What RAID recovery costs. In sentences, because a table would imply more certainty than a failed set allows.

A set is priced for its members, because every one of them is imaged and every image is part of the reconstruction. Two to four members that made up one set, a RAID 1 pair, a RAID 5 or 10 of four, a two-bay or four-bay Thunderbolt box, a four-disk Storage Spaces pool, a small mdadm or ZFS set, start at £500 + VAT after the free look, with the figure in writing before any chargeable work. That covers the imaging, the geometry, the assembly in software, the file-system repair on the virtual volume, and the extraction of virtual machines or the repair of a database where the set carried them.

Sets of five or more, 15K SAS sets, storage shelves of twelve or twenty-four members including the 520-byte aggregates from NetApp and EMC, vSAN and Storage Spaces Direct clusters, and enterprise storage such as MSA, Storwize and VNX start at £1,250 + VAT, rising with the number of members and with what was done to the set before it arrived. Priority handling, where the set goes to the front of the bench, is quoted separately and only when you ask for it. VAT is added to every figure at the standard rate.

One thing is not an array job. If the set is healthy and one member failed, and you want that member imaged onto fresh media so your own rebuild can run from a drive that answers every read, that is the drive site's job at £300 + VAT, and it is the commonest way to keep a degraded set out of the second band altogether. raiddrivedatarecovery.co.uk covers it, and the same bench does both.

The number of members, first. Then what was done before the parcel: a rebuild that ran into a second failure, a foreign configuration cleared, F2 pressed at a 1779 prompt, a drive forced online, a new virtual disk created and initialised, chkdsk or fsck run on a mis-assembled volume, a pool rewound with -F on the originals. Each is honestly described as more work, or as the end of the road, and each is quoted before you agree. On images, most of what those commands did is reversible; on the originals, the bench works out what they reached.

What does not move the figure: the level, the controller, the capacity, the file system, or a member inside the set that needs a head swap, a service-area repair or a 520-byte translation. A RAID 6 costs the same as a RAID 5 of the same size; a ZFS pool the same as an mdadm array; a set of SEDs with their keys the same as an unencrypted one. A set another laboratory has already worked on, or drives that have been opened at home, sit outside the no data, no bill promise, and you are told that at the outset.

Logging, a number on every member matched to its slot, photographs, and a read-only assessment on the imager each drive's interface needs, never on a controller: the metadata read from every member, the level, the order, the stripe size, the event counts that say which members are current and which dropped out first, and the controller's own log where you sent one. We identify what happened to the set, what has been written since, and how much can honestly be read back. You then get an account of what we found, what we expect to recover, and one price. If you say no, the drives come back to you at our expense, and nothing is owed.

The vendor's rebuild under warranty works when every surviving member is healthy and you have a backup: the engineer fits the new drive, the controller rebuilds onto it from clean survivors, and the failed drive leaves with the courier. On a set whose survivors were bought the same day as the drive that failed, the rebuild reads every sector of a drive that is about to do the same thing, and the failed drive that left with the courier was the only place the stripes it stopped on still existed. The bench does the same rebuild on images, which cannot stall, and keeps the failed drive until the set is whole.

RAID recovery software has a narrower place than its advertising suggests: every member readable on the host you have, the metadata intact, the geometry known or guessable from a preset list, and the software run against images rather than the drives. It cannot image a weak member, cannot read a 520-byte drive, cannot unlock a SED, cannot prove a RAID 6 by Q, and against a set with a failing member every scan is thousands of extra reads of the weakest surface. The free look is where you find out which yours is, and it costs nothing.