Azure doesn’t fix what you move into it.

Detailed summary

The four stages between a legacy estate and a working Azure platform, and why most of the work happens before anything moves.

Brennan recently assessed a health and nutrition manufacturer’s production estate, ahead of a disaster recovery move to Azure. The virtual machines were running in a third-party private cloud and were headed for a straight lift and shift.

The assessment found that many workloads could be rehosted, but their technical readiness did not remove the need to examine licensing, supportability and cost.

The problem we identified sat in the licensing. Several Windows Server operating systems in scope required closer review of their support and licensing position.

Flexera’s 2026 survey of more than 750 cloud decision makers puts the share of cloud spend going to waste at 29%, the first rise in five years, while Gartner’s May 2026 forecast has Australian organisations spending A$33.6 billion on public cloud this year, up 17.9% on 2025. A growing share of a growing number.

A migration tool moves a workload. It doesn’t repair one. Whatever was wrong with a server on Friday is still wrong on Monday, now running on metered infrastructure with an invoice attached.

So, the decisions that matter happen well before anything moves. Four stages, and each one turns on a decision. What deserves to move (Assess). What needs to exist before anything does (Establish). How much of the old estate travels with it (Modernise). Whether it still earns its cost a year later (Optimise).

We start by measuring. Azure Migrate goes into the environment and collects live performance and inventory data from the servers themselves, over a window long enough to catch real load rather than a quiet afternoon. For the manufacturer, that picture differed from the documentation.

Alongside it, discovery workshops establish what the business is trying to solve: data centre lease ending, perhaps a compliance deadline. In our example, it was reliance on a third-party private cloud with limited resilience and little cost visibility.

Then every workload gets rationalised against the 5 Rs: rehost, refactor, rearchitect…

Opening of the article, excerpted from iTnews. To read the rest, open the article on the publisher's site.