Microsoft Fabric · Data Factory ADF to Fabric Migration
Microsoft Fabric · Data FactoryMicrosoft Fabric Next project
The ADF → Fabric migration experience

Getting an enterprise migration aligned and off the ground.

Azure Data Factory had become the go-to tool for data pipelines across industries for over a decade. Microsoft built Fabric as its newer all-in-one platform, but the challenge was building a migration that didn't break while getting everyone internally aligned. I partnered with my team, brought stakeholders together across two large orgs, and led the end-to-end design to make migration something teams could actually trust.

Role
Principal Designer · Lead
Team
PM, Eng, UX · Azure + Fabric
My focus
Strategy → end-to-end UX
i
Plan ahead with Fabric
Explore what's available and plan your migration.
Migrate to Fabric
Migration assessment
Factory · Sales-Analytics-ADF
142 pipelines · 1,038 activities scanned
Ready{{ readyCount }}
Needs review{{ needsReviewCount }}
Coming soon{{ comingSoonCount }}
Not compatible{{ notCompatibleCount }}
Recreated UI, assessment categories match the shipped experience.
01 · The mandate

Leadership wanted everyone on Fabric. Nobody agreed on how.

Azure Data Factory is mature and trusted by enterprises to run production pipelines at scale. But the data landscape was evolving and ADF was falling behind. Microsoft made the call to build Fabric, an entirely new all-in-one SaaS platform, and made it publicly available in 2023.

Both platforms have run in parallel since. But "move everyone to Fabric" was a mandate, not a plan. There was no agreed approach, no internal confidence, and no fundable project to make it happen.

It's not a simple migration
Users had complex pipelines built on ADF features that Fabric was still rolling out. Many were exploring Fabric for other workloads, but fully committing to the move was a different challenge entirely.
Two orgs, one direction
ADF and Fabric had different roadmaps, customers, and definitions of done. Any path had to work for both, even when not everyone was ready to move.
Breaking things wasn't an option
Real businesses run on these pipelines. Users needed to trust that Fabric would keep building out the features they relied on in ADF before they'd make the switch.
14K+
companies on ADF globally
2 → 1
orgs, one roadmap
0
funded projects at kickoff
02 · What I owned

I was the lead designer who got this off the ground.

UX strategy, end-to-end
Defined the strategy, shaped scope with PMs, and mapped the full migration journey, from factories with 10 pipelines to ones with thousands.
Cross-org alignment
Built trust across Azure, Fabric, PM, and eng leadership by keeping conversations transparent and bringing everyone to a final review where the whole org aligned.
Communication strategy
Helped shape messaging to ADF users without triggering panic about ADF's eventual sunset. Refined phase announcements, from opt-in to a required migration.
Foundational research
Led the end-to-end customer study, from running sessions to synthesis, that defined scope and revealed what it would actually take for users to migrate.
03 · Research

Before designing anything, I went and listened.

I ran a foundational study with 10 enterprise architects and BI leads running data factories in ADF. Three archetypes emerged, and their differences shaped the entire strategy.

Refactor & rebuild
Kay · Enterprise Architect

Inherited tangled no-code pipelines. Wants tooling to extract the hidden logic so her team can rebuild cleanly in code.

"Even if you use no-code, you never really get away from code."
Modernize & consolidate
Milan · Data Architect

Juggling many disparate sources. Sees migration as a chance to simplify but needs a clear picture of what moves and what doesn't before making any decisions.

"I'm the pilot and it is the copilot."
Parallel run, no regression
Ben · BI / Program Lead

Runs production at scale. Not moving until Fabric reaches parity. Refuses to ship a step backward.

"We don't want to go backwards to a worse process."
Archetypes based on research participants.
What the research made undeniable
01
Migration is a refactor, not a copy
The tool should guide a better rebuild, not blindly carry technical debt forward.
02
Assessment before action
Teams need to understand what's ready, what needs work, and what won't move before they commit to anything.
03
Let people move incrementally
Users migrate at their own pace, one pipeline at a time. On their timeline, not ours.
04
Do the heavy lifting for them
Automate the work so users can focus on the decisions that actually matter.
04 · The strategy

Mount first, Assess first.
Migrate when ready.

The insight that unlocked alignment: don't make customers choose between staying and leaving. Bring their Azure Data Factory into Fabric as a live, mounted item, let them assess what's ready, then migrate incrementally on their own terms.

This approach gave both orgs a direction they could finally agree on.

1Mount
ADF, living inside Fabric

Bring the factory into a Fabric workspace as a native item. See and manage pipelines side-by-side. Nothing breaks.

2Assess
Know before you move

An assessment-first report categorizes every pipeline and activity and surfaces the hidden logic underneath.

3Migrate
Incrementally, with help

Select pipelines, auto-map connections, and migrate in waves, keeping anything unsupported running in place.

"

Why it aligned the org: mounting gave Azure customers a low-risk path in, and gave Fabric a clear growth strategy. Both sides could finally say yes to the same plan, because nobody had to lose to make it work.

05 · The experience

Designing the path end to end.

From a single banner inside Azure Data Factory to a guided migration inside Fabric. Here are the moments that mattered most.

a.
The entry point
A persistent banner in ADF. The message changes as the deadline approaches.
Azure Data Factory
i
Plan ahead with Fabric. Data integration features are being delivered in Fabric first. Explore what's available and plan your migration.
Migrate to Fabric
×
2026 · Explore  dismissible, soft nudge
2027 · Transition  persists, encourages
2028 · Critical  deadline, required
b.
The assessment
Every pipeline assessed for migration readiness, then drilling into the logic underneath.
Migration summary
142 pipelines · 1,038 activities
72% ready to migrate as-is
Ready{{ asmReadyCount }} pipelines
Needs review{{ asmNeedsReviewCount }} pipelines
Not compatible{{ asmNotCompatibleCount }} pipelines
Migrate to Fabric
Export report
Pipeline detail, expanded
CopyOnPrem_SalesReady
Transform_Daily_MDFNeeds review
Mapping Data Flow → converts to Dataflow Gen2. Lookup query surfaced for review.
Orchestrate_MasterReady
Legacy_USQL_BatchNot compatible
c.
Mounted ADF in Fabric
The factory, living natively in a Fabric workspace, fully editable.
Fabric · Sales-Analytics workspaceSales-Analytics-ADF (mounted)
Factory resources
▸ Pipelines (142)
CopyOnPrem_Sales
Transform_Daily_MDF
Orchestrate_Master
▸ Datasets (61)
▸ Data flows (12)
▸ Connections (28)
CopyOnPrem_Sales · canvas
Lookup
Get config
ForEach
Per region
Copy data
Load to OneLake
OutputParametersVariablesSettings
d.
The migrate flow
Select → review & auto-map connections → migrate. Three panels, no scripts.
Step 1
Select pipelines
CopyOnPrem_Sales
Refresh_Reports
Legacy_USQL_Batch
Transform_Daily
112 of 142 selected
Step 2
Review connections
SalesDB · Azure SQL
Auto-mapped
BlobArchive · ADLS Gen2
Auto-mapped
OracleFusion · Oracle
Needs mappingMap
26 of 28 connections mapped automatically.
Step 3
Review & migrate
Migration complete
112 pipelines migrated to
the Sales-Analytics workspace.
Open in workspace
Accessible by default
Keyboard focus & screen-reader labels for the whole mounted experience.

Every state of the migration flow was annotated for engineering: focus order, roles, names, and screen reader notifications, so the experience met the accessibility bar enterprise tools require.

Responsive reflow
From full-screen canvas to mobile, without losing the thread.

The slide-in migration panel, explorer, and assessment views were specified across breakpoints so the experience held up wherever data teams work.

06 · Rollout & communication

A migration this big needs time, and the right message.

Phased rollout
Nov 2025
Internal dogfood
Validated the assessment and mounting experience with internal teams.
Jan 2026
Private Preview
Mount ADF into Fabric, assess pipelines, and migrate, guided in the UI.
Next
Auto-map & scale
Automatic connection mapping; large-factory background migration.
Roadmap
Public & GA
Synapse entry points, Fabric panel, path to general availability.
Communication plan · escalating urgency
2026 · ExploreMedium

No deadline communicated. Soft messaging guides users to explore Fabric and migrate when ready, with a dismissible banner.

2027 · TransitionHigh

The tone gets more direct, actively pushing users toward migration. The banner persists and the migrate action is always present in the toolbar.

2028 · CriticalRequired

ADF support is ending and the deadline is clear. Users must move to Fabric. The banner is unmissable and the action is required.

Getting the org to commit meant designing the right messaging, not just the experience. The tone had to reassure users, and the timeline had to give both orgs something they could stand behind.

07 · Outcome & reflection

From a mandate to a plan
the whole org backed.

2→1
Orgs aligned on one path
Azure and Fabric agreed on a
single mount-first migration approach.
3
Archetypes that shaped the strategy
Refactor, modernize, and parallel-run
patterns drawn from the research.
Official
Assessment framework now published by
Microsoft as the standard migration path.
Funded and shipped

The approach was funded and reached internal dogfood and private preview.

The new Microsoft standard

Mount-first, assessment-first is now the migration path Microsoft publishes, with clear states showing what's ready to migrate and what needs a closer look.

Alignment, earned

Consensus across Azure, Fabric, and exec leadership, with patterns that fed into Microsoft's broader migration guidelines.

Reflection

The hardest design problem here wasn't a screen. It was getting a large, divided organization to agree on a direction. Research gave me neutral ground to make that happen. Good design strategy is often just making the right path the obvious one.

{{ fabIcon }} {{ fabLabel }}