Data Recovery Case File · Formatted & Logical Faults · Partitions Are Real Boundaries

The Installer Wrote Inside One Partition and Nowhere Else

His enquiry describes a sequence with one very fortunate detail. A drive that failed after a freeze: "I have 3 partitions. I reinstalled Windows on the first. The majority of my data is on the second and third. The reinstall was successful, but the problem came back 2 days later. Afterward no reinstall was working. I bought another drive, replaced the faulty one" and set it aside. Reinstalling to one partition did not touch the others, and replacing the drive rather than persisting with it was the right call at exactly the right time.

The decode: what the partition boundary protected, and what the reinstalls revealed

Why the data partitions were not touched: a partition is a real boundary, not a label. An installer directed at one partition creates its filesystem, writes its system files, and operates entirely within that partition's defined extent. It does not write outside it, because everything it does is addressed relative to the partition it was given. So the second and third partitions hold exactly what they held before any of this started.

Why that is worth stating clearly: the archive is full of cases where a reinstall wrote tens of gigabytes over somebody's data — but those are almost always reinstalls onto a single-partition drive, or ones where the wrong target was chosen. A correctly targeted reinstall on a partitioned drive is one of the few destructive-sounding operations that is genuinely contained. His data was never at risk from the installer.

What the reinstall history actually tells us: something more useful than it appears. The first reinstall succeeded, the machine ran for two days, and then the fault returned — after which no reinstall would complete at all. That is a progression, and it describes a drive whose ability to read and write was declining across those days. The reinstalls were not failing because of software; they were failing because the medium beneath them was deteriorating.

Why that matters for expectations: the data partitions are intact in the sense that nothing overwrote them, and they sit on a drive with physical read problems. So the question is not whether the files are there but how much of the surface still reads — which is measured rather than assumed.

What he did right, and it is the decisive thing: he bought a replacement drive and took the failing one out of service. Most people in this position keep trying reinstalls, keep the machine running, and keep asking the drive to work. Every one of those attempts is sustained read and write load on a degrading surface. Stopping when he did preserved whatever margin remains.

What is done now: error rates measured first to establish how the failing regions are distributed, then imaging under strict per-sector timeouts with weak areas revisited on later passes — and capture directed at the data partitions first, since those hold what matters and the system partition holds nothing worth recovering.

On the bench

The original drive was kept out of service, the owner's decision to replace it rather than continue reinstalling having preserved the remaining margin. The Atola Insight Forensic assessed error rates to establish the distribution of failing regions — the progression from a successful reinstall to none completing describing a surface deteriorating across days. Imaging ran write-blocked under per-sector timeouts with weak regions revisited, capture prioritised to the data partitions since the system partition held nothing of value, and files validated by opening.

The outcome

Error rates measured first, the drive imaged under timeout control and capture directed at the data partitions. Free assessment, one fixed written figure including VAT; where a drive has to be opened, 50% of parts and labour is payable upfront with the balance only on success — otherwise no recovery, no fee. The decode: a partition is a real boundary and an installer directed at one writes only within it, so your data partitions were never touched. And the reinstall history is informative — one succeeding, then none completing, describes a surface deteriorating rather than a software problem. Replacing the drive rather than persisting was the decisive call.

Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.