Data recovery shortlist

0 · Decision tree — N old Macs, mixed condition

Cheap reversible checks first, escalate only when the tier fails.

1. Triage every Mac into (a) boots fine, (b) boots but flaky/slow/SMART warnings, (c) won't POST or screen dead.
2. (a) Bootable — clone the user-data volume to an external SSD with Carbon Copy Cloner in data-only mode, in parallel across machines. Don't bother making them bootable. Park the SSDs as "raw extract" before any destructive step.
3. (b) Flaky — image first, browse later. ddrescue via Homebrew (attach via Mac Sharing Mode / TDM / USB enclosure) to pull a sparse image. Mount the image read-only and triage what you actually need.
4. (c) Won't POST — first ask: Intel/no-T2 (pre-2018) → T2 Intel (2018-2020) → Apple Silicon (2020+)? That single answer determines whether DIY is still possible.
   • Intel/no-T2: pull the drive, USB-enclose it, back to bucket (a)/(b).
   • T2 Intel: SSD is soldered AND encrypted. DIY ends here. Ship to a board-level lab if data has any value.
   • Apple Silicon (M1+): same as T2 but worse. Lab work is the only option; success depends on whether the M-chip itself survived.
5. Consolidate onto the new Mac by mounting the cloned SSDs and pulling what you actually want — not by running Migration Assistant N times. MA onto a fresh Mac is fine; onto an already-set-up Mac is a nightmare (duplicate users, broken permissions).
6. Stop and decide at every step: is what's on this drive worth the next dollar of effort? Most old Macs have <5% genuinely irreplaceable data.

1 · Target Disk Mode + Mac Sharing Mode

TDM (hold T at boot, plug a cable, source Mac shows up as an external disk) is gold-standard for Intel Macs that won't fully boot but whose drives are healthy. Apple replaced it on Apple Silicon with Mac Sharing Mode — Recovery → Options → Utilities → Share Disk.

• T2 blocker: T2 Macs technically support TDM, but the host sees ciphertext. Decryption needs a functional T2 on the source. Dead logic board = no decrypt.
• Cable trap: USB-C charge-only cables (USB 2.0 in a USB-C jacket) silently fail Mac Sharing Mode — it hangs at "Waiting for other Mac". Rule: use a known-good Thunderbolt 3/4 cable.
• Apple Silicon caveat: Mac Sharing Mode shares the Data volume, not the System volume. User files only, not a bootable image.

2 · Migration Assistant — over-network / Thunderbolt / Time Machine

Reliability ranking (consensus): Time Machine restore > Thunderbolt direct > Ethernet > Wi-Fi. TM is most reliable because it's restoring from a known-good static snapshot.

• The Thunderbolt-default gotcha: even with a Thunderbolt cable plugged in, MA defaults to peer-to-peer Wi-Fi at ~20 Mbps. To get real Thunderbolt speeds (~400 MB/s) you manually configure fixed IPs on the Thunderbolt Bridge interface on both machines and enable Internet Sharing. One user reported 460GB in ~30 min with the manual config.
• What breaks: kernel extensions from pre-Catalina silently dropped; cross-major-version permissions need sudo chown -R newuser ~; duplicate usernames get a phantom 1 suffix.
• Best practice: migrate onto a freshly installed Mac (throwaway admin → MA → promote migrated user → delete throwaway), not one you've already set up.

3 · Clone-the-drive — CCC / SuperDuper / ddrescue / dd

Bootable clones are functionally dead on Apple Silicon. Mike Bombich (CCC author) says: "the new security of macOS makes it impossible to know if everything was copied, or copied correctly." Pattern: clone the Data volume only, install macOS fresh on the destination, migrate from the clone.

• Carbon Copy Cloner: data-only clones from healthy / mildly-flaky drives. Best in class for APFS. Pay for it.
• SuperDuper!: simpler UI, slightly behind on APFS. Fine for HFS+ era.
• ddrescue (Homebrew): the right tool for failing drives. Map file + retry-bad-sectors-last + re-runnable. CCC and SD will crash on read errors; ddrescue will limp through.
• dd: don't. No error recovery, no map file, one bad sector kills the run.
• Rule: if smartctl reports any reallocated sectors, skip CCC and go straight to ddrescue with a map file.

4 · FileVault unlock when the Mac is dead

Three recovery paths, in descending order of likelihood:

1. Personal recovery key (24-char string Apple showed at setup). If saved (1Password, paper), unlocks from any Mac. Only reliable path.
2. iCloud-escrowed key (if opted in). Apple can release after identity verification. Rising rejection rates for migrated/merged accounts.
3. Account password of an enabled user. Useless if Mac won't boot (FileVault unlocks before login).

T2 trap: on a T2 Mac, even with the correct recovery key, you need a functional T2 chip on the original logic board. The Secure Enclave does the actual AES; move the SSD to another T2 Mac and the new chip has no key material → ciphertext. No key + no escrow + dead Mac = data is gone — even DriveSavers can't break FileVault.

5 · T2 + Apple Silicon — "the Mac died, can I get my data back?"

Architectural shifts that determine everything downstream:

• Pre-2018 Intel: SSD is a removable NVMe/SATA stick. Logic board dies → pull SSD → enclosure → done.
• T2 (2018-2020 Intel): chips soldered, T2 holds AES key in Secure Enclave, key never leaves chip. Data bound to that specific T2 die.
• Apple Silicon (M1+): same as T2 but the M-series SoC is the secure enclave.

Apple Community summary: "If the flash chips are separated from the M1 chip, the data becomes unreadable encrypted hash. You cannot pull data off the internal drive unless the Mac can be repaired in a way that leaves both the M1 and the flash chips intact."

• BridgeOS corruption (T2 OS bricked, silicon fine) → recoverable by board-level shops.
• Cracked T2/M die, water-killed chip, fire → cryptographically destroyed. No lab recovers it.
• Power-delivery / GPU / display / USB-C controller died but T2/M is fine → recoverable by board-level repair (Rossmann-tier bread and butter).

6 · Physically failed drives — when to stop DIY

• Clicking, beeping, whirring on a spinning drive → power off immediately. Every spin-up grinds platters. Stop. Ship.
• Doesn't enumerate on USB (no system_profiler SPUSBDataType entry) → PCB-level or worse. Stop.
• SSD vanishes mid-write → controller failing. Hours not days. Image now or ship.
• Already tried more than two DIY approaches → every attempt makes lab work harder. Stop digging.

Cost reality: indie / board-level shops $100-2,000 (HDD) / $200-1,500 (SSD) / $200-2,500 (NVMe). Big-name shops (DriveSavers, Ontrack, Secure Data Recovery) quote $1,500-7,000 for the same work. Apple Silicon board-level: $800-2,500 at honest shops; $3,000-7,000+ at the big names. Get two quotes. If one is 3× the other for the same described work, the expensive one is paying for Google Ads, not skill.

7 · What recovery shops do that you can't

• Head-stack transplant (HDDs) — donor + ISO-5 cleanroom + sub-micron alignment. Period.
• PC-3000 firmware tricks — speaks to drive firmware directly, bypasses logical layers, rebuilds translator tables.
• DeepSpar Disk Imager — hardware imager, handles unstable drives better than ddrescue (faster head-park between retries).
• PCB swap with ROM transplant — modern drives store unique adaptive calibration in a ROM on the PCB. Swap without moving the ROM and you brick the drive.
• Chip-off / NAND reading — desolder NAND, raw-read with a programmer, reconstruct the FTL. Useless against T2/M-series encryption.
• Microsoldering / board-level repair — replacing the dead IC. Rossmann-tier; big-name shops outsource this.

8 · Picking a shop — red flags

• "Diagnostic before we can quote" + non-refundable fee → red flag. Honest shops give a range from your description, refine after seeing the drive, no fee.
• Non-refundable "engagement / research / down payment" → red flag. True no data, no charge shops keep nothing if they fail.
• Quote rises after they've started → hostage pattern. Get the full quote in writing before authorizing.
• Won't name who actually does the work → many "national" brands ship your drive to a sub-contractor; 3× markup for the middleman.
• Heavy Google Ads + "as seen on TV" → fine in itself but it's in your quote. Documented $150+/click on data-recovery keywords.
• Cleanroom-tier price for SSD work → SSDs don't need ISO-5. Anyone quoting cleanroom rates for soldered-NAND / NVMe is fleecing you.
• Refuses model-era flat rate for Apple Silicon → work is well-characterized; honest Mac specialists publish per-model pricing.

Green flags: published price tiers on the website; no data, no charge with no exceptions in fine print; willing to name the technician; willing to send photos of the drive on their bench; turnaround quoted in business days, not "as soon as possible".

9 · Call script for the first-pass diagnostic visit

• Bring: all Macs (or at least their drives if removable), the original recovery keys for any FileVault-encrypted ones, a hand-written list of what's worth recovering on each.
• Ask: free diagnostic? Per-drive flat rate to image healthy + flaky drives? Which referrals do you use for clean-room work? Will you give a per-drive quote BEFORE doing anything destructive? What's your "we got nothing" refund policy?
• Refuse to pay for: non-refundable diagnostic; cleanroom-tier price on an SSD; escalation without a re-quote; vague "as soon as possible" turnaround.

Caveat: WebSearch returned zero direct Reddit thread URLs (de-indexed since 2024). Synthesis above draws from MacRumors Forums, Apple Community, recovery-lab posts, and Reddit's standing positions. For primary sources, search inside reddit.com directly or use old.reddit.com.