← Back to all work
05
Shaka Guide Senior Product Designer 2026 · Paid Design Trial Mobile App (iOS)

Improving the Early Discovery Funnel

The drop-off wasn’t one broken screen — it was one screen serving two different intents. New-user discovery already works. The real opportunity was a home that recognises returning intent.

🌞 View Figma File

The Brief

Shaka Guide is a GPS-based audio tour app. Users browse, purchase, and download tours that play automatically as they drive or walk. The brief: review the early app experience (app open → home → discovery → tour detail page) and propose how to get more users to a relevant tour detail page. A meaningful drop-off was visible before users reached that page.

This was a paid 15–20 hour design trial with real funnel data. I was asked to diagnose the problem, identify user intent patterns, and propose a focused UX/UI direction.

🌞 View Figma File

The Funnel Data

The 60-day blended funnel showed a meaningful drop before users reached a tour detail page. But blended numbers hide the story — so the first thing I did was split every step by new vs. returning users.

Funnel Stage All Users New (55%) Returning (45%)
Launch → Home 87.8% 93.0% 81.5%
Home → Product View 72.1% 84.7% 54.5%
Product → Purchase 39.0% 44.6% 27.0%
Launch → Purchase 24.7% 35.2% 12.0%

The Diagnosis

New users: Discovery already works

93% reach home. 84.7% go on to view a product. Their real leak is Product → Purchase — out of scope. There’s limited upside in “fixing” new-user discovery; it’s largely working.

Returning users: A measurement mismatch

Only 54.5% reach a product view — but many aren’t trying to shop. They open the app to resume a tour, access a download, start navigation, or listen to audio they already own.

The composition effect

The blended 72.1% is dragged down by a large returning cohort with different intent. A “product view” was never their goal — so counting them as a funnel drop is partly a measurement mismatch, not a broken screen.

The real problem

The home screen serves one discovery-shaped experience to everyone. It does almost nothing for the returning user who came back with a specific goal in mind. ≈10,000 returning users are lost here.

What I Saw In the App

  • The detail page already sells. Photo/video carousel, audio preview, “Trusted by 7,000+”, 4.7★ from 538 reviews, clear price, obvious Buy Now. The problem is getting people to it — not the page itself.
  • Signing in barely changes anything. The logged-in home is almost identical to guest. The only addition is a “Recently Viewed” row. No “Continue”, no “My Tours”, no history-based picks. The product doesn’t adapt to why a returning user came back.
  • Tour cards give you almost nothing to act on. Photo + title only — no price, rating, duration, or number of stops. Everything needed to judge a tour already exists on the detail page, but cards don’t surface it. Since a Product View = a card tap, weak cards are a direct drag on the metric.
  • Real tours sit deep in the scroll. A “10 Years” banner, Discover tiles, and Destination Spotlights all appear before the first tappable tour. A lot of orientation cost before anything shoppable shows up.
  • Navigation overlaps itself. Category chips, Discover tiles, island tiles, and bottom tabs all route to similar places — many paths, no clear “start here.”

Four Intent Patterns

Grounding the funnel in behaviour, I identified four intent patterns. One of them is structurally different in kind — and the product was ignoring it entirely.

Well served
The Decided

Knows the destination, wants the fastest path to a detail page. Search → detail works for them today.

Enriched cards help
The Orienting

Has a trip but not a tour. Needs ratings, duration, and comparison cues to narrow the shortlist.

Weakly served
The Explorer

No fixed plan yet, deciding if the app is worth it. The current home doesn’t give them a strong “start here.”

Structurally underserved
The Owner / Resumer

Already bought. Opened the app to resume, download, or navigate — not to shop. ≈10,000 showing up as “drops” in the data.

The Direction — An Intent-Aware Home

A
Returning users: Serve the goal first

“Continue / My Tours” at the top of the home — resume the tour they’re mid-way through (with a progress bar and one-tap Resume), owned tours with the right verb per state (Resume · Navigate · Download), and history-based “Keep planning” picks. Discovery stays available, just lower. This routes owners straight to what they came for — converting ≈10,000 “lost” users into people who reached their actual goal.

B
New / guest users: Strengthen discovery with enriched cards

The cheapest high-impact change. Every card surfaces rating, price, duration, number of stops, and a drive/walk badge — all data that already exists one screen deeper on the detail page. The same enriched card powers home, All Tours, and search. Whole-card tap matches the card-tap = Product View definition exactly.

C
Consolidated navigation

Overlapping chips, Discover tiles, and island tiles collapse into one filter bar (Region / State / Nearby + Filters) and one category rail. Tabs re-scoped: Home · Tours · My Tours · Explore — so “browse tours” vs “read content” is no longer blurry. A denied location permission no longer dead-ends; the home leads with the last-planned destination instead.

Before & After

Two problem areas, two targeted fixes. The home screen ignores returning intent; the discovery cards hide the information needed to act.

Home screen — returning user

Current home — generic discovery, no returning-user state Before

Generic discovery home. No resume, no owned tours. Returning users see the same view as guests.

Proposed home — Continue where you left off, My Tours at top After

Intent-aware home. Resume + progress at the top, owned tours below — discovery stays, just lower.

Tour cards — discovery

Current tour cards — photo and title only, no rating or price Before

Cards show photo + title only. Rating, price, duration hidden one tap deeper.

Proposed tour cards — rating, price, duration on the card After

Enriched cards surface rating, price, and type upfront. Everything needed to decide without tapping.

Prioritisation

Priority Opportunity Who it helps Impact
First bet Intent-aware home (returning first) Returning · ≈10k gap High
Supports A Enrich tour & bundle cards New + planners High
Next Reorder discovery home New + planners Med–High
Later Clarify tours vs content, trim nav All Med
Later Nearby / location fallback Planners Med

What I deliberately didn’t touch: onboarding, purchase/checkout, pricing, post-purchase, and audio/tour-start — all out of scope. I also resisted “fixing” new-user product-view rate (84.7% is already strong) and resisted pushing returning owners toward product views — that’s the wrong goal for them.

Design System

Airbnb Patterns, Shaka Brand

Reused Airbnb’s mature marketplace patterns — 16px photo cards, soft shadows, round-pill chips, whole-card taps — recolored to Shaka’s teal (#1C97AB) and navy (#123C4C). A trustworthy language users already know, made native to the brand.

Booking-Grade Trust at Card Level

Ratings to two decimals, 2px line icons (star & heart), plain-spoken microcopy (“Show map” not “SHOW MAP”), one-time price with strikethrough MSRP. Every cue already lived on the detail page — the card just surfaces it.

Reusable Everywhere

The enriched card component is one recolour of a mature design system. It extends consistently to home, All Tours, search, and bundle cards — no bespoke one-off work required.

How I’d Measure Success

0k Returning users addressable with intent-aware home
0pp Gap between new and returning product-view rate
0% New-user Home → Product rate (benchmark to protect)
  • Returning users: “Reached intended action” — resume / download / navigate. Leading signal: taps on the Continue / My Tours surface. Not product-view rate.
  • New / discovery users: Home → product-view rate and card tap-through by section.
  • Guardrails: Don’t slow the Decided user’s fast path; don’t let view-to-purchase quality slip.
  • Method: A/B the new home vs control, always reported split by segment — a blended readout would understate the returning win (and is how the problem was hidden in the first place).