The migration that goes badly almost never goes badly on migration weekend.

It goes badly six weeks earlier, on an unremarkable Tuesday, when nobody looks closely at the instance and a few thousand work items quietly carry a data problem that no one has thought to check for yet. The weekend is simply when you find out. By then you have a change window, a rollback plan, an executive on a bridge call, and roughly none of the time you would need to fix the actual cause.

I mention this because Atlassian published its Q1/Q2 2026 Cloud Transition Round-Up in late June, and if you skim it you will read a list of tooling updates. If you read it properly, you will notice that nearly every item addresses the same moment — the weeks before anyone touches a migration plan. That is a more interesting editorial choice than the release notes let on, and it happens to match where migrations actually succeed or fail.

Portfolio Insights got substantially more opinionated

Portfolio Insights is the free assessment feature in Atlassian Administration that inventories your product portfolio, evaluates your data, and hands back recommendations before you commit to a plan. Four things changed, and they compound.

Connecting is now one install. Wiring a Data Center instance to Portfolio Insights previously meant collecting several Marketplace connector apps. Those capabilities are now bundled into a single installation. Unglamorous, and precisely the sort of friction that quietly determines whether an assessment happens at all.

The assessment logic updates itself. The data and metrics app is what lets Portfolio Insights run Atlassian-supplied assessment queries against your instance, so your results reflect current logic rather than whatever shipped the year you installed it. It rides along with Jira Cloud Migration Assistant 1.12.50 and later, and Confluence Cloud Migration Assistant 3.13.9 and later. Update your migration assistant and you have it; there is nothing else to install.

Apps are finally in scope. Marketplace app assessments and custom app identification are now part of the cloud readiness report. Anyone who has run an enterprise migration knows apps are where the schedule goes to die — not because they are hard, but because their complexity surfaces late, after the plan is already socialized. Moving that discovery earlier is worth more than it sounds.

Performance and Security insights arrived. Performance insights use the Apdex framework to measure how satisfying your key workflows actually feel to end users, with instance-specific recommendations attached to detected symptoms. Security insights score your Data Center configuration against best practices across five areas — vulnerability management, system configuration, application configuration, access management, and apps and integrations — with step-by-step remediation for anything that fails.

Here is the detail I did not expect: these two are not really migration features. A security score that refreshes roughly every two hours, with daily snapshots over the trailing week, is instance health tooling. It is useful on a Tuesday when you have no migration planned at all. And cloud readiness assessments now rerun automatically every 28 days — opt-out available in the UI — so your picture stays current as your data shape drifts.

From prep to cutover

Three more items, moving down the timeline.

Preflight remediation is the one I would flag to anyone mid-planning. Run a pre-migration check and you now get actionable specifics — the affected project, the item type, the recommended fix — and you resolve them using a structured remediation file supporting both required and optional updates across many items at once. Upload the corrected file, get to green, proceed. The critical part is that none of it changes your source environment. You are fixing the migration’s view of the data rather than editing production to satisfy a tool. It has been available to all customers since April 2026. A guided remediation UI inside JCMA was planned for the end of Q2 2026, so check whether it has landed in your instance before assuming the CSV route is your only option.

Migration timestamps in JCMA are an opt-in checkbox on the Select Projects step: add migration timestamp details to entity descriptions. Enable it and JCMA appends a short tag during export — something along the lines of migrated on a given date from Jira Data Center — to migrated projects and issues. Left off by default, and off means genuinely untouched. Small feature, real value for auditability and for the six-month-later conversation about which content came from where.

The cloud-hosted migration assistant is the structural change. Rather than running the migration from your production Data Center environment, you copy your data into a dedicated Atlassian-hosted migration environment and run from there. It supports on-demand incremental transfers, so subsequent runs move only what has changed. That makes test-validate-repeat cycles cheap, which is the single biggest lever on migration downtime. It is currently limited to selected customers in an early access program, so plan around it as a direction rather than a resource.

What is coming, and one thing worth watching

Atlassian flagged four things in flight: usage-based insights showing which apps, automations, and configurations are actually in use so you can clean up before you move rather than after; an Action Required summary tile surfacing only the guardrails currently blocking a test migration; expanded remediation coverage; and a cloud transition agent laying groundwork for AI-assisted troubleshooting and migration support.

That last one is the one I would keep an eye on. An agent that can reason across your instance health, your assessment history, and your remediation state is a very different proposition from a wizard with better error messages — and it is the same underlying bet we described when we wrote about closing the gap between AI and your actual business. Usage-based insights will be optional and opt-in, and Portfolio Insights remains fully functional without it.

The honest caveats

Portfolio Insights is free, but it requires an Atlassian organization — free to create if you do not have one — and access is limited to organization admins in Cloud and system admins in Data Center. If you cannot connect your instance for network or policy reasons, you can still get the insights by collecting and uploading instance health data manually. The cloud-hosted migration assistant is early access. And the automatic 28-day refresh applies to cloud readiness specifically; security insights refresh on their own, much faster cadence.

The Avaratak Take

Every one of these features rewards the same behavior, and it is the behavior migration teams are worst at: assessing early, repeatedly, when nothing is at stake.

The reason we are bad at it is structural rather than lazy. Assessment produces bad news, and bad news before a project is funded feels like a reason not to fund it. So the check gets deferred until the plan exists, by which point the plan has a date, and the date has an audience. Automatic 28-day reassessment quietly solves this by making the news arrive whether or not anyone asked — which is a design decision I respect more the longer I sit with it.

So here is the small piece of work I would do this quarter, in this order, whether or not you have a migration on the roadmap. Connect Portfolio Insights and let the first assessment run — you are gathering a baseline, not committing to anything. Read the security score before you read the cloud readiness report, because that one is useful to you today regardless of where you end up hosting. Then run a pre-migration check on one representative project and work the remediation file end to end, so your team learns the mechanics when a mistake costs nothing. And leave the 28-day reassessment on. The opt-out exists for good reasons, and none of them are ‘we would rather not know.’

Do that and a migration becomes a scheduling problem, which is the kind of problem organizations are genuinely good at solving. Skip it and it becomes a discovery problem, discovered on a Saturday.

The best migration story is one nobody tells afterward, because nothing happened worth retelling. If you would like a second set of eyes on your cloud readiness report, your app inventory, or the sequencing of a move you have been putting off, that is exactly the conversation we enjoy at Avaratak. Come find us at avaratak.com. Bring your assessment; we will bring the good questions.