS models are built this way. They are slow to image even when nothing is wrong, and painfully slow when something is, so the timescale on one of these is longer for a reason that has nothing to do with how busy the bench is.
A machine went in for a fix, came back working, and the documents are gone. What is left depends entirely on which route the installer took and how much has been written since. There is a page on this site devoted to exactly that situation, and the useful instruction in the meantime is to stop using the computer altogether.
Two disks that were kept identical, one of which dropped out of the set weeks before anybody noticed. The surviving member carries current data and the removed one carries a snapshot of whenever it left. Send both. Which of the two holds what you actually want is a question the bench answers by reading them, not by looking at the labels.
Take the power off first and read the rest afterwards. Inside the case, glass platters turn at several thousand revolutions a minute while the read heads ride a few nanometres above them on moving air. Robust enough while everything is flying, and fragile the instant one head touches down. The tick described on the telephone is the actuator parking and swinging out again, over and over, because the servo pattern it steers by will no longer resolve, and each pass takes a damaged head over more of the surface your files are on. So the least technical instruction on this page is the one worth acting on: switch it off, leave it off. Searching for somebody to ring at midnight is understandable and the answer will keep until morning. What follows runs at the opposite speed. Filtered air, a donor head stack matched to your model and fitted by hand, a seized spindle released where its bearing has gone, then an unhurried read that banks the cooperative regions first and returns to the awkward ones at the end. None of that changes the price. Any single hard drive is £300 + VAT, and where the fault proves mechanical or electronic, half is payable before work starts, because the donor parts are ordered against your specific drive and not held on a shelf for drives at large.
Everything that arrives here belongs to one of four groups, and the group decides the method, the odds and the terms. Mechanical failure means the heads or the motor: clicks, scrapes, a drive that reaches speed and thinks better of it, or one that will not turn at all. That is filtered-air work with donor parts. Electronic failure means the board on the underside, and it turns up more often than people expect after a storm or a tired power supply, because a small protection component is built to short itself out and absorb the surge rather than pass it into the mechanism. Swapping a board is not a matter of buying the same model, since the calibration data unique to your drive lives in a chip that has to be carried across by hand. Firmware failure is the odd one out: a perfectly sound mechanism reporting the wrong capacity, or none at all, because the service area holding the drive's own tables has become inconsistent. Nothing has happened to your files in that case. Logical failure is the fourth and much the commonest — a deletion, a format, a partition table that no longer parses — and it is the group no fix, no fee covers. The other three are physical work and carry the deposit the published exclusions describe.
Seagate, Western Digital and Toshiba account for most of what comes through the door, and the first two have pages of their own on this site because their habits differ enough to be worth separating. Seagate work splits between burnt boards, translator faults that leave a healthy mechanism insisting it holds nothing, and shingled desktop drives that image slowly because their tracks overlap by design. Western Digital fails somewhere else: a service area silting up with relocation records until a sound disk reads at walking pace, or a fall that calls for a matched head stack. Behind those two sit names that stopped being companies years ago. Seagate absorbed Maxtor in 2006 and took on the Samsung disk line in 2011, which leaves a DiamondMax or a Spinpoint in a family that has had no new parts made for it in well over a decade. Deskstar and Travelstar started life at IBM, went across to Hitachi, spent years badged HGST and ended up inside Western Digital. The practical consequence is short: donor stock for old drives is thinning, and a disk that has started to misbehave is worth posting this month rather than next year. The figure is identical whichever badge is on the lid, and the examination costs nothing whether the drive is two years old or twenty.
Nothing on this list is here for show. Each bench answers a class of failure that cannot be reached any other way, and between them they account for what actually arrives in Oxford from Coventry and the Warwickshire towns. None of it writes to your disk at any point in the process.
Head replacements, platter transfers and lifting a stack that has settled onto a surface are all done inside a filtered enclosure with the disk supported and the internal parts handled by tooling rather than by hand. At the scale a head flies, a speck of household dust is a rock in the road, which is why a disk opened on a kitchen table rarely reads afterwards.
Translator rebuilds, defect list surgery, modules read back out of the service area, a hung boot process interrupted and stepped through. This covers roughly half of the disks that arrive looking mechanically dead, and none of it is visible to an operating system, because the area being worked on is not part of the capacity the machine can see.
Copies are taken head by head with current draw watched throughout, and retry behaviour is limited at sector level by the imaging hardware rather than by the disk's own firmware. A region that would occupy an ordinary copy for forty minutes is skipped, logged and revisited at the end, once the straightforward ninety-odd per cent is already safe.
Purpose-built tools for lifting a head assembly clear of the platters without letting the sliders touch anything, and for moving an entire platter pack from one chassis into another with the alignment between the discs preserved. Improvised versions of these tools are the reason some disks arrive already beyond help.
Shelved by model, firmware version and factory code, because a head stack that is nearly right is worth nothing at all. Most jobs find their match on the shelf on the day. When one does not, the wait for the correct donor is the honest reason a date moves, and you are told that rather than left guessing.
Burnt components replaced, lifted pads rebuilt, damaged tracks bridged, and the disk's own tuning figures moved across onto whichever board finally goes on. That transfer is the step most home board swaps miss, and it is the difference between a disk that reads your files and one that reads convincing nonsense.
Everything your media is plugged into has a hardware blocker between it and the machine. From the moment a parcel is opened to the moment it goes back out, nothing on your disk is altered, no journal is replayed and no volume is mounted for writing. The work happens against copies, which is also why a failed attempt costs you nothing but time.
That is a description of the post, not a definition of what can be worked on. Two part numbers lead it, the WD40EZAZ and the ST4000DM004, because more Midlands towers and cheap external cases were built around those two than anything else of their era. The IronWolf and Red units come out of four-bay boxes in spare-room offices around Kenilworth, Warwick and Leamington. SkyHawk disks come out of recorders that have been writing without a pause since the installer packed up his van, and recorder jobs appear by name in the exclusions from no fix, no fee. Travelstars and old Deskstars turn up when a house is cleared, and they read perfectly well once their heads have been persuaded off the surface. Whatever the badge, one disk is 300 pounds + VAT. Looking at it is free and finishes on the second working day after it is logged in, and the recovery itself takes a further two to four working days from the moment you accept the figure. Logical faults are covered by no fix, no fee; mechanical and electronic ones want half in advance and are excluded from it. PATA is still worked on, if the disk has spent twenty years in a shed.
Between them, Seagate and Western Digital account for most of the booking sheet, and each has been given a full page on this site rather than a paragraph in the middle of this one. Three places tend to go on a Seagate: the electronics, the firmware translator, and the overlapping write tracks of a shingled BarraCuda, which image slowly because of the geometry rather than because anything is wrong. Donors for them are filed by firmware revision, since on this maker the revision matters far more than the capacity printed on the front. Western Digital arrives with different complaints. Either the reserved area fills up with relocation records until a mechanically sound disk reads at a crawl, or something was dropped and matched heads have to go in under filtered air. Read the colour on the lid as a life history: Blue lived in a family tower, Red in a NAS, Black in a workstation, Purple in a recorder, Gold in a rack. Whichever it is, it is one disk at 300 pounds + VAT once a free assessment has said what is wrong with it.
Which Hitachi turns up, and what it needs when it does, follows a fairly reliable pattern. Notebook Travelstars of the 7K1000 generation come first by some margin, then Deskstar 7K3000 desktop units of much the same age. Stand either of them in a cupboard for a few years and the sliders bond to whichever part of the surface they were resting on, until the motor no longer has the torque to break them free. Neither is especially difficult to deal with under filtered air. Ultrastar disks pulled out of small servers show up with hour counts in six figures and, so long as the seal has held, produce beautiful images. The approach is the same across all three families: take one surface at a time, secure everything the sound heads can still reach, and only then bring a donor stack in to cover the remainder. Parts for these lines are bought and shelved in advance instead of chased when a job appears, because each year they are a little harder to find. That shelf is precisely why a disk made in 2004 gets the same treatment as one made last year.
This badge arrives in bigger numbers than people expect, mostly because Toshiba carried on supplying laptop makers after several rivals gave the business up. Its MQ01ABD and MQ04ABF100 mechanisms shed head performance in stages instead of all in one go, which is why the owner spends a couple of months thinking the computer has got slow before there is ever a noise to hear. The X300 desktop line comes out of home builds and small workstations, and anything sold under the Dynabook name has the same mechanisms in it with different lettering on the lid. There is a practical conclusion, and it is worth acting on rather than filing away: on the day a Toshiba begins to limp it can usually still be read from end to end, and four weeks later a good deal of that has gone.
Maxtor was absorbed by Seagate in 2006 and the Samsung disk business followed in 2011, which means a DiamondMax or a Spinpoint F3 belongs to a product line nobody has manufactured a spare part for in well over ten years. Only one consequence really matters, and it is a supply problem. A Spinpoint that starts to falter today can end up sitting on a shelf waiting for a match to appear, and those shelves get emptier every year. Fujitsu notebook mechanisms are in the same position, as are the Quantum Fireballs that occasionally come out of a garage in Nuneaton still attached to an IDE ribbon. Age has never yet been a reason to give up on what was recorded, but a disk with no supply chain behind it is one to send while there is still something to match it against.
Part the storage from the machine before the box is made up, because a bare drive is all that should reach Oxford. Beaten by the screws? Ring 0800 689 0668 and somebody will walk you through it, or let a repair shop on almost any high street free it for a few pounds.
Nothing but the bare drive is taken in here, so the disk has to come out of the machine before the parcel is made up; dismantling computers is not part of what is on offer. Four screws and two plugs covers most towers, and any repair counter will do it in minutes for a few pounds if you would rather not. The address is Oxford Data Recovery, John Eccles House, Oxford Science Park, Robert Robinson Avenue, Littlemore, Oxford OX4 4GP, and it wants a tracked, insured service. Coventry to that door is about fifty-five miles of M40, an hour or so, and if you would sooner not trust it to the post then reception takes it by hand between 9:00am and 5:30pm on any weekday. Book your own courier if that suits better. Nothing is collected anywhere in this network and there is no Coventry counter to walk into, so the journey south is your part of the arrangement. Three things do not follow the drive-only rule: an external in its own enclosure goes whole, a NAS goes whole, and a Fusion Mac has to arrive as both of its drives with a label on each. RAID sets and servers send member disks alone, each one marked with the bay it came from. If you would like the packing checked before it goes, ring 0800 689 0668 during those hours. Two limits, stated without hedging: nothing can be done with storage soldered onto a mainboard, and phones and tablets are not accepted at all.
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.
The service set out in full, for readers anywhere in the country
Blue, Black and Red disks, and the ones sold inside a plastic case
BarraCuda, IronWolf, Exos, and shingled surfaces that rewrite slowly
Steadying a disk to copy it is not the same as putting it right
Prices, timings, and how a disk from here reaches the Oxford bench
A scan that writes is the last thing a dying disk needs
The name goes immediately; the contents leave a good deal later
What moves when a job is marked urgent, and what will not move
The county page: ten towns, one bench, everything by tracked post
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.