Azure Quickstart Center

Timeline

2022 – 2024

Focus

Growth Design

Impact

Phase 1

+33% Activation

+41% Trial → paid

+30% Retention

Phase 2

+12% Activation

+15% Trial → paid

Validated across 90-day A/B tests vs. the original.

My Role

Lead Designer — owned end-to-end design across both phases and the follow-on project.

Collaborated with PM, 2 researchers, 2 engineers, and partner teams (MS Learn · Azure Portal · Service Extensions).

Context

Azure Quickstart Center is the designated front door for new Azure users — a place to learn the platform and deploy their first service. But three signals told us it was failing at that job: users took months (sometimes years) to deploy their first service, ~600K visitors came each quarter (80% first-timers) and most left without deploying anything, and a meaningful share of users didn't even know AQC existed.

Original Azure Quickstart Center — a directory of links

The original Quickstart Center

🔍

Discoverability wasn't the bottleneck

Before committing to a redesign, we tested the cheapest hypothesis first: maybe users just couldn't find Quickstart Center. We redirected new users to AQC immediately after first login. The result was a null — no statistically significant lift. That null result became the case for redesigning the experience itself.

Challenge

That null result reframed the question — this wasn't about discoverability. It was about rebuilding the experience itself.

🎯

Problem Statement

How might we redesign Azure Quickstart Center to help first-time users learn Azure fundamentals more efficiently and deploy their first services faster?

Strategy

Research surfaced four insights. Three could ship additively, without touching the existing IA. The fourth required restructuring it. That split shaped a two-phase approach — sequenced by risk.

🧭

Two phases, not one rebuild

Phase 1 ships the three additive insights — checklist, multimedia, in-context deployment — on top of the existing IA. Phase 2 restructures the IA itself, but only after Phase 1's results have earned the case for it. The bet: split the work by risk — get value out faster with the safe additive bets, then use those results to justify the bigger structural ask.

Phase 1 · MVP

Layer in a guided start

A guided starting point that lets new users learn a concept and deploy a working service in the same flow — without leaving the page. Phase 1 was deliberately additive: a new first tab in AQC, no changes to the existing IA.

Problems we tackled

The original Quickstart Center was a directory of links — no scaffolding, no in-context learning, no momentum.

📚

No scaffolded learning path

New users didn't know where to start. There was no first step, no order, no sense of progress.

📝

Text-heavy documentation

Abstract cloud concepts were explained in dense docs. Users wanted videos and visuals to build mental models faster.

🚪

Every link took users out

Each card sent users to a different site. They lost their place, broke focus, and rarely came back to finish.

Design Highlights

Step-by-step Starter Checklist

Replaced the wall of links with a scaffolded checklist — expanded by default so users can see the full path before they begin. A persistent progress bar builds momentum, and a completion notification rewards the finish before pointing to what comes next.

Starter checklist with expanded steps and progress bar

Video tutorials over text

Testing showed users built mental models faster from video than from text or graphs — particularly for abstract cloud concepts. Tutorials open in a dedicated focused view (testing also showed side panes were distracting).

Video tutorial in a focused page within Quickstart Center

In-context service deployment

Service deployment slides in as a context pane on the checklist page instead of redirecting users out of AQC. Staying in one place reduced confusion and was the most strongly-preferred concept in user testing.

Phase 1 impact · 90-day A/B test

+33%

Activation rate increase

+41%

Free trial → paid conversion

+30%

Returning user rate

Phase 2 · IA Redesign

Restructure around the journey

Phase 1's checklist worked — but it was layered on top of an underlying IA that was never designed for the catalog Azure had grown into. Phase 2 restructured the IA around how users actually progress through the platform.

Problems we tackled

📊

The tab structure couldn't scale

The original tabs were built for a smaller catalog. As Azure's services and templated solutions multiplied, the structure got harder to navigate, not easier.

🧭

No clear journey through the content

Users couldn't tell what to do first as content grew — "lots of resources but nothing told me what was most important."

Design Highlights

Three-stage information architecture

Reorganized the landing page into Get Started (starter checklist + in-context deployment), Create (service deployment + templated solutions), and Learn Further (documentation + training). The progression maps directly to user intent at each stage.

Final Quickstart Center landing page with three-stage IA

Persistent table-of-contents navigation

Added a left sidebar that organizes content by Azure service category, so users can scan available topics and jump anywhere in the flow without losing their place.

Phase 2 impact · 90-day A/B test

+12%

Activation rate increase

+15%

Free trial → paid conversion

Follow-on

From clearer paths to clearer choices

Phase 2 cleared the path to Create, but users still got stuck at the service card itself — two friction points that called for two different patterns.

Problems we tackled

❓

"What should I deploy?"

Even with a curated catalog, new users couldn't tell which service matched their goal. They defaulted to inaction or guesswork.

🤔

"What did I just pick?"

Service cards led straight into deployment wizards. Users were configuring resources they didn't fully understand.

Design Highlights

Copilot — translating goals into services

For "What should I deploy?" — an AI assistant on the deployment page where users describe what they're building in plain language. Copilot returns the most relevant service with rationale. AI's role is translation, not authority — it surfaces options, the user still decides.

Copilot recommending services on the Azure deployment page based on user goals

Service explainer — context before commitment

For "What did I just pick?" — a plain-language layer with text and images that sits between selection and the deployment wizard. Users see what the service is, when to use it, and what they're committing to before configuring a single field.

Service explainer page — plain-language overview between service selection and the deployment wizard

Takeaways

🔍

Research clarifies ambiguity

"AQC isn't working" could mean many things — research turned it into four specific, actionable design directions.

🧭

Design for the user's mental model

The three-stage IA came from how users want to progress, not how Azure organizes its services.

🎯

Strategy is what to leave alone

Phase 1 deliberately didn't touch the IA. That constraint kept scope tight and earned trust for the bigger Phase 2 conversation.