Data Recovery · October 23, 2026 · ComputerKing Team

Clicking Hard Drive? DIY vs Pro Recovery: Vetting, Verifying

Your hard drive is clicking — should you try to recover it yourself or send it to a professional lab? This guide covers valuing your data, choosing a recovery provider, a pre-send checklist, and how to verify recovered files. It also covers replacing the drive and rebuilding backups under the 3-2-1 rule.

Summary

Your hard drive is clicking — should you try to recover it yourself or send it to a professional lab? This guide covers valuing your data, choosing a recovery provider, a pre-send checklist, and how to verify recovered files. It also covers replacing the drive and rebuilding backups under the 3-2-1 rule.

---

Brief recap: what the earlier guide already covered, not repeated here

This is the decision guide, and it assumes you have already stopped using the drive. The earlier guide covered three things in detail, summarised here in three sentences only: first, when you hear clicking, grinding, or beeping, power off immediately and do not keep trying to boot; second, different sounds may point to different causes, and a clicking sound does not necessarily mean the read/write heads have failed — an accurate diagnosis requires proper testing, not guesswork from sound alone; third, folk remedies such as the freezer trick, tapping the drive, opening it yourself, or swapping heads or circuit boards are consistently described in public sources as unreliable and potentially damaging, so do not do any of them. Everything from here on is about what to decide after you have stopped.

---

1. After you stop, assess three things first

Do not rush to buy recovery software, and do not rush to collect quotes. First, calmly assess three things. These three assessments decide which path you take next.

First, the current state of the drive. Write down when the sound started, whether the drive was dropped, exposed to water, hit by a power cut, whether there was a burning smell, what error messages appeared, how many times you have powered it on since, and what steps you have already tried. This information is genuinely useful to the people doing the diagnosis, so write it down now rather than relying on memory. Photographing the outside of the drive and the error messages on screen is also a good habit.

Second, whether the data has any other way out. Ask yourself: is this drive really the only copy? Is there another backup, a cloud sync, a copy sent to someone else, or an older version on another machine? If you already have a complete, verified backup, restore from it — there is no recovery fee to pay at all. If the data can simply be re-downloaded, such as films, music, games, or applications, it is normally not worth paying for professional recovery.

Third, what the data is worth. This is not about what the drive is worth, but about what happens if the data is gone. The next section explains how to judge that.

Only once these three assessments are done are you ready for the next decision: recover it yourself, or send it to a professional.

---

2. DIY vs professional: the triage table

This is the most important table in the whole guide. The dividing line is simple: is the drive itself physically healthy?

Two red lines deserve special emphasis, because old articles and videos online are most misleading on exactly these points.

Red line one: never open the drive, swap heads, or swap platters yourself. The gap between the heads and the platters is measured in nanometres, and a single dust particle is many times larger than that gap — enough to cause a head crash and permanent data destruction. Opening a drive in an ordinary room contaminates the platters with dust and will almost certainly make recovery harder, if not impossible. Platter swapping is even riskier: on a multi-platter drive, platter-to-head alignment is precisely calibrated by the manufacturer, and moving platters yourself can destroy any remaining chance of recovery.

Red line two: do not assume swapping in a circuit board from an identical model will work. On a modern hard drive, the circuit board holds adaptive data unique to that individual drive, including head flying height, positioning offsets, motor start parameters, and the defect map. Without transferring that data, a swapped board may refuse to initialise the drive, or worse, drive the heads into the spinning platters. In the worst case, incorrect calibration parameters can send a head straight into a platter spinning at thousands of revolutions per minute, scraping off the magnetic coating and causing irreversible physical damage. Conceptually, the correct approach is to transfer the original board's ROM chip contents to the donor board, which involves micro-soldering and specialist tools — not something an ordinary user can safely do. Board swapping sometimes worked on drives from before 2003; on modern drives it basically does not, and the exception should not be treated as "worth a try."

One more point: running recovery software on a physically faulty drive forces the drive to keep re-reading damaged sectors, which can cause electronic burnout or further platter damage. Software is for logical problems; mechanical problems belong in a lab. Do not blur that line.

---

3. Is the data worth recovering? Four questions to ask yourself

Professional recovery is not cheap, so before paying, ask yourself four questions.

First, irreplaceability. Is this data the only copy in the world? Wedding photos, baby photos, videos of family members who have passed away, years of research data, or work that cannot be re-shot or redone — these are usually in the "almost certainly worth it" category. Conversely, anything that can be re-downloaded or remade at any time is worth far less.

Second, business or legal consequences. Accounting records, customer databases, project files, or documents you are legally required to keep: the cost to count is not just the recovery fee, but the cost of redoing the work, the cost of downtime, and possible penalties or compliance consequences. Sometimes a recovery fee that looks expensive is trivial next to a week of stopped work.

Third, backups or alternative sources. This was covered above, but it bears repeating: if a complete backup exists, restore it and stop there. Only proceed to paid recovery when there is no backup.

Fourth, urgency and time cost. A delivery deadline, a court date, or a semester deadline raises the opportunity cost of "trying slowly yourself." But remember: urgency is a reason to stop and send the drive for assessment, not a reason to experiment. One wrong DIY attempt can end the whole opportunity.

One statement that must be made plainly: professional recovery never guarantees success. Even after paying, there is a risk that some or all of the data cannot be retrieved. Any claim of "definitely recoverable" made before the drive has even been seen should immediately raise your guard.

Warranty is not data recovery — a confusion many people make. Take Seagate's official limited warranty terms as an example: the limited warranty does not cover data loss, nor does it reimburse costs arising from data recovery, disassembly, or installation; unauthorised removal of components or the cover can void the warranty; and a drive sent in for warranty replacement is generally not returned. The practical meaning is direct: "the drive is still under warranty" does not mean "someone will pay for my data." Before sending a drive away for a warranty replacement, you must first decide what happens to the data — because once the original drive is gone, the data may be gone forever. The order is: assess the data first, deal with the warranty second.

---

4. How recovery is charged: diagnose first, quote in writing, start only with approval

4.1 The mainstream model

The model described publicly by multiple labs is the same: assess or diagnose first, quote a firm price, and start only after the customer agrees. Do not trust a fixed price quoted before the drive has been examined. When asking about price, do not just ask "how much" — ask each item:

1. How much is the diagnostic fee? How many days does it cover? Is it refunded on failure? Is it credited against a successful recovery?

2. Is the quote a fixed total, or itemised (diagnosis plus labour plus donor parts plus target drive plus shipping plus tax)?

3. How does "no data, no fee" define "success"? Does a partial recovery count as success? Which files count as the critical files?

4. How much is the rush fee — a percentage of the standard price, or a fixed amount?

5. How long is the quote valid? If platter scratching is discovered midway, will there be a revised quote?

Why ask in such detail? Because one lab's published terms demonstrate the key detail: if there is no prior written agreement to the contrary, a partial recovery of critical files counts as success. Before signing anything, you must list which files or folders are critical and define success in writing. That definition is the single most important sentence in the whole engagement.

4.2 Published overseas price lists (overseas prices only — never Hong Kong prices)

Every price below comes from an overseas lab's public price page, in US dollars, excluding tax and the target drive. They are shown only so readers can understand how charges are graded by fault type — not as Hong Kong market rates, and not for direct comparison with a Hong Kong quote.

Two structural lessons stand out. First, charges are graded by fault type: logical problems are cheapest, head swaps and platter damage the most expensive, differing by several times — there is no single flat rate. Second, some labs grade by capacity instead — Dave's, for example, states prices may be higher above 2TB. Do not mix the two models up.

4.3 Hong Kong prices: this research round found no published price list

To be honest about it: this research round found no citable published source for Hong Kong hard drive cleanroom recovery diagnostic fees or graded prices. The Hong Kong-related results found only say things like "expect to pay a diagnostic fee before a quote" or "free evaluation, charges depend on the nature of the fault" — none of them lists an amount. This guide therefore states no Hong Kong dollar price at all. For readers asking for prices in Hong Kong, the only reliable approach is: go by the written quote after diagnosis, and ask about four items — the diagnostic fee, donor parts fees, target drive fees, and failure charges. Any Hong Kong dollar "firm price" quoted before the drive has been seen deserves a question mark.

---

5. Choosing a recovery provider: six must-ask items plus decoding success-rate claims

5.1 Six must-ask items

1. Is the no-data-no-fee term written down? If nothing is recovered, is there no service charge? Are the terms in writing? Multiple guides list this as the first criterion.

2. Free or low-cost initial assessment plus a fixed quote before work starts. After assessment, the provider should state how much is expected to be recoverable, how long it will take, and what the fixed charge is — and only then start. Refuse a "mystery bill."

3. Work on a copy first, never experiment on the original. A reliable lab images or clones the faulty drive first, then recovers from the copy.

4. Cleanroom and whether the work is done in-house locally. Is physical recovery done in a certified cleanroom? Is it done locally, or handed to a third party or shipped overseas? Handing work on is not necessarily a problem, but it must be disclosed in advance — a quiet mark-up on a hand-off is.

5. Privacy and confidentiality. Is there a written non-disclosure agreement? How are recovered data encrypted for handover? How are working copies handled or destroyed afterwards? Is there a retention period? This matters most for business or sensitive data.

6. A file list to confirm before payment. Ask for a list or catalogue of recoverable files, check the most important files first, and only then confirm payment and handover. Do not judge by a total like "X GB recovered."

5.2 How to ask about cleanrooms so the answer actually means something

The industry standard commonly cited is ISO Class 5 under ISO 14644-1, formerly known as Class 100: no more than 3,520 particles of 0.5 micrometres or larger per cubic metre of air, expressed in the old system as fewer than 100 such particles per cubic foot. The smaller the ISO number, the cleaner the air — Class 4 is ten times stricter than Class 5. Why does it need to be that clean? Because the gap between heads and platters is extremely small, and a single dust particle can cause a head crash.

But "having a cleanroom" is not the same as "the whole room is a cleanroom." ISO 14644-1 itself covers cleanrooms, clean zones, and separative devices such as laminar-flow benches. What matters is the particle count where the drive actually sits — not the sign on the door or what the technicians are wearing. Ask three questions: in what certified environment is the lid-opening work done, and to which ISO 14644-1 class, with certificates or test reports available to see? Is the whole room compliant, or only the workbench area? And does this drive's fault actually require opening at all — if not, the charge and process should be different.

One more clarification: a cleanroom is not magic. It only provides a safe environment in which a technician can open the drive, swap parts, and then image it. The steps that actually save the data are the imaging and logical extraction after the swap — "it went into a cleanroom" does not equal "the data was saved."

5.3 Decoding success-rate claims

Never take any company's success-rate figure as fact. There are three angles to pick apart.

The denominator trap. Some companies only accept cases they believe are recoverable — for example, drives never opened, with no platter damage and no physical problems — which naturally inflates the rate. Others refer difficult cases elsewhere and then quote post-referral figures as their own. ADRC states plainly that over 90% of such claims are never entirely accurate: recovery rates depend heavily on the drive model, the failure mode, and what the user did afterwards, with some situations exceeding 95% and others only 20% or lower.

The numerator trap. Does a "partial recovery" count as success? As noted above, one lab's published terms state that, absent a prior written agreement, partial recovery of critical files counts as success. Another angle: some companies report the number of files recovered rather than whether those files are usable — a file name appearing on a list does not mean it opens or that its contents are complete.

The verifiability trap. WeRecoverData states openly that it does not advertise a single company-wide success percentage, because results depend on device type, the nature and extent of damage, prior attempts, encryption, media degradation, and the customer's own definition of the recovery goal — a single percentage without those variables is misleading. Counter-examples must be attributed: for instance, datarecovery.net's own website claims a 98% success rate across all failure types, and Wondershare's summary of Ontrack material cites a 90% success rate. Those figures are the vendors' own claims, not independently verified, and must not be written as industry facts.

Questions to put to the provider, one by one: what is the denominator of this percentage — all drives received, or only the ones you chose to take on? How is "success" defined? Is there a breakdown by fault type? Has the figure been verified by a third party, over what period, and with what sample size? Or is it just a slogan on a website?

5.4 Warning signs: stop and think if you see any of these

Vague quotes, a price quoted with no assessment, refusal to quote in writing; claims of "100% guaranteed recovery"; unwillingness to say who will actually handle the drive; demands for full payment before assessment. Any one of these is worth pausing over and getting a second opinion. A second opinion is itself reasonable: one lab saying "unrecoverable" does not mean the whole industry would agree — it may simply be that the first shop lacks the equipment. But before going to a second lab, stop using the drive and do not try anything else yourself, or each attempt will reduce what remains possible.

---

6. Pre-send checklist: follow it item by item

6.1 Do

6.2 Do not

Do not keep powering on, plugging and unplugging, or switching on and off to test your luck. Do not run chkdsk, formatting, initialisation, system reinstallation, partition repair, full scans, or recovery software repeatedly on the faulty drive. Do not open the drive yourself, put it in the freezer, shake it, or knock it. The earlier guide covered these in detail; here they appear only as the pre-send do-not list, not expanded again.

6.3 Packing and shipping

Place the drive in an anti-static bag — aluminium foil can substitute if no bag is available — and never wrap a bare drive directly in bubble wrap. Surround it with at least 2 inches of bubble wrap on all sides; for multiple drives, bag and wrap each one separately, never bundle them together. Use a sturdy double-walled cardboard box, fill all empty space so the drive cannot move at all in transit, seal all seams with H-tape, and mark the box FRAGILE. Do not use padded mailing envelopes or loose polystyrene beads — the protection is insufficient. Send it tracked and insured, include the engagement form and contact details in the box, and notify the provider of the tracking number.

---

7. After recovery: how to verify before you accept

7.1 Before payment: look at the list, do not just hear "everything is recovered"

Ask to see the recoverable file list before paying; some labs provide an online platform or a remote verification session so customers can confirm critical files before charges are settled. Check each item against the critical file list you made beforehand — do not judge by total file count or total capacity. A large number does not mean the critical files are complete and usable. If a critical file will not open, raise it before payment or sign-off, not after the original drive has been returned or the working image destroyed.

7.2 After delivery: open and check files by type

First step of acceptance: make a copy for verification purposes, and check files on that copy — do not edit the only recovered copy directly. Any repair tools must only operate on the copy.

Suggested acceptance order: most important files first, then large or frequently used folders, then the rest. Hash-value checks only apply when you already have old hash values or an old copy to compare against — without an old copy, do not assume a hash check proves anything.

7.3 Two things to do immediately after acceptance

First, fold the accepted data into a new backup straight away — the next section explains how to rebuild it. Second, confirm the lab's arrangements for its working copies (destruction or retention period), in writing for sensitive data especially.

7.4 Five decision- and acceptance-related misconceptions (beyond the earlier guide)

These five are limited to misconceptions related to decisions and acceptance that the earlier guide did not cover; the earlier guide's six misconceptions are not repeated here.

1. "Recovered means 100% complete" — wrong. Recovery never guarantees success, and even files that are retrieved may be only partially intact. When a file is damaged, recovery tools can only restore what is still on the disk; they cannot know what the original "should" have been.

2. "If the file name is visible, the file opens" — wrong. Recovery software can find a file name, size, and starting position from file system metadata, but the content may have been partly overwritten, damaged, or missing entirely — leaving a file that exists but will not open.

3. "If the thumbnail or preview looks fine, the whole file is fine" — wrong. A photo thumbnail may survive while the main image is damaged; a document's first pages may be fine while later pages, formulas, or embedded images are not; a video must be played at the beginning, middle, and end before it counts.

4. "If the file size is right, the content is right" — wrong. A reasonable size is only preliminary evidence, not complete proof. A zero-byte file definitely has no data, but a normal-looking file can still be internally damaged.

5. "A drive under warranty means the manufacturer will recover the data too" — wrong. Manufacturer warranty generally covers a replacement drive only, not data recovery costs, as covered in Section 3.

One more reality to accept: if some files will not open, first work out whether it is missing content or a format/application problem. The former is usually unrecoverable; the latter may be fixed with the original application's repair functions or specialist repair tools — but always try on a copy.

---

8. Replacing the drive and rebuilding backups: recovered does not mean repaired

8.1 The principle

A drive that has made clicking sounds should not go back into service as your main drive, even if this recovery succeeded. Replacing the drive and rebuilding backups is the proper finish — "recovered" does not mean "fixed." One more clarification: the handover medium (the new drive the recovered data arrives on) should not itself be treated as your only backup. It is a temporary handover drive; once acceptance passes, copy the data again immediately.

8.2 The 3-2-1 backup rule, made practical

The 3-2-1 rule means: 3 copies, 2 different storage media, 1 off-site copy, with the original file counting as one of the three. Practical steps:

1. Take stock. List what matters most — original photos, financial or legal documents, creative projects, anything that cannot be re-downloaded — then decide capacity and backup frequency.

2. Separate the media. A local external drive plus a NAS or another location plus cloud or off-site storage — never three copies on the same machine or the same account. "Two media types" means different kinds of media — for example, a local disk plus cloud backup, not just two hard drives.

3. Automate. A backup that depends on someone remembering to do it will eventually stop. Set an automatic schedule and keep multiple versions, so a damaged latest copy does not overwrite all good versions.

4. Lock one copy down. At least one backup should be offline or immutable with a retention period, so a single account cannot delete every copy — ransomware targets backups first.

5. Practise restoring. A backup that has never been restored is only an assumption. Regularly restore a few files to a temporary location, open them, and check content and dates; run a larger restore test at least quarterly and time it.

6. Write a one-page backup map. Where each copy is, how to log in, who is responsible — stored somewhere outside the protected system, and reviewed whenever you change phones, computers, or storage setups.

Three supplementary reminders: RAID is not a backup — mirroring only protects against a single drive failure, while accidental deletion or ransomware affects all copies simultaneously. A sync service is not an independent backup either — syncing propagates deletions and corruption, so an off-site copy is best kept as a backup service or a regularly disconnected external drive. And a backup must be restore-tested on a sample file, or it is effectively an unverified backup.

A simplified home version: working files on the computer, plus automatic backup to an external drive (unplugged once the backup finishes), plus cloud backup running in the background. Photos and tax records are the two categories most commonly brought in for rescue when only one copy existed — prioritise off-site copies for those two.

8.3 Choosing the replacement drive: HDD or SSD, CMR or SMR

HDD or SSD — choose by use, not by price alone. HDDs still win on cost per terabyte at large capacities, suiting NAS, media libraries, cold backup, and long-term powered-off storage; HDD writes have no TBW-style write endurance limit, and the usual cause of death is mechanical wear. SSDs have no moving parts, resist shock, and are fast; lifespan is measured by TBW (total bytes written), typical home write volumes fall far below rated TBW, and SSD failure modes differ from HDDs, making recovery more complex. The framework: SSD for the system drive or frequently used programs, HDD for large storage, NAS, and backup, with a hybrid (SSD for hot data, HDD for the backup layer) being a common setup. Lifespan figures must be stated conservatively: multiple sources put HDDs at generally 3–5 years, NAS drives at 3–7 years, and consumer SSDs at around 5–10 years — but all stress that this depends on usage, temperature, and vibration, so none of these is a guaranteed lifespan or an expiry date.

CMR or SMR — a must-check for NAS or RAID. CMR (conventional magnetic recording) writes each track separately; SMR (shingled magnetic recording) overlaps tracks to increase density, so rewrites must also rewrite neighbouring tracks, and sustained writes or RAID rebuilds can slow down dramatically. The consistent rule in 2026 guides: any RAID, ZFS, or NAS array setup should use CMR and avoid drive-managed SMR; SMR suits only cold storage with very few writes and mostly reads, not rebuilds. How to avoid the trap: do not judge by the brand series name alone — check the exact model's datasheet, because CMR/SMR, rotation speed, and cache can differ between capacities in the same series. CMR NAS lines listed in 2026 guides include the Seagate IronWolf series, WD Red Plus or Red Pro, and the Toshiba N300 series — but model tables change, so the manufacturer's latest datasheet at purchase time is the authority, and this list must not be treated as permanent. And once more: RAID is not a backup.

---

Frequently Asked Questions

Q1: Does a clicking sound immediately mean an expensive cleanroom job?

Not necessarily. Clicking is only a symptom of mechanical trouble; the exact cause requires diagnosis, and firmware problems, power problems, and head problems are handled differently and charged at different tiers. The mainstream model is diagnose first, quote in writing, and start only with approval, so the actual tier is only known after diagnosis. On Hong Kong prices, this research round found no published price list, so everything goes by the written quote after diagnosis — do not trust a fixed price quoted before the drive has been seen.

Q2: Should I try recovery software first, and only send the drive to a lab if that fails?

If the drive makes unusual sounds or shows signs of physical damage, that only makes things worse. Software scanning forces damaged heads to sweep the whole platter surface for a long time, accelerating surface scratching, and failed DIY attempts often make later professional recovery harder and more expensive. Software is only for logical problems, and even then it should work on an image, not on the original.

Q3: Why can't I just buy a matching head assembly or circuit board and swap it myself?

Heads require precise matching, specialist tools, and a cleanroom environment — opening a drive in an ordinary room already contaminates the platters. For circuit boards, each modern drive's adaptive data is unique to that drive; without transferring the original board's ROM contents, a swapped board may refuse to initialise or even drive the heads into the platters. Neither job is one an ordinary user can safely carry out.

Q4: A provider claims a success rate in the high nineties — can I trust that?

Ask about the denominator and the numerator. Is the denominator all drives received, or only the ones the provider chose to take on? Does "success" mean all critical files recovered and usable, or any file at all recovered? Is there a breakdown by fault type? Has the figure been third-party verified? Vendors' own claims — for example, datarecovery.net's self-claimed 98% success rate — are not independently verified and must not be treated as industry facts. An honest answer is a per-case confidence level after assessment, not a guarantee before the drive has been seen.

Q5: Does "no data, no fee" mean I pay nothing at all?

Not necessarily. Ask about the details: is the diagnostic fee charged separately, is it refunded on failure, are ordered donor parts refunded, does a partial recovery count as success, and how long is the quote valid? One lab's published terms state that once diagnosis has started the fee is non-refundable, and approved donor parts are non-refundable even on failure. Ask about each item and get it in writing before handing over the drive.

Q6: How do I verify recovered files before accepting them?

Before payment, look at the recoverable file list and check it against the Top 10 critical files you listed beforehand, item by item. After delivery, open files on a copy: read documents through to the later pages and formulas, view photos at full size, play videos at the beginning, middle, and end, and open project files in their original application to check layers and fonts. A visible file name, a correct-looking size, or an intact thumbnail none prove the content is complete.

Q7: The drive is still under warranty — should I just send it to the manufacturer for a replacement?

Not straight away. Manufacturer warranty generally covers a replacement drive only, not data recovery costs, and the original drive is generally not returned. Before sending it for a warranty swap, decide what happens to the data first — the order should be assess the data, then deal with the warranty.

Q8: After recovery, can I keep using the old drive?

Do not put it back into service as your main drive. A drive that has made clicking sounds should be retired even if this recovery succeeded. Replace it with a new drive and rebuild backups under the 3-2-1 rule: three copies, two media types, one off-site, with restore tests. The handover drive itself should not be your only backup — copy the accepted data again immediately.

---

Sources

• Rossmann Repair Group published pricing (USA, USD, overseas reference only)
• Dave's Computers October 2026 price snapshot (USA, USD, overseas reference only)
• MDrepairs pricing and terms (USA, USD, overseas reference only)
• Ontrack guide to choosing a data recovery company
• Ontrack Hong Kong: consultation, assessment, approved recovery, data return
• Stellar clean room data recovery explained
• Stellar recover data from a clicking hard drive
• Rossmann: do you really need a clean room for hard drive data recovery
• DriveSavers certified ISO Class 5 cleanroom
• 300 Dollar Data Recovery: understanding success rates
• ADRC recovery success rate explained
• WeRecoverData: why no single success-rate percentage is advertised
• eProvided: how to choose the best data recovery company
• Prime Data Recovery Ontario guide
• MDrepairs: why data recovery is expensive — value assessment framework
• PandaOffice: is it worth paying for data recovery
• Seagate official limited warranty terms: data loss not covered
• How-To Geek: why you cannot swap the PCB to fix your hard drive
• Rossmann PCB swap myth
• CBL fallacies about data recovery
• MDrepairs: how to ship a hard drive for data recovery
• DriveSavers shipping and packaging best practices
• Secure Data Recovery shipping and packaging
• Asoftech: why recovered files will not open
• Flyriver: how to verify recovered files
• GoodFirms: how to find a data recovery service
• WhereToRepair: data recovery explained
• 3-2-1 backup rule explained
• ComputingForGeeks: CMR vs SMR drives for NAS
• WunderTech best hard drives for NAS
• BestInHood Hong Kong data recovery comparison (triage reference only, no published prices)

---

Related articles

• Clicking Hard Drive? Power Off Now: 2026 Recovery & Myth Guide (HK)
• Hong Kong Backup Strategy Guide: The 3-2-1 Rule Done Properly — Time Machine, Windows, Cloud and Offline Drives
• Backup Is Not Recovery: Restore Tests and Immutable Backups