
Solution and impact
- A short path for people who already know their ticket, the price on the card opens checkout, and a long path with details and Compare for everyone else.
- A ticket card that explains itself, plain names, what is included in 3 lines and Compare in every category.
- Every ticket on one page from the first screen, with filters that refine instead of hide.
- One language and currency choice that holds from the first page to payment.
- A new visual direction, chosen in 4 review rounds, carried to every page.
Constraints that shaped the work
- A third-party ticketing engine, Ventrata. Checkout changes went to another vendor as proposals, each with its own business case.
- Speed was already good, LCP 1.39 s and CLS 0.01. The redesign could not cost performance.
- Many languages and currencies. 44.8% of visitors came from abroad, and the language had to be chosen twice.
- A release date set at kick-off, with a smaller first launch.
- WCAG built into the concepts from the start.
Discovery research that changed the framing
- Visitors arrive ready to buy. Buy Tickets was the most clicked element on Home, on every device.
- Most of the offer was out of sight. On a phone, 27% of visitors reached 5% of the Buy Tickets page and 11% reached 20%.
- Ticket tiers confused people. "Not sure which ticket to buy" was the second reason for buying elsewhere, and the 86th + 102nd floor page bounced 48.6% against 19.3% for the 86th floor.
Problem: a last-minute decision, made on a phone.
Decision: a mobile-first Buy Tickets page with 2 paths through it. Prices sit on every button, and the photos are a swipeable carousel.
KPI: checkout funnel entry rate, the share of all visitors who reach the ticketing flow. Target set before launch: +20% against the pre-launch baseline, measured in GA4.
The short path
For visitors who already know their ticket, the price on the card opens checkout on the same page.
The long path
For visitors still deciding, from Home to Buy Tickets, then everything a ticket includes before booking it.
Problem: tickets people couldn't tell apart.
Decision: a card that explains the ticket on its own, and Compare in every category. Compare ends every category and opens a table of what each ticket includes, one ticket at a time on a phone.
KPI: premium ticket conversion, the share of premium tickets in all sales. Baseline 55.7% from the Empire State Building revenue team, target +5%.

- A photo of the real view, 3 per card, swipeable.
- The full name. Empire State Building instead of ESB, which participants didn't recognize.
- The starting price on the button, which participants used to shop within a budget.
- What is included, in 3 lines.
- More details, the full list with Compare and a book button.
What didn't ship: Day and Night.
Plain ticket names across the whole range, and Day and Night, a single photo that turns from day to night instead of 2 stacked pictures. Both were agreed with the Empire State Building, then neither was tested nor built: the content management system couldn't hold them.


Problem: offers nobody saw.
Decision: every ticket on one page from the first screen, filters that refine instead of hide. We first built filtered categories. In Round 1, 7 of 10 people didn't understand the filter names and 5 of 10 expected to see everything first, so all tickets became the default and the filters a shortcut. In Round 2, all 10 found every ticket type.
KPI: ticket type visibility rate, how many visitors see or tap a ticket beyond the top 3. Baseline 3% on mobile and 5% on desktop in Clarity, target 20–25%, 1 visitor in 4.
Problem: international visitors fighting the language and the prices.
Decision: one language and currency choice that holds from the first page to payment. The language had to be chosen twice and prices stayed in dollars until checkout, yet non-English visitors converted better, Chinese at 4.67% and Japanese at 3.21% against 2.51% for English. The proposal: one menu, the language detected and remembered, local prices and date formats.
KPI: localization effectiveness, checkout completion of international visitors. The baseline is set after launch.
Problem: a look that didn't make the official site feel like the place to buy.
Decision: a new visual direction, chosen in 4 review rounds. Concepts came from several designers. Stakeholders judged them by user impact, not taste, and kept mine every round; Round 2 testers called it modern and on-brand, and it is the identity on the live site today.










The chosen style, carried to every page
The first UI round, the moment the concept became templates. Details changed in later rounds.
A component library and a section library, so new pages are assembled, not designed.
- 226 tokens, 117 component sets, 1,213 variants, each with Desktop, Tablet, Mobile and Wide Desktop modes.
- Sections on top of the atoms, so a first-time event page still looks finished, and templates readable by people and AI search.
Design
Figma
Claude
Data
Google Analytics 4
Microsoft Clarity
Ventrata
Standards
WCAG AA
Not a line. A loop with 2 test rounds and 4 feedback rounds in it.
| Stage | When | Deliverables | What sent us back and trade-offs |
|---|---|---|---|
| Kick-off and data | Oct 2025 | GA4, BrightEdge, competitor and Baymard research. | — |
| CRO audit | Oct–Nov | Funnel, heatmaps, Core Web Vitals. | 92% drop, so checkout became the core of the scope. |
| IA | Oct 27 | A reimagined structure: Experiences, Plan your visit, About. | Gave up the safety of a structure that already converted, because visitors arrive in planning mode and the museum and exhibits were invisible. |
| Quick wins for Ventrata | Nov 12 | Checkout recommendations I defined from the audit. | — |
| KPI framework | Nov → Jan | 3 versions. | Baselines updated as data arrived. |
| UX-1 → UX-4 | Nov → Jan | All templates, checkout v2. | Reviews each round. |
| Visual concepts 1–4 | Jan → Feb 18 | 4 rounds, my concept kept each time. | — |
| First-launch scope | Feb 24 | A smaller first release, search limited to tickets. | Phased the release: the first launch left out the AI ticket assistant, group sales, planning sections and site-wide search; the second launch shipped the full scope. |
| Usability testing, Round 1 | Feb 26 | 10 sessions, mobile wireframes. | Filters renamed, show all, add-ons before payment. |
| Usability testing, Round 2 | Mar 12 | 10 sessions, mobile UI. | Deferred comparison across categories, the most requested feature, 7/10 then 10/10, to the second launch, because the ticketing platform couldn't support it for the first one; compare inside a category at launch. Full name, real views. |
| UI-1 → UI-3 | Mar → May | 4 feedback rounds. | Reused system components over bespoke special-experience sections, saving 3 days of build. Price order, included in every ticket, $5 fee. |
What I'd do again, and what I'd do earlier
- Put the vendor in the room on day one. The most requested feature in testing, comparison across categories, waited for the second launch.
- Treat baselines as drafts. Funnel entry moved from 2.4% to 6.5% between versions; targets set on the first number would have been meaningless.
- Test copy with the layout. Most hesitation in both rounds came from words, "ESB", "Green Room", "From $X", not structure.




































