Blog
Rationalise before you migrate: why the application estate comes first
Windows 10 support ends on 14 October 2025, and the instinct in most organisations is to plan a migration for every application they can find. The evidence says most of those applications do not need to go anywhere. Retiring them first is the single biggest lever you have on cost and timeline.
The “just in case” migration
Ask a large organisation how many applications it runs and you will get three different answers: what procurement licenses, what the CMDB records, and what discovery actually finds on devices. The third number is always the biggest, and it is usually a shock. Faced with that list and a fixed deadline, the default plan writes itself: package everything, test everything, migrate everything, just in case somebody needs it.
It feels like the safe option. It is actually the expensive one, and with October approaching it may not even be the achievable one.
Every application you carry charges you three times
An application that crosses to Windows 11 is not free just because the installer already exists. It has to be compatibility assessed, packaged or repackaged for modern deployment, put through UAT, and then supported and patched for the rest of its life. If it is commercial software, it also has to be licensed, and renewal quotes are based on the entitlement you hold, not the usage you make of it.
Here is the uncomfortable arithmetic: an application nobody has opened in a year costs exactly the same to migrate as one your business runs on every day. Packaging effort, test cycles and licence spend do not know the difference. Only usage evidence does.
What the usage data actually shows
When our discovery tooling meters an estate over a sensible window, typically 30 to 90 days, the same picture emerges almost every time. A large share of the discovered software has not been launched at all in the period. Multiple versions of the same product sit side by side because upgrades never cleaned up behind themselves. Several different tools do the same job in different departments. And a long tail of applications has no identifiable owner, because the project that bought them closed years ago.
An application nobody has opened in a year costs exactly the same to migrate as one your business runs on every day.
The scale of the opportunity is not marginal. On one enterprise engagement our discovery and rationalisation work took an inventory of 26,000 applications down to about 1,000 that the business genuinely needed to carry forward. That is not an outlier produced by a uniquely messy estate; it is what decades of accumulation look like when someone finally measures usage instead of counting installations.
Rationalising with a defensible method
Rationalisation fails when it is done by opinion, because every application has a hypothetical user somewhere. It succeeds when every decision carries evidence. The method is straightforward. Build a single inventory from device-level discovery, then enrich it with metered usage, version data and ownership. Sort the estate into three piles: retire, where usage is zero or the function is duplicated elsewhere; consolidate, where several products do one job and the business can standardise; and retain, the applications that earn their place and get packaged, tested and migrated properly.
Two safeguards make the programme resilient. First, publish the evidence behind each retirement decision, so a challenge from the business is a conversation about data rather than a veto. Second, keep a rapid reinstatement path for the rare case where something retired turns out to be needed. Knowing an application can come back in days makes stakeholders far more willing to let it go.
The deadline is the argument, not the excuse
The common objection is that there is no time to rationalise before October. In practice the opposite is true: rationalisation is how you buy the time back. An estate cut by a half, or far more, means fewer packages to build, fewer test cycles to run, fewer licences to renew and fewer compatibility problems to chase. Every application retired in June is a workstream that never appears on the migration plan at all. And for any devices that do slip past the deadline into Extended Security Updates, a smaller estate shrinks that exposure too.
The sequencing matters because usage metering needs weeks of data to be credible. Started now, a rationalisation exercise delivers its evidence in time to shape the migration rather than trail behind it. Started in September, it becomes a report nobody has time to act on.
Windows 10 end of support is usually framed as a device problem: how many machines, how fast can we build them. The estates that land the migration calmly treat it as an application problem first. Rationalise before you migrate, and the migration you are left with is a fraction of the one you feared.
How big is your migration, really?
Our optimisation and rationalisation service puts usage evidence behind every retire, consolidate or retain decision, before the packaging bill lands.