monchis · Home Redesign Showcase

Context & problem

A home built for browsing,
for people who weren't browsing.

❌ Current home · before

Outdated · Disorganized

↓ and it kept
scrolling

The mandate

I wasn't handed a brief. I was asked to take the lead on the home.

61K users, 300K orders, 1,268 vendors, and a home everyone agreed was underperforming without agreeing on why. Open scope meant I owned the framing: research first, then two directions, then a defended decision. The output of this project isn't a screen. It's the argument that produced it.

Outdated & unattractive UI

Disorganized layout. Icons don't match their categories. A negative first impression that costs trust before the first tap.

Discoverability failure

Users can't quickly find products. Too much time searching → frustration → abandonment.

No reorder or personalization

Can't find favourite stores. Can't repeat the last purchase. The app feels generic to a user who orders the same lunch every week.

Content blind to context

The same promos at 8am and 8pm, for every user. Merchandising that ignores time, preference and location.

Research · Discovery

Nobody could say why
the home wasn't converting.

I was brought in without a defined scope: the home "worked", but nobody could explain why it wasn't converting. So instead of redesigning a screen, I started by defining the question the home actually had to answer: is this a place to show what monchis is, or a place to get to the next order fast? Everything that follows, including the two paths, comes from that tension.

Method 01

Behavioural data

Amplitude across 61K users and 300K orders. Where taps land, which verticals get entered, where sessions die.

Method 02

Active listening

User feedback and support inputs, coded into themes: discoverability, speed, personalization, promo visibility.

Method 03

Heuristic audit

Mapping what the old home declared as hierarchy vs. what it actually delivered above the fold.

Method 04

Marketplace benchmark

Regional delivery and marketplace patterns for header, search persistence and category legibility.

What the numbers said

60%

of users engage with search, on a static placeholder

Search was the front door, disguised as a form field

80%

go straight into Restaurants from home

Users arrive with intent, not curiosity

67%+

of category traffic is burgers & pizza

High intent, low browsing: they already decided

27%

of navbar taps go to cart, the highest of any tab

Cart is the real destination. Home is just a corridor

The pattern was consistent: people arrive knowing roughly what they want. The old home was built for browsing, for an audience that wasn't browsing.

What users said

Discoverability

It's very difficult to find the products I want; I have to click a lot.

Speed expectation

The app is very slow; I generally want to buy something to eat quickly.

Personalization gap

I always order the same thing for lunch, but I can't mark stores as favorites.

Context-awareness

I opened the app this morning and saw beer offers. The app doesn't know what time it is.

Shipping costs are high and it's very difficult to find stores with promotions.

Multi-store

I'm with my friends and I want to shop at different stores, but I can't.

Findings · What the research forced

Five findings,
and the constraint each one set.

I only kept findings that changed a decision. Each one is paired with the constraint it imposed on any design direction, which is exactly what made two paths possible instead of one.

1 / 5

Search was already the front door, dressed as a form field

60% of users engage with search, yet it sat as a static input with a generic placeholder, scrolling out of view the moment anyone explored.

Constraint

Search must be permanently reachable, and it has to say something.

Users arrive with intent, not curiosity

80% go straight into Restaurants; burgers and pizza alone carry 67%+ of category traffic. The home was optimising for discovery for people who had already decided.

Constraint

Everything actionable has to fit above the fold. Brand cannot own the first screen.

The product didn't know what time it was

Beer promos at 8am. The same content for every user, every hour. Users read this as the app not knowing them: a trust problem, not a merchandising one.

Constraint

Home content must be time- and preference-aware, or not shown at all.

Zero memory between sessions

No favourites, no reorder, no trace of the last purchase, in a product where the same lunch repeats week after week.

Constraint

The home needs a personalization slot high in the page, not a "recommended" afterthought.

Categories had to be decoded, not read

Icons didn't match their verticals and unlabeled imagery forced interpretation: a bag of groceries could plausibly be Mercado or Comidas.

Constraint

If it's a vertical, it carries a label. No guessing.

Wireframes · Exploration

The arguments were settled
in low fidelity.

Low fidelity was where the real arguments got settled: block order, how many verticals survive above the fold, where the personalization slot lives, and what happens to the header on scroll. It's also where the sharpest idea of the project appeared, a contextual box driven by the moment of the day, and where I worked out the two things that make an idea like that survivable: how much space it's allowed to take from commercial content, and what happens when the user dismisses it. Carrying those unresolved trade-offs into review is what made two complete paths necessary instead of one polished opinion.

What the wireframes resolved

Search as primary entry point, not a field

One component, four moments, instead of four sections

Banner placement matched to onboarding prefs

Search sheet and box share one context engine

The same box across one day. For the person, the right craving shows up without searching. For Monchis, every moment of the day is a new chance to turn a craving into an order.

The exploration · Time-of-day box

Where the box lives: after the user, before the catalogue

Greeting and favourites open the page, the offer banner and category row follow, and only then the contextual box: "¿Qué te tienta esta mañana?" with two stacked rails: an intent shortcut ("¿Fan de los desayunos?" · medialunas, café, tostados) and "Más favoriteados a esta hora". Commercial sections keep running underneath.

Rule

The box is dismissible, and if it's closed, it returns the next day with a different moment. Personalization that can't be refused stops being a service and becomes a banner.

The two paths · V1 vs V2

Two directions:
brand-led and task-led.

The findings pointed clearly at what the home had to do, but not at how much brand we were willing to trade for speed. Rather than pick for the business, I built both answers to the same brief, each internally consistent down to error and loading states. It moved the review from "I like this one" to "which cost are we choosing".

V1 · Brand-ledV2 · Task-led

Path V1 · Brand-led

The home as
a statement.

An extended red header owns the entry: saved address, greeting and search all live inside the brand block. Promo carousel takes the hero slot, verticals follow as an image mosaic.

Where it wins

✓ Address is impossible to miss: lowest risk of ordering to the wrong place

✓ The proximity alert ("estás lejos de tu casa") has a natural home inside the header

✓ Strongest brand recall on first frame

Where it costs

✕ The red block eats the first screen: verticals get pushed below the fold

✕ Search inherits the visual weight of the header and competes with the promo

✕ Unlabeled category mosaic leaves finding F05 unresolved

Path V2 · Task-led

The home as
a shortcut.

Neutral surface. Search is the first thing on screen, address one level below it. Labeled vertical cards move up; the promo becomes a module instead of the doorway.

Where it wins

✓ Search, address and all five verticals fit above the fold

✓ Photo + label kills the guessing game: Comidas, Mercado, Tiendas, Pick up, ueno

✓ Scales: a sixth vertical doesn't break the layout

Where it costs

✕ Much quieter brand presence on the first frame

✕ Promo loses the hero slot, and needed a separate conversation with commercial

✕ The proximity alert loses its container and has to be re-solved

V1 · Contextual alert & states

The red header earned its space by preventing an expensive error.

Ordering to the wrong address costs money, time and trust, so in V1 location isn't just displayed, it warns: "¡Cuidado! Estás lejos de tu casa. ¿Querés hacer el pedido igual?", dismissible and never blocking. The same pass covered the states nobody designs: skeleton loading on the real grid, and a persistent offline snackbar.

· Proximity alert docked inside the header block

· Dismissible with an explicit "order anyway" path

· Skeleton mirrors the final layout, no content jump

· Offline snackbar over the header, never blocking nav

V1: proximity alert and offline / loading states

V2 · Scroll behaviour & states

The header had to survive the scroll.

Putting search at the top only works if it stays there. In V2 the header collapses into a compact sticky bar (search plus abbreviated address) and a floating "Volver arriba" returns the full context without manual scrolling. What V2 had to solve separately was the alert: without a red block to dock into, the proximity warning becomes a contextual snackbar.

· Collapsed sticky header: search never leaves the screen

· Back-to-top after deep scroll

· Address abbreviated, never dropped

· Same skeleton and offline rules as V1: states are shared, not per-path

How I decided

Criterion, tied to a finding	V1 · Brand-led	V2 · Task-led
Time to first action (F02)	High: verticals below the fold	Low: everything above the fold
Search persistence (F01)	Scrolls away with the header	Sticky, collapses but stays
Vertical legibility (F05)	Image only, requires decoding	Photo + label
Room for personalization (F04)	Competes with hero promo	Dedicated high slot
Address error prevention	Highest: address is the headline	Needs a dedicated pattern
Brand presence	High	Medium, carried by the system
Category scalability	Limited	High

The call

V2 won on four of the five findings. V1 won on the one that costs real money.

So the final home isn't V2. It's V2's spine with V1's safeguards grafted on. Task-first structure, labeled verticals and the sticky search from V2; the contextual proximity alert and the address prominence from V1, re-expressed as a contextual snackbar instead of a permanent red block. The promo kept its reach by moving from doorway to module, which is the trade the commercial team ultimately signed off on.

The final version

What actually
shipped.

V2's structure survived contact with delivery, and the personalization layer the wireframes had been arguing for finally landed on top of it: favourite places, time-aware carousels, and a search bar that stopped being a field and started being a conversation.

Task-first spine · from V2

Search on top, labeled verticals above the fold, promo as a module. The sticky collapsing header carries search through the whole page.

Contextual safeguards · from V1

Proximity alert and address confirmation preserved, re-expressed as contextual snackbars rather than a permanent brand block.

The personalization layer · new

Favourite places, reorder and time-aware sections. This is the piece that later produced the highest CVR of any carousel on the page.

Final Monchis home on a tilted phone: search at the top with the saved address below, labeled verticals for Restaurantes, Mercados, Locales and Retiro en el local, a ueno bank promo module, food categories, and the bottom tab bar

Final · V2 spine + V1 safeguards

Feature focus · Animated dynamic search

Both paths agreed on one thing and disagreed on everything else: search had to lead. V2 won the structure argument, but this is the component that made the decision worth making: once search is permanently on screen, it stops being a field you fill and becomes the surface that carries context.

The search
that thinks.

The animated search bar is the emotional centerpiece of the new home. Instead of a static placeholder, it rotates contextual suggestions based on time of day, user history, and active promotions, making the app feel alive, personal, and aware.

Time-aware: "¿Ya desayunaste?" at 9am · "Para el almuerzo" at noon · "¿Qué cenamos?" at 7pm

Preference-matched: reflects onboarding choices: "Mirá el sushi con delivery gratis"

Promotion-aware: surfaces active deals: "Hamburguesas al 50% OFF hoy"

60% of users click the search: 23.2K unique users · 90.1K clicks (Amplitude, Nov 2025)

9:41

▪▪▪▪ WiFi ▪▪

¿Una comida rica?

💬

11 de septiembre de 1888, 2979... ›

🍔🥗🍦

Restaurantes

🥦🛒🍌

Mercados

🎧

Locales

🏪

Retiro en local

Locales favoritos

Chu

Chuchos

25 min

🍔

McDonald's

40 min

SUB

Subway

16 min

Burger King

25 min

🏠

Home

🏷

Promos

🛒

📋

Pedidos

👤

Cuenta

Time-aware · 9am

Rotates every 2.5s

60% click rate

Search placeholder variations

🌅

Morning

"¿Ya desayunaste?"

☀️

Lunch

"Hamburguesas o pizza"

🌆

Afternoon

"¿Algo fresco?"

🎯

Preference

"Sushi con delivery gratis"

🏷

Promo

"Pizzas al 50% OFF"

Design system · Home components

★ Hero component

The search that thinks.

The animated search bar is the emotional centerpiece of the new home. Instead of a static "¿Qué vas a pedir?" it rotates contextual suggestions based on time of day, user preferences and active promotions, making the app feel alive and personal.

Time-aware: "¿Ya desayunaste?" at 9am, "¿Algo para cenar?" at night

Preference-matched: reflects onboarding choices: "Sushi con delivery gratis"

Promotion-aware: surfaces active deals: "Hamburguesas al 50% OFF"

60% of users click the search: 23.2K unique · 90.1K clicks (Amplitude)

Buscar

"¿Una comida rica?" "Hamburguesas o pizza" "¿Algo fresco?" "¿Ya desayunaste?" "Sushi con delivery gratis" "Pizzas al 50% OFF"

↑ Live demo · rotating every 2.5s

Search input · Static state

Dynamic placeholder · Variations

A smiling woman at home holds a phone toward the camera showing the new Monchis home: search at the top, labeled verticals, the ueno bank promo and food categories

Full flow · End-to-end experience

From address to
checkout, in one flow.

Impact & metrics · Amplitude Nov 2025

What Amplitude showed
in November 2025.

44.6%

CVR · Killer feature: customization (Favorite Places)

60%

Users interested in search · 23.2K unique · 90.1K clicks

80%

Users enter Restaurants vertical from home

77.8%

Complete the onboarding flow

67%+

Hamburgers & pizzas dominate food categories traffic

60%

Checkout completion rate

99.1%

Address loading success rate (1'29'' load time)

90%

Users prefer to search by map location

Carousel CVR breakdown

Favorite Places

44.6%

Pick up without delays

16.0%

New places near you

15.4%

Ready for Burgers?

15.0%

Pizza Craving

10.5%

27%

Shopping cart

25%

Home

16%

Account

12%

Orders

Key design decisions

The seven decisions
behind it.

01
Two paths, not one proposal
V1 and V2 turned a taste debate into an explicit trade-off.
Process
02
Animated dynamic search
A static placeholder became rotating, contextual suggestions.
60% click it
03
Time-aware carousels
No more beer promos at 8 am. Content follows the moment of the day.
Trust
04
Favourite places as the hero
The best-performing carousel on the page, by far.
44.6% CVR
05
Personalized promo banners
Three types: top locals, onboarding match, payment incentive.
Relevance
06
Facelift first, features later
Ship the visual upgrade fast, then let data decide what to build.
Speed
07
A reusable component system
Documented components cut QA cycles for the next iterations.
Scale

The result

Personalization was the killer feature. Users convert when the product feels like it knows them.

What I'm taking to the next project

Show the trade-off, not an opinion.

Two fully resolved directions made the decision about cost, not taste.

Context is trust.

Knowing the time and the user changed how people felt about the app, and the numbers followed.

Documentation ships features.

A real system is a product investment: less QA, faster iterations.

The search that thinks.

The seven decisionsbehind it.

The seven decisions
behind it.