Emma Gu

I've shaped how people advertise on and sell on — turning the most ambiguous problems into something that just flows.

Selected Work

01 — Meta

Creatives

Restructured how Marketing messages are created in Ads Manager — introducing multi-text options and AI-powered text generation for advertisers.

Meta Creatives
02 — Meta

Conversations

Created the Conversations experience for Marketing messages in Ads Manager, enabling businesses to manage and automate how they respond to customers at scale.

Meta Conversations
03 — Walmart

Seller Center

Built the first Seller Center homepage for Walmart sellers and redesigned the onboarding and communications experience, driving a +12 NPS lift.

Walmart Seller Center

About

Hi, I'm Emma.

A product designer based in the Bay Area (San Jose).

Most recently I was at Meta, leading the Marketing message experience team — building 0→1 AI-powered creative tools and working closely with PMs, engineers, and data scientists along the way.

Before that, I was at Walmart Marketplace leading the seller acquisition and communications team — designing the Seller Center homepage, onboarding, notifications, and the overall seller experience.

Earlier in my career I did a completely different kind of design — experiential and brand work. Different medium, same job: I've always been a designer.

Emma Gu
01 — Meta · Ads Manager

Creatives

Creatives
+42.67%
CTR — multi-text vs single text
Marketing messages with multiple body text variants performed 42.67% better in CTR than those with a single text option — the core finding that validated the entire direction of this work.

Marketing messages inside Meta Ads Manager

Marketing messages (MM) are a message-based format — distinct from standard ads — that lets businesses send promotional messages directly to customers who have opted in via WhatsApp, Instagram, or Messenger. This project focused on WhatsApp.

I led design across multiple projects combined into three phases — restructuring how MM fits into the creative flow, building multi-text optimization on top of it, and defining a unified future direction. Since the Ads team owns the broader Creative space in Ads Manager, close alignment with them was a constant throughout — not just a one-time handoff.

Role
Lead Designer · Project Owner
Team
PM · Engineering · Data Science · Content Designer
Collaboration
Ads team alignment throughout

A WhatsApp Marketing message is a promotional message businesses send directly to customers who have opted in — delivered to their WhatsApp inbox, not a feed. Unlike standard ads, when a consumer replies, it opens a real conversation.

Marketing messages can be sent through multiple channels. One of them is Meta Ads Manager — the largest advertising platform at Meta — where MM runs as a placement alongside existing ads, letting businesses leverage their ad spend to reach opted-in customers on WhatsApp.

New product launches Promotions & deals Re-engagement
Marketing message example
Media
Tap Target
Headline
Primary Text
Call to Action
Quick Replies

Restructuring the IA — and building on top

MM features were buried in their own isolated sub-section inside Ads Manager — completely separate from where ad creative lived. Research showed that advertisers think about ads and MM as part of the same creative workflow, not as separate tasks. The structure didn't match their mental model, and that mismatch was hurting adoption.

Original creative setup
Ad creative setup — MM buried at the bottom
MM setup modal
Clicking "Edit marketing message" opened a separate modal
Before
Name · Identity · Setup
Destination
Creative
Ad setup
Media
Text
Enhancement
Marketing message setup
Suggested optimizations
Text
Auto reply
Advanced features
Quick replies
Advantage+ · Language · Tracking
After
Name · Identity · Setup
Destination
Creative
Ad + Marketing message setup
Media · Enhancement
Text
Ad text
Marketing message text
Advantage+ creative
Conversations new
Auto reply
Quick replies
Language · Tracking
Text setup

MM was a zero-to-one product — new to Ads Manager and unfamiliar to most advertisers. Giving MM its own separate text field made it legible as something distinct from ad copy. To lower the barrier, the MM field pre-fills from the ad text already written — so it's never blank, but clearly its own thing.

Gen AI

We introduced Gen AI text generation natively within the creative wizard. The MM field surfaces performance-optimized alternatives so advertisers can accept a suggestion or refine the pre-filled text manually — without ever leaving Ads Manager.

The IA restructure brought MM into the advertiser's natural workflow — but adoption data showed that visibility alone wasn't moving the needle. Customization rate stayed at 1.4%, which became the starting point for Phase 2.

Multi-text options — bringing ads-level optimization to MM

Even after the IA restructure, only 1.4% of MM texts were customized by advertisers. Meanwhile, multi-text on standard ads was already proving its value — delivering a +16.33% CTR lift. MM had the same potential, but advertisers weren't leveraging it.

Ad text multi-text options
23%
of ads adopted
multi-text options
1.4%
of MM texts
were customized

Advertisers were already doing the work on the ads side — we just weren't leveraging it for MM. We extended their existing multi-text options to automatically apply to MM, without asking them to do extra work. The key challenge: surface this without overwhelming them visually.

"Due to low adoption, we planned to leverage the multi-text options already added by advertisers for ads — and extend them to MM."

The leadership direction was clear: make it immediately obvious to advertisers that their ads copy is being used for MM by default — and give them a way to customize if needed. The design challenge was surfacing this without adding friction to an already complex workflow.

Option A — Progressive disclosure

Displays options only when needed. By default, ads copy is used for MM. If advertisers want to customize their marketing message, they can select "Provide your own text" to enter their own.

+Keeps the UI simple and focused
+Reduces visual noise and cognitive load
MM customization may be missed unless advertisers expand
Option B — Full visibility

Displays all options at once for immediate comparison. Makes MM customization more visible and easier to discover.

+Greater visibility and prominence for MM
+Advertisers can compare options at a glance
+Fewer clicks to explore
UI may feel visually dense or cluttered
Higher cognitive load when multiple elements compete

We measured MM customization rate — how many advertisers chose to write their own MM text. We ran a 10% rollout followed by a 50% rollout to validate at scale.

Option A — Dropdown
27.8%
customization rate at 50% rollout
Option B — Radio · Winner
32.4%
customization rate at 50% rollout

Full visibility made it easier for advertisers to discover and act on the customization option. Option B was selected as the winning design.

To further support customization adoption, we redesigned the Gen AI experience for MM.

Before
Single replacement
Gen AI generated one suggestion at a time — advertisers had to replace their text entirely to use it.
After
Multi-select
Up to 5 MM-specific variants presented at once — advertisers pick which ones to include, adding multiple options in one step.
MM-specific best practices

We synthesized ~17,000 Marketing messages to identify what drives high CTR in MM — tone, phrasing, and structure that works in direct messaging is different from standard ads. This informed how GenAI generates MM-specific suggestions.

Do
✓ Include an emoji, used selectively
✓ Personalize the message
✓ Keep under 550 characters
Don't
✗ Overload with sale-heavy language
✗ Overuse emojis

Together, these changes — the radio button design and the Gen AI redesign — formed the final experience.

1
Default
Ads copy is used for MM by default — "Optimize with ad text" is pre-selected
2
First option selected
Advertiser adds multiple ad texts — MM automatically inherits all variants for system testing
3
Customize selected
Advertiser writes MM-specific text — Gen AI surfaces up to 5 tailored variants to choose from

Multi-text drove significantly higher CTR for marketing messages

The more text options an advertiser adds, the more variants the system tests — and the better it gets at identifying which message resonates most with each consumer. The best-performing variant is then delivered to the majority of the audience, compounding the impact over time.

+42.67%
CTR lift for MM with multiple body text variants vs. single text

Unification — questioning the structure, not just the features

As the ads team began exploring changes to the creative setup, I partnered with them early to ensure MM wasn't designed around — the two surfaces share the same space, and decisions on one directly impact the other. That collaboration opened up a bigger strategic question: rather than maintaining two separate inputs, what if we unified them?

01
One unified field
Single primary text input serving both ads and MM — one workflow, no duplication
02
Differentiated GenAI
Same source, tailored output — GenAI surfaces distinct suggestions for ads vs. MM based on their separate best practices
03
Simpler interface
Unification reduces screen complexity — less real estate, cleaner creation flow

This is what that unified experience could look like.

Sketch
Text
Add text
+ Add text option
Call to action
Learn more, Shop now…
AI routes to each placement
Advantage+ text generation ✦
Strategy 1
Ad text
Marketing message text
Strategy 2
Strategy 3
Cancel
Next
What's next

This strategy was defined and handed off as the north star for where MM and ads creative should converge next — a single input, with GenAI surfacing tailored suggestions for each surface behind the scenes.

02 — Meta · Ads Manager

Conversations


Turning marketing messages into conversations

Marketing messages don't end at publish. When a consumer receives an MM and replies on WhatsApp, a conversation begins — but that post-publish experience had no dedicated home in Ads Manager. It was buried inside the creative setup flow, designed for pre-publish decisions, not post-publish conversations.

I led the end-to-end design to give conversations their own surface — a standalone Conversations card — and defined the future direction for how AI could make those conversations scalable for advertisers.

Role
Lead Designer
Team
PM · Engineering · Content Designer · Data Science

A WhatsApp Marketing message is a promotional message businesses send directly to customers who have opted in — delivered to their WhatsApp inbox, not a feed. Unlike standard ads, when a consumer replies, it opens a real conversation.

Marketing messages can be sent through multiple channels. One of them is Meta Ads Manager — the largest advertising platform at Meta — where MM runs as a placement alongside existing ads, letting businesses leverage their ad spend to reach opted-in customers on WhatsApp.

New product launches Promotions & deals Re-engagement
Marketing message example
Media
Tap Target
Headline
Primary Text
Call to Action
Quick Replies

Conversations drive MM performance — but they're not happening

Marketing messages are one-directional today. 60% of consumers read them — but only ~1% end up in a real conversation.

FINDING 01
60% read rate
Consumers see Marketing messages. The drop-off happens after — there's no conversational partner on the other side to continue the exchange.
FINDING 02
3.26% → 19.31%
Only 3.26% of consumers reply to MMs. Of those, just 19.31% become deep conversations — defined as 2+ messages from each side. The rest drop off with no follow-through.
FINDING 03
+11% CTR
MMs with deep conversations show 11% higher CTR compared to those without. When conversations happen, performance follows.

Of our MM advertisers, 61% manage replies directly — without any third-party tool. Of those, 89% have no conversation tool at all. For them, Chat Builder wouldn't be an upgrade. It would be their first ever reply management solution. Consumers were already replying to their MMs — and getting no response.

"The data showed a clear opportunity: if we could make conversations happen more reliably, CTR would follow. The product just didn't have the right surface to support that yet."

The automation setup was in the wrong place entirely

Auto Reply and Quick Replies were collapsed into Creative — a pre-publish flow with no connection to post-publish conversations. Two different moments in the advertiser journey were sharing the same card.

"Most advertisers never found the automation setup at all."
Pre-publish · Creative

This is where advertisers configure how the marketing message looks before it goes live — media, copy, and creative decisions. It's a setup flow that happens once, before the campaign launches.

Post-publish · Conversations

This is where automation lives — Auto Reply and Quick Replies are only relevant after a consumer sends a message in response to a marketing message. It's a response layer that operates independently of how the marketing message was built.

1
Working session
With Business Messaging team's designer to formalize the pre vs. post framework
2
Leadership presentation
Brought the framework to both Ads and BM leadership
3
Cross-team alignment
Both sides approved the restructure — argument was structural, not subjective
Before
Name · Identity · Setup
Destination
Creative
Ad setup
Media
Text
Enhancement
Marketing message setup
Suggested optimizations
Text
Auto reply
Advanced features
Quick replies
Advantage+ · Language · Tracking
Recommended
Name · Identity · Setup
Destination
Creative
Ad setup
Media
Text
Enhancement
Conversations new
Auto reply
Quick replies
Advantage+ · Language · Tracking

Chat Builder — a new home for conversations

Auto Reply and Quick Replies moved out of Creative into their own dedicated card — Chat Builder — a new Conversations section at the same level as Creative in the L1 flow.

For the design pattern, we didn't start from scratch. Click to WhatsApp (CTWA) already had a Chat Builder — and most MM advertisers are also CTWA users. We built on the same pattern to give advertisers one consistent surface for conversation setup, no matter which product they're using.

"We didn't introduce something new. The mental model transferred directly."
Chat Builder — where it lives in the L1 flow
Template Management

Choose from previously saved configurations to apply a full automation setup in one click. Every time you publish, Chat Builder auto-saves your setup as a template — so returning advertisers start with their preferred configuration already filled in.

Automation Preview

See what's configured before you publish — both automations shown inline so advertisers can review at a glance.

Quick Replies

Quick reply buttons delivered with the MM — when a consumer taps one, that response is sent instantly.

Auto Reply

Triggered automatically when a consumer sends any message. Pre-filled by default so advertisers can publish immediately without any setup.

Before

No live preview. Advertisers published without seeing the consumer experience.

After
Interactive live preview

Right panel updates in real time as advertisers configure.

Consumer POV

Preview shows exactly what the consumer will see and receive.

Improved grouping

Replaced nested fields with flat layout — button and response relationship is clearer.

Placeholder association

Updated placeholder text to better match the quick reply configuration component.

"Advertiser feedback confirmed that setup was significantly easier — many could configure their automations immediately, without any guidance."

Automating the conversation gap at scale

Business AI closes that gap by turning every marketing message into an intelligent two-way conversation. When a consumer replies, AI responds instantly — keeping the exchange within the 24-hour free messaging window so advertisers incur no additional cost to reopen conversations.

Existing MM advertisers complete a one-time BizAI onboarding. Once enabled, the feature surfaces directly inside the Conversations card and Chat Builder at the L1 level — no new surface to learn.

The opportunity
Every unanswered reply is a missed conversion. Business AI turns that gap into a scalable conversation engine.
24-hour free window
When a consumer replies to an MM, a free 24-hour messaging window opens. Business AI responds instantly within that window — so advertisers pay nothing extra to continue the conversation.
One-time onboarding
Existing MM advertisers complete BizAI setup once — then it surfaces at L1 inside Conversations card and Chat Builder.
Strategy exploration · Not launched

I also redesigned the quick reply placement — moving the buttons below the CTA to create a visual separation between the marketing message and the conversation entry point. The hypothesis: advertisers who clearly distinguish the two will see higher conversation initiation rates than those using the current placement.

"Without automation, every consumer reply goes nowhere. Business AI closes that loop."
03 — Walmart Seller Center

Seller Center

↑ 1200 bps
Landing Page NPS
Improvement in NPS score after launch
1600
Homepage Card CTR
Average clicks per day across homepage cards
↑ 82%
Promotion Adoption
Increase in WMP Capital Program adoption

The platform powering Walmart's third-party sellers

Walmart Seller Center had no dedicated homepage — sellers landed directly on the items page with no sense of context, priority, or what to do next. I led the end-to-end design of the first-ever homepage from scratch, from competitive analysis through to launch and design QA.

Role
Lead Designer · Project Owner
Team
PM · Engineering · Content Designer

There was no homepage — sellers had nowhere to start

Without a dedicated landing page, sellers went straight to the items page. Unrelated notifications appeared on this page to alert sellers, with no sense of priority or context. There was no homepage that could help sellers understand their business, complete tasks, or grow their store.

Before
"The website is lacking in general functionality related to the tasks needed each day."
"A lot of tasks are too manual to remember."
"We would like to have more complete information in one place."
"I hope there is a dashboard to tell new orders, return and refund requests — just an overview of my account."

Every major competitor was ahead

I audited Amazon, eBay, Etsy, and Shopify to understand what sellers experience elsewhere. All four had dedicated homepages with task guidance and growth features. Walmart had none of these — and was the only platform without a dismissible notification system or a promotions discovery surface.

Platform Guided tasks Contextual metrics Growth tips Lifecycle-aware

Defining what belongs on the homepage

I ran a content workshop with the product team to determine the essential requirements across three core areas: key business metrics, tasks, and growth. From there, we mapped each piece of content to the seller's lifecycle stage — whether they're onboarding for the first time or already an active seller — so the homepage surfaces the right information at the right moment.


Three pillars. One homepage.

A clear mission emerged from the research: the homepage should help sellers run their business, complete their tasks, and grow their store — in that order of urgency.

01
Run the Business
Surface the metrics that matter — revenue, orders, conversion — with context for trend, not just the number.
02
Complete Tasks
Prioritize pending actions by business impact. Turn the homepage into a daily triage layer.
03
Grow the Store
Offer contextual growth tips — promotions, listing health, programs — relevant to where sellers are in their lifecycle.

Mapping needs by lifecycle stage

A week-one seller and a three-year veteran need completely different things. I mapped the full journey to design an experience that adapts to where sellers are.


Exploring the layout

I explored several directions before converging on the final approach — testing how to balance metrics density with task clarity.

Initial homepage with three core sections

After several iterations, we developed the initial version of the homepage — a single layout organizing everything sellers need into three focused sections: To-dos, KPIs, and Promotions.

Testing comprehension — new design vs. original onboarding

We tested this first-round design using the PURE (Practical Usability Rating by Experts) method with 5 evaluators. To set a baseline, we first tested the original onboarding flow, then integrated that onboarding into the new homepage to see how the To-do section improved seller comprehension.

Original onboarding flow

The original onboarding prioritized the "Add Items" banner over the actual setup steps — sellers were pushed toward listing products before completing essential account setup, causing confusion and drop-off.

New onboarding flow

By integrating onboarding into the To-do section, sellers could clearly see what setup steps remained — in the right order, with the right priority — without being distracted by promotions.

New design outperformed across every task

5 Evaluators
  • 2 Designers
  • 2 Products
  • 1 Researcher
User Mindset
  • Prospective Walmart Marketplace seller
  • Already sells on Amazon and eBay
Original onboarding 17
Become a Verified Seller 4
Add Payments 5
Review Status 3
Add Shipping Information 5
New onboarding 11
Become a Verified Seller 4
Add Payments 3
Review Status 2
Add Shipping Information 2
PURE score — lower is better. Teal = easy, yellow = moderate, red = difficult.

A homepage that knows who you are

The redesigned homepage brings together three focused sections in one scannable layout — adapting to each seller's stage and surfacing the right information at the right time.

TO-DOS
Action-first
Sellers can clearly see what needs to be done — ordered by urgency and business impact, so nothing falls through the cracks.
BUSINESS KPIs
Performance at a glance
Key metrics like revenue, orders, and conversion shown in context — with trend signals so sellers understand their business without digging.
PROMOTIONS
Contextual growth
Relevant programs and promotions surfaced based on where the seller is in their lifecycle — not one-size-fits-all banners.

A framework for what shows up and when

In collaboration with the Content Designer, we developed a framework to determine what content belongs on the homepage and where — including where seller alerts should be routed, how each module is prioritized, and when content appears or retires based on the seller's lifecycle stage.

Banner lifecycle

Promotional banners were a major pain point — they'd pile up, go stale, and drown out more important tasks. I designed a lifecycle system that controls how banners appear, surface, and expire.

What sellers said

"I LOVE the homepage you've created."
"Love the new homepage, I can even see new orders everyday!"
"It's really easy to work with."
"This is new and interesting. It's a good landing page."
"If we can have a checklist of the actions needed to keep the site at top shape, that will help."

Measurable impact across the board

↑ 1200 bps
Landing Page NPS
Improvement in NPS score
1600 /day
Homepage Card CTR
Average clicks per day across homepage cards
↑ 82%
Promo Adoption Rate
Increase in WMP Capital Program adoption