sr. product designer • The Knot • iOS & web • 2026
couple onboarding | targeting +16% completion
Redesigning sign-up and onboarding so fewer couples drop off and the product learns enough to help them from the first session.
Pre-redesign sign-up drop-off
New components + colour scale for the design system
Onboarding completion — Web
Onboarding completion — iOS
tl;dr
About 70% of people dropped off at sign-up. Couples, guests and vendors all went through the same flow, and the homepage ignored the answers people had just given. I was the only designer on the redesign, working with a PM, a UX researcher, a UX writer, four web engineers and one iOS engineer. We added a first screen that sends guests and vendors where they actually need to go, moved account creation after the questions that show couples what they’ll get, and changed how we ask for wedding date, location and vendor needs so the answers are usable later. Two rounds of unmoderated testing shaped the flow before development. It’s now in A/B testing, aiming for an 11–16% relative lift in onboarding completion on web and iOS without lowering core planning actions or account quality.
overview
The Knot is one of the leading wedding planning platforms in the US, where couples find vendors, manage their guest list and keep track of everything else.
Onboarding is the product’s front door, and too many people were leaving before The Knot had a chance to show them anything.
The goal
Lose fewer people at sign-up, ask for each piece of information when it makes sense to ask, and use the answers to point couples to their first core planning actions (CPAs): booking a vendor, contacting one, or making a purchase.
problem
I started with the drop-off data, past experiment results and existing research. There were years of studies to get through, so I used AURA, The Knot’s internal AI research tool, to pull out the patterns that kept coming up across them.
- Couples, guests and vendors all landed in the same flow, so it didn’t fit any of them well. About 17% of people going through onboarding weren’t couples at all.
- The sign-up form appeared before people had seen anything, so we were asking for an account before giving them a reason to create one.
- Guests and vendors had no path of their own: they either signed up as a couple or left.
- Couples who add a wedding date are 59% more likely to go on to complete a core planning action. But the date question was optional and near the end, and more than half of users skipped it. Without a date, checklists, countdowns and email retargeting can’t be personalised.
- People were happy to share information, but expected to see it used. When the product ignored it, they trusted it less.
sign up modal displayed right away, no way out for guests or vendors
confusing categories and allowing contradictory options to be selected at once
not enough flexibility for users that were not sure about their wedding date yet
not all vendor categories were displayed and data collection was not granular enough
discovery
I was the only designer, working with a PM, a UX researcher, a UX writer, four web engineers and one iOS engineer, with input from design and product directors. I looked at how other consumer products handle onboarding and ran internal critique sessions to test ideas with the team before testing them with users.
How I prototyped
I built early concepts as clickable prototypes in Claude Design, Claude Code and Figma Make, then took the strongest ones into Figma for detailed design and handoff. Prototyping got much faster, so I could try more directions for each problem and get them in front of the team for critique before spending time on high fidelity. I now work this way on my other projects too.
What we were betting on
The old flow felt like a marketing form: fill this in so we can email you. We wanted it to feel like: tell us what you’re planning, and we’ll set up your workspace around it. If that worked, fewer couples would drop off, their profiles would be more complete, and more of them would take a planning action in their first session.
Onboarding doesn’t create planning actions by itself. Its job is to understand the couple and pass that on to the homepage and the Next Best Action (NBA) engine, which decide what to show them next.
So we weren’t trying to get more sign-ups, we were trying to get better ones. Sending guests and vendors elsewhere would lower sign-up numbers, and we were fine with that.
design decisions
Most of the redesign came down to order: find out who someone is, show couples what they’ll get, then ask for an account.
-
Intent routing as the first screen
About 17% of people in onboarding weren’t couples, and none of them needed an account. We considered a lighter sign-up form and a skip link, but both still treat guests and vendors as people who should sign up. So the first screen now asks “Which one best describes you?” Guests go to Find a Couple, vendors go to WeddingPro, and couples continue into their own flow. In testing, the “Just browsing” option made people more comfortable carrying on without committing.
-
Earning the sign-up
Account creation now comes after couples have added their names and wedding date. By then they’ve seen what The Knot will do with their answers, so asking for an account makes sense. There was a technical reason too: the wedding date decides which email sequence a new user gets, so it had to exist before the account was created.
-
The right data, at the right granularity
The date question now accepts a month and year, or several, for couples who haven’t picked a date yet. The backend used to store these rough dates as exact ones, which made them misleading downstream. Venue and location became separate questions, and location now works outside the US, which matters because 12% of couples are planning a destination wedding. Guest count became a range. For vendors, I first explored asking, for each category, whether the couple had already booked it and how interested they were (interested, maybe, not interested). It was too much to answer during onboarding, so I cut it down to a simple booking question (no / not sure / yes) and kept that for the A/B test.
-
Connecting to the homepage
Everything couples answer is saved to their wedding profile, which the rest of the product can use to personalise what it shows. The homepage hasn’t been redesigned yet, but it can now use this data to show relevant planning actions from the first visit.
Guests and vendors don’t need an account, so they get a way out. We expect fewer sign-ups, but more of them will be real couples.
guest opt-out, redirects to the find a couple site
vendor opt-out, redirects to the sub brand focused on wedding professionals
smoother transitions and more aligned with the new brand refresh
allowing multi-select across months and years to reduce step drop-off rates
asking what couples need help with, which also shows them what the product can do
displaying all vendor categories and offering more granularity on need level
validation
I ran two rounds of unmoderated testing, with 10 participants per round (5 on iOS, 5 on web). I wrote the script and built the prototypes using our research team’s template, and launched each round once the UX researcher had approved it. We analysed the results together as a team. I used NotebookLM to go through the recordings and notes faster and compare findings across iOS and web.
-
User type
Everyone got through this step easily. “Just browsing” made people more comfortable carrying on, but “Other” confused them. → We removed “Other”.
-
Couple names
Some people felt asking for full names this early was too much. → We made last names optional and only ask for first names upfront.
-
Email auth
Most people treated it as a normal step, though one participant asked why we needed an account at that point. → We moved sign-up after the wedding date question. One comment wouldn’t have been enough on its own, but it lined up with the email-sequence dependency above.
-
Wedding date
Most people completed it, but a few didn’t notice they could select more than one month. → We’re watching how people use it after launch before changing the UI or copy.
-
Wedding location & guest count
No issues. People understood the broad location search and the guest-count input. → This confirmed the changes we’d made after the first round.
-
Planning & vendor needs
People understood both steps, but one participant felt the vendor question was “too much too soon”. It may work better asked later, after onboarding. → We’ll decide based on post-launch data.
-
Attribution & offers
No issues. People saw both as normal onboarding steps and read the offers as sign-up perks. → We’re tracking whether the new design moves revenue in either direction.
-
Home screen & transitions
Transitions worked and the home screen met expectations; people felt the top section was personalised to them. → Even so, the homepage needed more work than this project could hold, so it became a project of its own.
challenges & trade-offs
-
Rescoping mid-project
The project started as onboarding plus a homepage redesign. Partway through, it was clearly too much for one project, so we split it: onboarding first (this case study), the homepage later. That changed how we measured success. Core planning actions mostly happen on the homepage, so judging onboarding on them alone wouldn’t be fair. We made onboarding completion the main metric, and tracked two side effects we expected: fewer sign-ups, because guests and vendors now leave the flow, and fewer dormant accounts.
-
A trade-off on the flexible wedding date
I designed the flexible date to cover several cases: consecutive months in the same year, consecutive months across years, the same month in different years, and more. Early on, our engineers and another squad’s engineers agreed the existing backend could support it. It later turned out it couldn’t, and we had to cut down to two cases: consecutive months (within or across years) and the same month across years. It’s less than planned, but still more flexible than the single fixed date we had before. Next time, I’d ask for any change to core logic to be checked in the code early on, instead of relying on an agreement.
impact
The project also added two components and a new colour scale to the design system. They were built for onboarding, but are meant to be used across the product.
-
Navigation tile
Links to another page or section. Used in lists, it helps users quickly scan and jump to related content, tools, or sections.
-
Selection tile
Lets users quickly compare and choose one or more options from a set of choices, in forms and guided flows like onboarding.
-
Warm gray scale
A new warm gray colour scale for the brand refresh, replacing the outdated champagne tone across the product.
status & next steps
Currently in A/B testing.
How we will know we succeeded
Onboarding completion decides whether the test succeeds. Planning actions and account quality are guardrails: they shouldn’t drop. The rest is tracked for context.
| Metric | Current | Target | Relative lift | Type |
|---|---|---|---|---|
| Onboarding completion — Web | 54% | 60% | +11% | Decision |
| Onboarding completion — iOS | 63% | 73% | +16% | Decision |
| Mean uncapped CPA (14d) | Baseline | Neutral or ↑ | Must not drop | OKR — Decision |
| 30-day new couple activation rate | Baseline | Neutral or ↑ | Must not drop | Informative |
| Sign-up volume | Baseline | Monitor | Expect down | Informative |
| Guest metrics (site visits, registry purchases, RSVPs) | Baseline | Neutral or ↑ | Expect up | Informative |
| Vendor metrics (WeddingPro traffic, leads) | Baseline | Neutral or ↑ | Expect up | Informative |
Open questions beyond the test: how well the homepage uses onboarding data to drive planning actions, and whether the first screen should change depending on where people come from, since someone arriving from a vendor page already tells us something different from someone arriving from a planning tool.
see other projects
travel search | +8 pts NPS
See project →subscription management | −27% tickets
See project →wedding vision | AI prototype
Case study coming soon