Pantry

Your kitchen, sorted. A voice-first, multi-surface smart grocery management platform for South African households.

Product Spike March 2026 SA Market

Executive Summary

The core insight: South African households manage groceries across multiple stores, multiple people, and multiple surfaces. No single product ties it all together. Pantry does.

The Problem

SA families juggle Woolies for premium items, Checkers Sixty60 for quick top-ups, and Makro for bulk buys. Grocery lists live across WhatsApp groups, Notes apps, and fridge whiteboards. Items get forgotten, food gets wasted (R400-R600/month per household), and budgets spiral.

The Solution

Pantry is a voice-first household grocery brain that passively tracks what you have, predicts what you need, and coordinates shopping across every household member and surface: mobile app, Google Nest Hub, WhatsApp, and web dashboard.

Key Numbers

R8K-12K

Monthly grocery spend (target household)

R400-600

Monthly food waste per household

Household personas identified

Interaction surfaces

How It Works

🧾 Passive Capture

Receipt scanning, delivery history integration, and predictive consumption modelling mean the system learns what you buy without you logging anything.

🎙 Voice-First

"Hey Google, tell Pantry we're out of milk." Kids, the domestic worker, grandma: everyone can talk to it. No app install needed.

💬 WhatsApp Bot

Grace already WhatsApps what's running low. Pantry just makes that message structured, trackable, and actionable.

📱 Multi-Surface

Melisa gets the full dashboard. Thabo gets a WhatsApp shopping list. Kids shout at the Nest Hub. Everyone uses what works for them.

Target User

Melisa, 38, Sandton. Marketing manager, two kids, domestic worker three days a week. Monthly grocery budget R8,000-R12,000. Shops at Woolies, Checkers Sixty60, and Makro. Uses iPhone, Google Nest Hub, and WhatsApp for everything. She doesn't want another app to manage: she wants the house to just tell her what it needs.

What's in This Report

👥 Personas

5 detailed household personas with interaction models

🚩 User Journeys

Per-persona, per-surface journey maps

📡 Passive Capture

6 data capture methods scored for SA feasibility

🎙 Voice Design

Google Nest Hub voice interaction scripts

🎨 UI Mockups

Mobile, Nest Hub, WhatsApp, Dashboard

🏆 Competitor Analysis

30+ products across 5 categories

🇿🇦 SA Ecosystem

Retailer landscape, loyalty programmes, delivery

📊 Market Sizing

TAM/SAM/SOM with ZAR figures

💰 Business Model

Revenue strategy and pricing

Persona Profiles

These personas represent the real people in a South African household who interact with Pantry, each with different needs, tech comfort levels, and motivations. Built from qualitative insights about middle-to-upper-income SA families managing groceries across Woolies, Checkers, Pick n Pay, and local delivery services.

👩

Melisa, the Household Manager

Marketing manager, 38, Sandton

Attribute	Detail
Household	Partner (Thabo), two kids (Lerato, 11 and Sipho, 7), domestic worker (Grace) three days a week
Monthly grocery budget	R8,000 - R12,000
Tech comfort	High. Uses iPhone, Google Nest Hub in the kitchen, WhatsApp for everything
Primary stores	Woolies (fresh, premium items), Checkers Sixty60 (quick top-ups), Makro (bulk monthly)

Goals

Never run out of essentials (milk, bread, eggs, kids' lunchbox items)
Reduce food waste: she throws away R400-R600 of expired food monthly and feels guilty about it
Control the budget without micromanaging every purchase
Delegate grocery tasks to Thabo and Grace without endless WhatsApp back-and-forth

Frustrations

"I only realise we're out of milk when someone complains at breakfast."
"Grace tells me we need things, but by the time I order, I've forgotten half the list."
"I have three different shopping lists going: Notes app, WhatsApp group, and the fridge whiteboard."
"Checkers Sixty60 is great for speed, but I overspend because I can't remember what we already have."

Key Quote: "I don't want another app to manage. I want the house to just tell me what it needs."

Interaction Model: Melisa is the approver and orchestrator. She reviews suggested lists, sets budget limits, and delegates shopping tasks. She wants dashboard-level visibility (web/mobile) and quick voice check-ins via the Nest Hub. She should never have to manually enter items.

🧒

Lerato (11) and Sipho (7), the Kids

Consumers and requesters

Attribute	Detail
Tech access	Lerato has a hand-me-down tablet (supervised). Sipho has none. Both use the Google Nest Hub.
Grocery role	Consumers and requesters. They eat the food and ask for more.

Goals

Get their favourite snacks without having to nag Mom repeatedly
Report what's finished in a way that actually reaches the shopping list
Sipho: just wants his yoghurt and doesn't care how it appears
Lerato: starting to help with meal planning for school lunches

Frustrations

"Mom said she'd buy more Frosties but she forgot again."
"I told Grace we need juice but nothing happened."
Lerato is old enough to be frustrated that her requests get lost in the household noise

Key Quote: "Can I just tell the kitchen we need more stuff?" - Sipho

Interaction Model: Kids interact exclusively through voice (Google Nest Hub) and possibly a simple visual display. Zero app, zero login, zero effort. Their requests go into a pending queue that Melisa can approve or dismiss.

👨

Thabo, the Partner/Co-parent

Software developer, 40, Rosebank office

Attribute	Detail
Tech comfort	Very high. Android user, but prefers not to install "yet another household app"
Grocery role	Occasional shopper, braai master, beer and snack buyer
Monthly personal grocery spend	R1,500 - R2,500

Goals

Help out when asked without having to learn a complex system
Pick up items on the way home when Melisa sends a list
Keep his braai supplies stocked (charcoal, boerewors, Mrs Ball's chutney)

Frustrations

"Melisa sends me a WhatsApp list while I'm driving and I can't read it properly."
"I bought 2% milk instead of full cream and you'd think I committed a crime."
"I don't mind helping, but I don't want to install an app I'll use twice a month."

Key Quote: "Just send me the list on WhatsApp. I'll get it done."

Interaction Model: Thabo's primary interface is WhatsApp. He receives assigned shopping lists with specific brands, sizes, and store suggestions. He can check items off in-chat, snap a photo of the receipt, and he's done.

👩‍🍳

Grace, the Domestic Worker

Frontline inventory manager, 52, Mon/Wed/Fri

Attribute	Detail
Tech comfort	Moderate. Uses WhatsApp and Facebook on a budget Android phone.
Data constraints	Limited mobile data (buys R50-R100 bundles). Wi-Fi access at Melisa's house.
Language	Fluent in Zulu and English. Prefers English for text, comfortable with either for voice.

Goals

Report what's running low quickly and reliably
Not get blamed when items run out
Keep things simple: she has enough to manage

Frustrations

"I WhatsApp Melisa when the Sunlight dish soap is almost finished, but she reads it hours later."
"Sometimes I tell her three things and she only buys one."
"I can't use too much data on apps."

Key Quote: "If I could just tell the machine and it goes on the list, that would save me so much trouble."

Interaction Model: Grace needs two channels: voice via Google Nest Hub (fastest while working in the kitchen) and WhatsApp for confirmation. The system must support Zulu voice input and low-data interactions.

👵

Gogo (Grandmother), the Extended Family Visitor

Temporary household member, 67, visits 2-4 weeks at a time

Attribute	Detail
Tech comfort	Low. Basic smartphone, uses WhatsApp (voice notes only).
Language	Primarily Zulu, conversational English
Grocery role	Temporary household member who changes consumption patterns significantly

Goals

Cook for the family using her own recipes (often requiring ingredients not usually stocked)
Not feel like a burden when requesting special items

Frustrations

"I need umfino (wild spinach) and Melisa doesn't know where to get it."
"These gadgets don't understand me when I speak."

Key Quote: "If Lerato can show me once how to tell the machine, I'll use it."

Interaction Model: Gogo is an edge-case user who reveals system flexibility. Her interaction is almost entirely through voice (Zulu preferred) or through Grace/Lerato as intermediaries.

Persona Comparison Matrix

Dimension	Melisa	Kids	Thabo	Grace	Gogo
Primary surface	Mobile app + Web dashboard	Voice (Nest Hub)	WhatsApp	Voice + WhatsApp	Voice (Zulu)
Role	Approver, orchestrator	Requester, consumer	Delegated shopper	Inventory scout	Temporary member
Tech comfort	High	Medium (voice-native)	Very high	Moderate	Low
Key need	Control + visibility	Zero-effort requests	Frictionless task completion	Reliable reporting	Inclusion + flexibility

Design Implications from Personas

Voice-first is non-negotiable. Three of five personas interact primarily through voice. The system must work well even if the app is never opened.

WhatsApp is the universal fallback. Every adult in the household already uses it. Building on WhatsApp reduces adoption friction to near zero.

Multi-language support is essential. Zulu voice recognition is not optional. It's required for Grace and Gogo, who are critical household contributors.

No persona should need to install an app. Only Melisa needs the full app experience. Everyone else should work through existing channels.

Data usage matters. Grace's data constraints mean WhatsApp interactions should be text-light, image-free, and optimised for low bandwidth.

The household is the user. Pantry must model a household with shifting membership, different permission levels, and varied interaction preferences.

User Journey Maps

Each journey maps a real scenario from a persona's perspective, showing touchpoints across surfaces, emotional states, and where Pantry adds value compared to current workarounds.

Journey 1: Melisa's Weekly Top-Up via Checkers Sixty60

▼

Persona: Melisa | Frequency: 2-3x per week | Surfaces: Mobile app, Web dashboard, Nest Hub

Phase	Action	Surface	Pantry's Role	Emotion
Morning check	Melisa asks "Hey Google, what do we need?" while making coffee	Nest Hub (voice)	Reads back prioritised list: "You're probably out of milk and eggs. Grace reported you're low on Sunlight. Sipho asked for yoghurt yesterday."	Informed, in control
Review	Opens Pantry app on phone during commute to review full list	Mobile app	Shows categorised list with confidence levels, estimated costs, and suggested store (Checkers Sixty60 for speed, Woolies for fresh)	Confident
Edit	Removes Sipho's request for sweets ("nice try"), adds avocados	Mobile app	Updates list, recalculates estimated total (R380)	Amused, efficient
Order	Taps "Send to Checkers Sixty60" button	Mobile app	Deep-links to Sixty60 with pre-populated cart (or copies list to clipboard if API isn't available)	Satisfied
Delivery	Receives Sixty60 delivery at home	None (passive)	Auto-detects order confirmation email, updates inventory: milk, eggs, yoghurt now "in stock"	Effortless
Feedback loop	Grace unpacks groceries, notices avocados are not ripe enough	Nest Hub (voice)	Grace says "The avocados need a few days." System notes this for future produce timing.	Heard

Key Insight: Melisa's journey spans three surfaces in one morning, each suited to the context: voice while hands are busy, mobile for detailed review, passive for delivery tracking.

Journey 2: Grace Reports What's Running Low

▼

Persona: Grace | Frequency: Each workday (Mon, Wed, Fri) | Surfaces: Nest Hub (voice), WhatsApp

Phase	Action	Surface	Pantry's Role	Emotion
Morning routine	Grace arrives, starts cleaning. Notices the dish soap is nearly empty.	Physical (observation)	None yet	Responsible
Voice report	"Hey Google, tell Pantry: Sunlight almost finished, and we need rice"	Nest Hub (voice)	Confirms: "Got it, Grace. I've added Sunlight dish soap and rice to the list."	Efficient, acknowledged
Midday check	While cooking lunch, notices the cooking oil is low	Nest Hub (voice)	Adds to list. Now shows 3 items from Grace today.	Thorough
WhatsApp confirmation	Receives WhatsApp message from Pantry bot	WhatsApp	"Hi Grace, today you reported: Sunlight dish soap, White Star rice 2kg, Sunfoil cooking oil 750ml. These are on Melisa's list."	Reassured, documented
End of day	Grace leaves for the day knowing her reports are captured	None	Melisa sees Grace's reports on her dashboard with timestamps	Professional, valued

Key Insight: Grace's journey replaces unreliable WhatsApp messages to Melisa with a structured system that still feels like talking. The WhatsApp confirmation gives Grace a record that she reported it.

Journey 3: Sipho Wants More Yoghurt

▼

Persona: Sipho (7) | Frequency: Ad hoc | Surface: Nest Hub (voice only)

Phase	Action	Surface	Pantry's Role	Emotion
Discovery	Sipho opens the fridge, sees the last yoghurt	Physical	None	Worried
Request	"Hey Google, tell Pantry we need more yoghurt!"	Nest Hub (voice)	"Okay Sipho, I'll let Mom know you want more yoghurt. I think you mean the Woolies strawberry one?"	Empowered
Confirmation	"Yes!"	Nest Hub (voice)	"Done! It's on the list. Mom will decide when to order it."	Happy, heard
Behind the scenes	Melisa gets a notification	Mobile app	Shows: "Sipho requested: Woolies Low Fat Strawberry Yoghurt 6-pack (R42.99)" with approve/dismiss buttons	(Melisa: amused, in control)

Key Insight: Sipho's interaction is under 15 seconds. No screen, no typing, no app. The system resolves "yoghurt" to a specific product using purchase history.

Journey 4: Thabo Picks Up Groceries on the Way Home

▼

Persona: Thabo | Frequency: 2-4x per month | Surface: WhatsApp

Phase	Action	Surface	Pantry's Role	Emotion
Trigger	Melisa assigns shopping task: "Thabo, can you grab these on the way home?"	Mobile app (Melisa)	Generates a curated list for Thabo, optimised for his usual store (Checkers Rosebank)	(Melisa: delegating)
Receive list	Thabo gets WhatsApp message from Pantry bot	WhatsApp	"Hi Thabo, Melisa asked you to pick up: Full cream milk 2L (NOT 2%, the blue one), Free range eggs x18, Woolies sourdough bread. Closest store: Checkers, Rosebank Mall"	Clear, no ambiguity
Shopping	Thabo shops, checks items off in WhatsApp	WhatsApp	Interactive checklist. Taps/replies to mark items as found.	Efficient
Substitution	Sourdough is sold out. Thabo replies "they don't have it"	WhatsApp	"No worries. Skip it or grab the Checkers ciabatta instead? I'll let Melisa know."	Supported, not stressed
Checkout	Thabo pays, snaps a photo of the receipt	WhatsApp (photo)	Processes receipt photo, updates inventory. "Got it! R187.50 at Checkers Rosebank. Inventory updated."	Done, easy

Key Insight: Thabo never opens an app. The entire experience happens in WhatsApp. Brand-specific instructions ("the blue one") prevent the wrong-item problem.

Journey 5: Monthly Makro Bulk Shop

▼

Persona: Melisa + Thabo | Frequency: Monthly | Surfaces: Web dashboard, Mobile app

Phase	Action	Surface	Pantry's Role	Emotion
Planning	Melisa reviews the monthly shopping plan on her laptop	Web dashboard	Auto-generated list based on: consumption history, current stock predictions, and items flagged low during the month. Shows estimated total: R4,200.	Prepared
Budget review	Compares to budget, removes non-essentials, adds Gogo's items	Web dashboard	Budget tracker updates in real time. Flags items over historical price. Visitor mode activated for Gogo.	In control
Shopping	Melisa and Thabo go to Makro together	Mobile app	Store-optimised list sorted by Makro's layout. Checklist mode for quick item marking.	Efficient
Checkout	Makro mCard scanned at till	Passive	Auto-imports purchase data. "Monthly shop complete: R4,087. Under budget by R113."	Accomplished

Key Insight: The monthly shop is where Pantry delivers the most strategic value: predictive list generation, budget tracking, and visitor anticipation.

Journey 6: Gogo Arrives for a Two-Week Stay

▼

Persona: Gogo + Melisa | Frequency: Several times per year | Surfaces: Voice (Nest Hub), Mobile (Melisa)

Phase	Action	Surface	Pantry's Role	Emotion
Arrival	Melisa tells Pantry: "Gogo is visiting for two weeks"	Mobile app / Voice	Activates visitor mode. Adjusts consumption predictions upward. Suggests stocking traditional ingredients.	Welcoming
First morning	Gogo wants phutu for breakfast, needs maize meal	Nest Hub (voice, Zulu)	Gogo: "Ngidinga impuphu" (I need maize meal). System responds in Zulu: "Ngizokufaka ohlwini, Gogo."	Included, comfortable
Melisa reviews	Sees Gogo's requests on her app	Mobile app	Shows "Gogo requested: Iwisa maize meal 5kg, dried sugar beans" tagged as visitor items.	Thoughtful
Departure	Gogo leaves after two weeks	Mobile app	Deactivates visitor mode. Reverts consumption predictions. "Gogo's visit added R1,200 to grocery spend."	Back to routine

Key Insight: Gogo's journey tests the system's flexibility and cultural sensitivity. Zulu voice support is not a nice-to-have, it's essential for inclusion.

Cross-Journey Patterns

Pattern	Observation	Design Implication
Multi-source list assembly	The shopping list is never built by one person. It's assembled from Grace's reports, kids' requests, predictive signals, and Melisa's additions.	The list is a living document with attributed sources and confidence levels.
Approval gates	Melisa always reviews before anything is ordered. She's the household "admin".	Permission model: anyone can add, only Melisa (and optionally Thabo) can approve/order.
Context-appropriate surfaces	Voice when hands are busy, WhatsApp when mobile, web for planning.	Each surface optimised for its context. Never force a user to switch surfaces.
Confirmation loops	Every input gets a confirmation: voice gets audio, WhatsApp gets text, app shows real-time updates.	Build trust through reliable feedback. "The system heard me" is baseline.
Store awareness	Different items come from different stores. Woolies for fresh, Makro for bulk, Checkers for speed.	Store-aware list generation and routing. Don't suggest buying 10kg rice on Sixty60.

Passive Data Capture Strategy

The core problem Pantry solves is knowing what's in the household without anyone having to manually log it. This section maps every realistic data capture method, scores it for feasibility in the SA market, and recommends a phased rollout.

The Data Entry Problem: Every smart pantry product fails at the same point: data entry fatigue. Users enthusiastically scan barcodes for two weeks, then stop. The system goes stale and trust collapses. Pantry must solve this with a layered approach where no single method carries the full burden.

1. Receipt and Invoice Parsing (Items In)

▼

Aspect	Detail
How it works	Photo of till slip, email receipt forwarding, or direct API integration with retailer loyalty programmes
SA feasibility	High. Woolies, Checkers, and Pick n Pay all issue detailed digital receipts via their apps and loyalty programmes
Data quality	Excellent: SKU-level detail, exact quantities, prices, store, date, and time
User effort	Low to zero. Email forwarding rules or API integration require one-time setup.

SA-Specific Implementation

Checkers Sixty60 / Woolies Dash: Order history is available via their apps. API integration gives exact item data for delivery orders.
Email receipt parsing: Woolies and Checkers both email receipts to loyalty members. A forwarding rule handles this passively.
Till slip OCR: Fallback for in-store purchases. SA till slips are standardised enough for reliable OCR.
Makro card: Makro's mCard tracks purchases. Bulk items can be auto-imported.

Confidence: 9/10 This is the primary "items in" signal.

2. Voice Reporting via Smart Speaker (Items Running Low)

▼

Aspect	Detail
How it works	Household members tell the Google Nest Hub when something is running low or finished
SA feasibility	High. Google Nest Hub is available in SA. English voice recognition is solid. Zulu support is emerging.
Data quality	Good for "items out" signals. Fuzzy: "the crunchy cereal" needs product resolution.
User effort	Minimal. Speaking is faster than typing.

Voice Capture Scenarios

Direct report: "Hey Google, tell Pantry we're out of milk"
Fuzzy request: "Hey Google, tell Pantry we need more of Sipho's yoghurt" (resolves to Woolies low-fat strawberry yoghurt 6-pack based on history)
Grace's report: "Hey Google, tell Pantry: Sunlight almost finished, need rice, and eggs" (batch input)

Confidence: 8/10 Voice is the most natural "items out" signal.

3. Predictive Consumption Modelling (Estimated Depletion)

▼

Aspect	Detail
How it works	System learns consumption rates from purchase history. If 2L milk is bought every 4 days, it predicts when milk will run low.
SA feasibility	High. Requires only purchase history data, which comes from receipt parsing.
User effort	Zero. Fully passive once purchase data flows in.

How the Model Works

Baseline period: 4-6 weeks of purchase data to establish consumption cadence per item
Categorisation: Items classified as staples (regular cadence), occasionals (irregular), or one-offs
Depletion estimate: "You bought 2L full cream milk 3 days ago. Based on your pattern, you'll need more in about 2 days."
Correction loop: Voice reports calibrate the model. If milk ran out a day early, the consumption rate adjusts.

Confidence: 7/10 The "smart" layer that makes Pantry feel proactive.

4. Barcode and Camera Scanning (Optional, Active)

▼

Confidence: 5/10 Optional, not core. Nice for power users, but the system must work well without it. Never gate core functionality behind barcode scanning.

When Scanning Makes Sense

Initial pantry setup: One-time scan of everything in the pantry to bootstrap the system
Unpacking after a big shop: Scanning items while unpacking the monthly Makro haul
Discard scanning: Scan the barcode on an empty container before throwing it away

5. Smart Device Integration (Future)

▼

Confidence: 2/10 Park this for v3+. SA's infrastructure and price points don't support this yet. Focus on software-first solutions.

6. WhatsApp Micro-Reports (Items Out, Low-Friction)

▼

Aspect	Detail
SA feasibility	Very high. WhatsApp is universal in SA. Zero app install needed.
User effort	Minimal. Same action Grace is already doing, just to a different contact.

Grace's Workflow: Current vs. Pantry

Current	With Pantry
WhatsApps Melisa: "The Sunlight is almost finished and we need rice"	WhatsApps Pantry bot: same message. Bot parses it, adds to list, confirms back.
Melisa reads it hours later, forgets half	Items appear on Melisa's dashboard immediately, flagged as "reported by Grace"
No record, no tracking	Grace gets confirmation: "Added: Sunlight dish soap, White Star rice 2kg"

Confidence: 9/10 This is the killer feature for the SA market.

Recommended Layered Capture Strategy

Priority	Method	Signal Type	Phase	User Effort
1	Receipt/invoice parsing	Items In	MVP	One-time setup
2	WhatsApp micro-reports	Items Out / Running Low	MVP	Near zero
3	Voice reporting (Nest Hub)	Items Out / Running Low	MVP	Minimal
4	Predictive consumption model	Estimated Depletion	v1.1	Zero (passive)
5	Barcode scanning	Items In / Items Out	v1.1	Active (optional)
6	Smart device integration	Real-time inventory	v3+	Zero (hardware)

Data Confidence Framework

High confidence (green): Receipt-confirmed purchase + within predicted consumption window. "You bought 2L milk 2 days ago. Likely still have some."

Medium confidence (amber): Purchase confirmed but nearing predicted depletion. "You might be running low on eggs."

Low confidence (red): Past predicted depletion date with no repurchase signal. "You're probably out of bread."

Unknown (grey): No data. Item hasn't been tracked yet or was purchased at an untracked store.

Voice Interaction Scripts

These scripts define how Pantry sounds and behaves on the Google Nest Hub. Voice is the primary interface for three of five personas (Kids, Grace, Gogo), making these interactions critical to product success.

Voice Design Principles

Warm, not robotic. Pantry sounds like a helpful household member, not a call centre IVR.

Brief by default. Responses are short. Detail is available on the Nest Hub screen or by asking follow-up questions.

Forgiving. Fuzzy language ("the crunchy cereal", "Sipho's yoghurt") is resolved using purchase history and household context.

Speaker-aware. Recognises household members by voice (Google Voice Match) and adjusts behaviour: kids' requests go to approval queue.

Multilingual. English and Zulu supported. Responds in the language spoken to it.

Non-blocking. Never asks more than one clarifying question. If unsure, make the best guess and confirm.

Script 1: Adding Items to the List

▼

Scenario A: Simple single item (Grace)

Grace: "Hey Google, tell Pantry we need more rice."
Pantry: "Got it, Grace. I've added White Star rice, 2 kilo bag, to the list. Is that right?"
Grace: "Yes."
Pantry: "Done. It's on Melisa's list."

System resolves "rice" to the household's usual brand and size from purchase history.

Scenario B: Multiple items (Grace, batch report)

Grace: "Hey Google, tell Pantry: we need Sunlight, eggs, and the blue milk."
Pantry: "Adding three items: Sunlight dish soap, eggs eighteen pack, and full cream milk two litres. All on the list."

Batch input is critical for Grace's workflow. "Blue milk" resolved to full cream from purchase history.

Scenario C: Fuzzy request (Sipho)

Sipho: "Hey Google, tell Pantry I want more of the crunchy cereal!"
Pantry: "Hey Sipho! I think you mean Kellogg's Crunchy Nut. I'll ask Mom to add it to the list."

Kids' requests are routed to Melisa's approval queue, not added directly.

Scenario D: Zulu input (Gogo)

Gogo: "Hey Google, tshela iPantry ngidinga impuphu ne-sugar beans."
Pantry: "Kulungile, Gogo. Ngifake impuphu ne-sugar beans ohlwini."

Responds in the language spoken. Zulu product names mapped to store product catalogue.

Script 2: Checking What's Needed

▼

Morning check (Melisa)

Melisa: "Hey Google, ask Pantry what we need."
Pantry: "You've got 7 items on the list. Top priority: milk and eggs are probably out. Grace reported you're low on dish soap and rice. Sipho wants yoghurt. Want me to read the full list?"
Melisa: "No, that's fine. Send it to my phone."
Pantry: "Sent. You can review it in the app."

Quick status check (anyone)

Thabo: "Hey Google, ask Pantry: do we have milk?"
Pantry: "Probably not. You bought milk 4 days ago and usually run out after 3. It's already on the list."

Predictive status based on consumption modelling. Honest about uncertainty.

Budget check (Melisa)

Melisa: "Hey Google, ask Pantry how much we've spent this month."
Pantry: "So far this month, you've spent R6,340 on groceries. That's about R1,600 under your R8,000 budget with 10 days left."

Financial data only shared with Melisa and Thabo (voice-matched).

Script 3: Removing and Modifying Items

▼

Melisa: "Hey Google, tell Pantry to remove the sweets from the list."
Pantry: "Done. I've removed the jelly sweets that Sipho requested."

Melisa: "Hey Google, tell Pantry to switch to Tastic rice from now on."
Pantry: "Noted. I'll default to Tastic rice for future orders. Want me to update the current list too?"
Melisa: "Yes."
Pantry: "Done. Tastic rice 2 kilo bag is now on the list."

Brand preferences are persistent. Once set, the system remembers for future voice-to-product resolution.

Script 4: Delegating Shopping Tasks

▼

Melisa: "Hey Google, tell Pantry to send the milk, eggs, and bread to Thabo."
Pantry: "Sending those three items to Thabo on WhatsApp now. He'll get them with store suggestions and brand details."

Melisa: "Hey Google, tell Pantry to order the top-up list on Sixty60."
Pantry: "I've prepared a Checkers Sixty60 order with 5 items, estimated total R340. Check the app to confirm and place the order."

Voice can initiate an order but cannot complete it. Final checkout always requires visual confirmation.

Script 5: Proactive Notifications (System-Initiated)

▼

Triggered by the system, not the user. They appear as gentle nudges on the Nest Hub screen.

Pantry (screen): "Heads up: you're probably running low on milk and eggs. Add to list?"
Melisa: "Hey Google, yes, add them."
Pantry: "Added milk and eggs to the list."

Proactive alerts are visual-first (screen card), not voice-first. Nobody wants their kitchen speaker randomly announcing grocery needs.

Pantry (to Melisa): "Gogo is coming next week. Based on her last visit, you might want to stock up on maize meal, cooking oil, and dried beans. Want me to add them?"

Script 6: Error Handling and Edge Cases

▼

Grace: "Hey Google, tell Pantry we need... [unclear]"
Pantry: "Sorry Grace, I didn't catch that. Could you say it again?"

One retry, then offer alternatives: "You can also WhatsApp me the items if voice isn't working."

Melisa: "Hey Google, tell Pantry to add milk."
Pantry: "Milk is already on the list, added by Grace this morning. Want me to increase the quantity?"

Lerato: "Hey Google, tell Pantry to order chips on Sixty60."
Pantry: "I can't place orders, Lerato, but I've added chips to the request list. Mom will see it and decide."

Kids can request, never order. The system recognises the voice and applies appropriate permissions.

Voice Personality Guidelines

Trait	Do	Don't
Tone	Warm, helpful, like a friendly family member	Robotic, overly formal, condescending
Brevity	"Milk added." / "Got it, Grace."	"I have successfully added full cream milk, 2 litres, to your household shopping list."
Certainty	"You're probably low on eggs" (honest)	"You are out of eggs" (can't know for sure)
Names	Use first names when voice-matched: "Got it, Grace"	Generic: "Item has been added to the list"
Kids	Friendly, encouraging: "Great idea, Sipho!"	Same formal tone as adults
Language	Match the speaker's language (English or Zulu)	Always respond in English regardless of input

Multi-Surface UX Framework

Pantry lives across four distinct surfaces, each optimised for different contexts, personas, and tasks. This framework defines what each surface does, what it doesn't do, and how they work together as a coherent system.

The Four Surfaces

Surface	Primary Personas	Context	Core Capability
Google Nest Hub	Grace, Kids, Gogo, Melisa	Kitchen, hands busy, quick interactions	Input (add items, report stock), quick status checks
WhatsApp Bot	Grace, Thabo, Melisa	On the go, at the store, asynchronous	Task delegation, shopping lists, confirmations, receipt capture
Mobile App (PWA)	Melisa, Thabo	Commute, couch, quick review moments	List management, approvals, ordering, notifications
Web Dashboard	Melisa	Evening planning, laptop	Analytics, budget tracking, monthly planning, settings

Permission Model Across Surfaces

Action	Melisa	Thabo	Grace	Kids	Gogo
Add items to list	Direct	Direct	Direct	To approval queue	Direct
Remove items	Yes	Yes (own)	No	No	No
Approve requests	Yes	Yes	No	No	No
Place orders	Yes	Yes (with limit)	No	No	No
View budget	Yes	Yes	No	No	No
Assign tasks	Yes	No	No	No	No
Change settings	Yes	Limited	No	No	No

Progressive Disclosure by Surface

Level	Nest Hub	WhatsApp	Mobile	Web
Glanceable	"5 items needed"	"3 items to pick up"	Badge count on icon	KPI cards
Summary	Voice readback of top items	Text list with brands	Categorised list view	Dashboard overview
Detail	N/A (go to mobile)	N/A (go to mobile)	Item cards with source, confidence, price	Full analytics, trends
Deep analysis	N/A	N/A	N/A	Historical trends, budget modelling

Notification Routing

Not every notification goes to every surface. Route by relevance and context:

Event	Nest Hub	WhatsApp	Mobile Push	Web
Item running low	Screen card (silent)	No	Yes (Melisa)	Dashboard badge
Item reported "out"	Voice confirmation	Confirmation to reporter	Yes (Melisa)	Activity feed
Shopping task assigned	No	Yes (to assignee)	Yes (to assignee)	No
Budget threshold (80%)	No	No	Yes (Melisa)	Alert banner
Deal on frequent item	Screen card (silent)	No	Optional	Specials section

Offline and Degraded Mode

SA's connectivity can be unreliable (load shedding, network issues). Each surface must handle degradation:

Surface	When Offline	When Back Online
Nest Hub	Queues voice commands locally: "Saved. I'll sync when we're back online."	Syncs queued items, resolves product names
WhatsApp	Messages queue in WhatsApp's own delivery system	Bot processes messages in order received
Mobile (PWA)	Full list management offline (cached data). Shows "last synced" timestamp.	Two-way sync with conflict resolution
Web	Read-only cached view. Shows "offline" banner.	Full refresh on reconnect

Cross-Surface Design Principles

Start Anywhere, Finish Anywhere. Grace reports "need rice" on the Nest Hub. Melisa reviews it on her phone. She assigns it to Thabo on WhatsApp. He snaps the receipt. No surface is a dead end.

Consistent Data, Adapted Presentation. The same shopping list appears on every surface, but formatted for the context: voice summary, checklist format, interactive cards, or full analytical table.

UI Mockups

Interactive mockups across all four Pantry surfaces: mobile app, Google Nest Hub, WhatsApp bot, and web dashboard.

Mobile App Mockups: Four core screens for the Pantry mobile app, designed for one-handed use while cooking, shopping, or managing the household. Dark mode default with warm, kitchen-friendly colours. All prices in ZAR, all stores South African. View full mobile mockups

The mobile app includes four key screens: Dashboard (quick-glance overview with budget tracker, alerts, and list preview), Pantry Inventory (full inventory with confidence-based stock indicators), Shopping List (grouped by store with price comparison), and Receipt Scan (camera-based OCR with SA retailer matching).

Dashboard

Budget status bar (R3,420 / R5,000), needs-attention alerts (eggs out, milk running low), this week's list preview (12 items, est. R684), and quick actions (voice add, scan, full list).

Pantry Inventory

Category tabs (All, Fridge, Pantry, Cleaning, Frozen). Colour-coded stock indicators: green (well stocked), amber (running low), red (out of stock). Search and filter, pull-to-refresh.

Shopping List

Grouped by store (Woolworths 5 items R198.95, Checkers 4 items R245.96, PnP 3 items R152.97). Price comparison chips, swipe-to-check-off, estimated total with budget remaining.

Receipt Scan

Camera viewfinder with alignment guide. Auto-detects retailer (Woolworths, Checkers, PnP). Parsed items with confidence scores, edit-before-confirm flow, running total calculation.

Google Nest Hub Mockups: Three key screens for the Nest Hub (landscape, 1024x600). Designed for arm's-length reading with large type and high contrast. View full Nest Hub mockups

Ambient Display

Always-on kitchen companion showing items running low (eggs out, milk ~1 day left), today's meal plan, and household status metrics (78% stocked, R3.4k spent, 12 items on list).

Voice Response Cards

Confirmation cards when adding items ("Added Full Cream Milk 2L, Woolworths R32.99") and summary cards for list queries ("13 items, R717 est., 3 stores"). Auto-dismiss after 10 seconds.

Family Scoreboard

Gamified household tracking with leaderboard (Thandi 2,450pts, Sipho 1,820pts), achievements (Budget Boss, Zero Waste), and household stats (waste score 92, R240 saved).

WhatsApp Bot Mockups: Four conversation flows showing natural language interaction, inline actions, and smart formatting. View full WhatsApp mockups

Quick Add Flow

"Running low on milk" triggers smart matching (Full Cream Milk 2L, R32.99 at Woolworths). Quick reply buttons: Looks good, Change to low-fat, Buy at Checkers, Remove.

Weekly Summary

Sunday evening overview: spending this week (R1,180), pantry status (78%, 4 running low), next week's draft list (13 items, R684 est.), smart suggestions (switch chicken to Checkers, save R12).

List Approval

Store-grouped list (Woolworths 5 items R198.95, Checkers 4 items R245.96, PnP 3 items R152.97). Natural language edits: "Remove the washing powder and add dishwashing liquid". New total recalculated instantly.

Exception Alerts

Proactive notifications: usage alerts (Jungle Oats finished 2 weeks early), price alerts (Woolworths chicken up 15%, Checkers alternative saves R12), budget warnings (80% of monthly budget hit).

Web Dashboard Mockups: Four full-width desktop screens showing Pantry's command centre for household analytics, historical trends, and family management. Built for focused evening sessions on a laptop. View full dashboard mockups

Analytics Dashboard

Spend by category (fresh produce R2,340, cleaning R890, pantry staples R1,560), store comparison bar chart (Woolworths R3,200, Checkers R2,100, PnP R1,480), waste tracking score (92/100), and monthly budget progress (R6,780 of R8,500).

Historical Trends

Month-over-month spending line charts, category breakdown over 6 months, seasonal patterns with annotations (December spike R12,400), price tracking per item, and year-over-year savings calculation (R4,200 saved).

Family Management

Member profiles with role-based permissions (Melisa: Admin, Thabo: Editor, Gogo: Voice Only), activity feed, device management (3 active: iPhone, Nest Hub, WhatsApp), and notification preferences per member.

Inventory Overview

Full pantry grid with stock levels, expiry tracking (3 items expiring this week), category filters, search, bulk actions, and smart reorder suggestions based on consumption velocity.

Design System

Pantry's visual language is warm, practical, and kitchen-inspired. It avoids the cold, corporate feel of typical SaaS tools and instead feels like a trusted household companion. The palette draws from natural, earthy tones: terracotta, warm sand, sage green, and deep charcoal.

Colour Palette

Two modes: dark (default) and light. Dark mode uses deep, warm backgrounds that feel cosy on kitchen countertops and bedside tables. Light mode flips to clean, airy surfaces for daytime use.

Charcoal
#1a1614

Cocoa
#2a2320

Terracotta
#c4653a

Warm Sand
#d4915a

Sage
#7a9e7e

Honey
#e8c468

Paprika
#c45c5c

Cream
#f0e6d3

Typography

Inter is the primary typeface: clean, highly legible, and excellent at small sizes on mobile screens.

Your Kitchen, Sorted

Weekly Grocery List

Budget Overview

Running Low Items

Your milk is running low. We've added it to this week's list at R32.99 from Checkers.

Last purchased 4 days ago

Design Principles

01. Kitchen-counter ready. Every screen should feel at home on a kitchen counter, bedside table, or in a busy parent's hand. Large touch targets, high contrast, glanceable information.

02. Smart defaults, easy overrides. The system should predict what you need, but never lock you in. One tap to approve, two taps to change.

03. Voice-first, screen-supported. Every action should be possible by voice alone. The screen adds context and confidence, not complexity.

04. Budget-aware always. ZAR pricing is never an afterthought. Running totals, store comparisons, and budget alerts are embedded in every flow.

05. Familiar, not foreign. Reference South African brands, stores, and shopping habits. Woolies, Checkers, Pick n Pay. Not Walmart, not Tesco.

Component Patterns

Status Badges

In Stock Running Low Out of Stock On List Paused

Device Frames

Mockups use consistent device frames: 375 x 812px for mobile (iPhone viewport), 1024 x 600px for Google Nest Hub, and standard chat-width for WhatsApp.

Competitor Landscape Analysis

Pantry enters a fragmented market with no single product combining inventory tracking, voice-first interaction, and SA grocery integration. Below is a comprehensive mapping of the competitive landscape across five categories.

1. SA Grocery Delivery Apps

The SA on-demand grocery delivery market is dominated by Checkers Sixty60, which holds over 80% market share among higher-income households.

Platform	Strengths	Gaps	Delivery Fee
Checkers Sixty60	Market leader (80%+ share). 694 locations. 60-min delivery. 47.7% sales growth (FY2025).	No inventory tracking. No voice integration. No cross-store shopping.	R36 (free on R500+)
Woolies Dash	Premium product range. 130 sites + 1 dark store. 40% YoY growth.	No inventory management. No voice ordering. Higher delivery fee.	R45
Pick n Pay asap!	600 locations. 48.7% online growth (FY2025). Re-platformed app with AI features.	No pantry tracking. No voice assistant. App still maturing.	R35
OneCart	Multi-store shopping (Woolworths, Dis-Chem, Pick n Pay, Makro) with one delivery fee per mall.	No inventory features. Limited brand awareness.	Varies
SPAR2U	525 stores. 380% volume growth YoY.	Late entrant. Less established UX. No smart features.	Varies

2. Grocery List and Inventory Apps

These apps solve list management and sharing but none offer deep SA store integration, voice-first design, or passive inventory tracking.

App	Strengths	Gaps	SA-Specific?
AnyList	Top all-rounder. Real-time syncing. Recipe organisation.	No SA store integration. No passive tracking.	No
OurGroceries	Simple shared lists. Google Assistant integration.	No inventory tracking. No SA store data.	No
Bring!	Visual, icon-based (great for non-tech-savvy users).	No inventory management. No SA retailers.	No
Listonic	Smart suggestions from past habits. Voice input.	No pantry tracking. No SA-specific data.	No

3. Smart Kitchen / Connected Appliances

Product	Strengths	SA Relevance
Samsung Family Hub	AI Vision Inside, expiry tracking, shopping list. R40,000+.	Available but very niche. High price limits adoption.
LG InstaView ThinQ	AI camera for food recognition. Knock-to-see-inside.	Limited SA availability. Premium pricing.

4. Voice Assistants + Grocery

Platform	Strengths	SA Relevance
Google Shopping List	Voice-add via Nest Hub. Lists in Google Keep. Free.	Google Nest Hub is available in SA. This is the voice platform Pantry should build on.
Alexa Shopping List	Voice-add via Echo. Alexa+ adds agentic grocery shopping.	Low. Echo has limited SA penetration. No local grocery partners.

Competitive White Space: Where Pantry Wins

No existing product combines these four capabilities:

Passive inventory tracking that knows what you have at home
Voice-first interaction via Google Nest Hub for hands-free kitchen use
SA grocery ecosystem integration with Checkers, Woolworths, Pick n Pay
Intelligent replenishment that learns household patterns

Pantry's moat is the integration layer: connecting what you have, what you need, and where to buy it, all optimised for the South African market.

SA Grocery Ecosystem Deep Dive

South Africa's grocery retail sector is highly concentrated among four major chains, each with distinct digital capabilities, loyalty programmes, and delivery infrastructure. Pantry must integrate deeply with this ecosystem to deliver value.

1. Major Retailer Digital Capabilities

Checkers / Shoprite Group (Sixty60)

▼

Dimension	Details
Delivery	Checkers Sixty60: 60-minute delivery from 694 locations. 80%+ of addressable online market. 47.7% sales growth FY2025.
Loyalty	Xtra Savings: Free tier with personalised deals. Plus: R99/month for unlimited free deliveries, double deals, 10% off one shop per month.
Delivery Fee	R36 (from July 2025). Free on R500+ orders. Free with Xtra Savings Plus.
Integration	No public API. Product catalogue accessible via web scraping. Deep linking into app for checkout possible.

Woolworths (Woolies Dash)

▼

Dimension	Details
Delivery	Woolies Dash: on-demand from 130 sites + 1 dark store. 90%+ of customer base. 40% YoY growth.
Loyalty	WRewards: Vouchers, promotional pricing, credit card linking. Strong brand loyalty among LSM 8-10.
Delivery Fee	R45 (from November 2025). Premium positioning matches brand.
Integration	No public API. WRewards card linking could surface personalised pricing. Deep linking into Woolworths app.

Pick n Pay (asap!)

▼

Dimension	Details
Delivery	PnP asap!: on-demand from 600 locations (expanded from 47 in mid-2025). Also on Mr D app. 48.7% online turnover growth FY2025.
Loyalty	Smart Shopper: Points-based, digital signup, in-app tracking. Historically largest SA grocery loyalty programme.
Delivery Fee	R35. Pre-authorisation model: charged only for delivered items.
Integration	Re-platformed app may be more integration-friendly. Mr D platform (Naspers) has API infrastructure.

2. Price Comparison

Factor	Checkers	Woolworths	Pick n Pay
Positioning	Value-to-mid. Aggressive promotions.	Premium. Higher prices, quality perception.	Mid-range. Competitive pricing.
Delivery Fee	R36 (or free)	R45	R35
Subscription	Xtra Savings Plus: R99/mo	None	None
Online Markup	Generally matches in-store	Slight premium on some items	Generally matches in-store

Pantry opportunity: Build a price comparison engine that shows users the best store for their specific basket, factoring in delivery fees, loyalty discounts, and current promotions.

3. Loyalty Programme Landscape

Programme	Members	Key Benefits	Pantry Value
Xtra Savings	30M+ cards	Personalised deals, Plus tier R99/mo	Surface personalised deals, alert to Plus savings based on order frequency
WRewards	5M+ members	Vouchers, promotional pricing	Show WRewards pricing, remind users to redeem vouchers before expiry
Smart Shopper	10M+ members	Points on purchases, digital tracking	Track points earned, suggest spend thresholds for bonus points

4. Integration Strategy

Phase 1 (MVP)

Web scrape product catalogues for matching and price comparison. Deep link to retailer apps for "Order Now" handoff. Allow users to link loyalty card numbers.

Phase 2 (Growth)

Pursue formal API partnerships (likely starting with PnP via Mr D/Naspers). Enable in-app ordering. Integrate real-time stock availability.

Phase 3 (Scale)

Become a super-aggregator: one basket, best price across stores, split orders. Negotiate affiliate/commission arrangements.

Market Sizing Analysis

This analysis sizes the opportunity for Pantry in the South African market, using a top-down TAM/SAM/SOM framework with locally relevant data points.

1. TAM / SAM / SOM

R1.1T

TAM: All SA grocery spending

R80B

SAM: Online grocery (LSM 7-10, urban)

R240-480M

SOM: Subscription revenue (Year 3)

2.8M

Target households (LSM 8-10 urban)

Revenue model context: Pantry's SOM is not a share of grocery spend but rather a subscription and commission layer on top of existing spending. The R100/month subscription covers smart inventory tracking, voice ordering, price comparison, and automated replenishment.

2. Target Demographics: LSM 8-10 Urban

Characteristic	LSM 8	LSM 9	LSM 10
Monthly Income	R13,200 - R19,000	R19,000 - R25,000	R25,000+
Smartphone	85%+	95%+	98%+
Online Shopping	Growing adoption	Regular	Frequent
Est. Households	~1.5M	~800K	~500K

3. Online Grocery Adoption

Metric	Value
E-commerce penetration	49% (2025), forecast 60% by 2028
Online grocery market	R80B projected by 2026
Growth rate	25-35% CAGR (5-7x faster than overall grocery)
Checkers Sixty60 growth	47.7% sales growth FY2025
Key drivers	Convenience, load shedding, safety concerns, COVID habit persistence
Key barriers	Delivery fees, can't select own fresh produce, data costs, payment trust

4. Growth Projections

Metric	Year 1	Year 2	Year 3	Year 5
Registered Households	25,000	120,000	400,000	1,000,000
Paid Subscribers	2,500 (10%)	18,000 (15%)	80,000 (20%)	250,000 (25%)
ARPU (blended)	R70/mo	R75/mo	R80/mo	R85/mo
Total Revenue	R2.6M/yr	R24.2M/yr	R116.8M/yr	R405M/yr

5. Market Tailwinds and Risks

Tailwinds

Online grocery growing 25-35% CAGR
Subscription conditioning (Xtra Savings Plus, Uber One, Netflix)
Smart speaker growth in SA homes
No local competitor combines inventory tracking with delivery integration
Food waste awareness is growing

Headwinds

Retailers could build similar features into their own apps
Price sensitivity and cost-of-living pressures
Data costs for always-on app usage
No public APIs from SA retailers
Behavioural change required for habit formation

Bottom line: The SA market presents a R80B+ online grocery opportunity with no local player solving the household inventory problem. Pantry's target segment represents approximately 2.8 million households spending R250B+ annually on groceries. The timing is right: online grocery adoption is accelerating, subscription models are accepted, and voice-first interfaces are maturing.

Feature Prioritisation Matrix

Every feature scored on four dimensions: user impact, development effort, technical risk, and SA market fit. Features are grouped into MVP, Phase 2, and Future.

MVP: ship first Phase 2: next quarter Future: later or experimental

Feature	Impact	Effort	Risk	SA Fit	Score	Phase
Voice-first grocery list Add, remove, check items via Google Nest or phone mic	5	3	2	4	4.0	MVP
Receipt scanning (OCR) Snap till slips from Woolies, Checkers, Pick n Pay	5	3	3	5	4.0	MVP
Manual pantry management Search, tap-to-add, quantity adjust	4	2	1	5	4.0	MVP
WhatsApp bot "What do I need?" / "Add milk" via WhatsApp	4	3	2	5	3.8	MVP
Budget tracking (basic) Monthly spend from receipts, category breakdown, ZAR	4	2	1	5	3.8	MVP
Family/multi-user management Shared household list, role-based access	4	2	1	4	3.8	MVP
Delivery history integration Pull order history from Woolies Dash, Checkers Sixty60	4	3	4	5	3.3	Phase 2
Predictive consumption ML-based "you'll run out of eggs by Thursday"	5	4	3	4	3.3	Phase 2
Smart scale / mat integration Weight-based pantry tracking via Bluetooth scales	5	4	4	3	3.0	Phase 2
Auto-ordering from retailers One-tap reorder via Sixty60 or Woolies Dash	4	4	5	5	3.0	Phase 2
Camera-based tracking Fridge/pantry cameras via computer vision	5	5	5	2	2.3	Future
NFC/RFID tagging Tap-to-track individual items	3	4	4	2	2.0	Future

Scoring methodology: Priority Score = (User Impact + SA Fit) / 2, penalised by high effort and risk. Formula: ((Impact + SA Fit) / 2) - ((Effort - 1) * 0.15) - ((Risk - 1) * 0.2). This deliberately biases toward features that solve real problems for SA families.

MVP Rationale

MVP: Data-Entry Friction

The MVP focuses on reducing data-entry friction through three complementary input channels: voice (fastest for known items), receipt scanning (bulk capture after shopping), and WhatsApp (accessible anywhere, no app install).

Phase 2: Intelligence

Phase 2 adds intelligence and automation. Predictive consumption needs purchase history, which the MVP builds over time. Auto-ordering depends on retailer partnerships.

Future: Camera Tracking

Camera-based tracking is the dream, but: high hardware cost, privacy concerns, and computer vision models still struggle with SA-specific products.

Passive Tracking Technology Review

The core challenge: how does Pantry know what is in your kitchen without you manually logging every item? We evaluated nine technology approaches against feasibility, cost for SA households, privacy implications, and readiness for the South African market.

Technology	Feasibility	Cost (SA)	Privacy	SA Readiness	Overall	Verdict
Receipt OCR	5/5	5/5	5/5	5/5	5.0	MVP
WhatsApp Business API	4/5	4/5	4/5	5/5	4.3	MVP
Google Home APIs	4/5	4/5	4/5	3/5	3.8	MVP
Predictive ML	3/5	4/5	5/5	3/5	3.5	Phase 2
Email/Order Parsing	4/5	5/5	3/5	4/5	3.5	Phase 2
Smart Scales	3/5	2/5	5/5	2/5	3.0	Phase 2
Weight Sensors	2/5	1/5	5/5	1/5	2.0	Future
Camera / CV	2/5	1/5	1/5	1/5	1.3	Future
NFC / RFID	2/5	2/5	5/5	1/5	2.0	Future

Receipt OCR (MVP Priority)

▼

How it works: User photographs their till slip. OCR extracts store name, item list, quantities, prices, and date. Items matched against a product database.

SA-Specific Considerations

Woolworths receipts are well-structured, clean thermal print. Easiest to parse.
Checkers/Shoprite use abbreviated product names (e.g., "CHK HM MLK 2L"). Requires SA-specific abbreviation dictionary.
Pick n Pay receipts vary by store generation. Newer stores have QR codes linking to digital receipts.

Tech: Google Cloud Vision API or Azure Computer Vision. Estimated cost: ~R0.03 per receipt.

WhatsApp Business API (MVP Priority)

▼

Why non-negotiable for SA: 96% of SA smartphone users are on WhatsApp. Minimal data usage (~1KB per text). Domestic workers and extended family far more likely to use WhatsApp than download a new app.

Tech: WhatsApp Business Cloud API (Meta-hosted). First 1,000 service conversations free per month, then ~R0.50 per conversation.

Google Home APIs (MVP Priority)

▼

SA context: Google Nest devices available via Takealot and Incredible Connection. Nest Mini ~R900, Nest Hub ~R2,500. Google Assistant supports SA English accent recognition. Competitor Alexa has almost no SA presence.

Tech: Google Home APIs platform with fulfillment via webhook. For Nest Hub visuals, use Interactive Canvas.

Risk: Google has a history of deprecating platforms (Conversational Actions sunset 2023). Build abstractions.

Predictive Consumption ML (Phase 2)

▼

Approach: Moving average of purchase intervals per item. SA-specific patterns: braai season (Sep-Mar), back-to-school (Jan), load shedding periods. Minimum 3 purchase cycles per item before predictions activate.

Why Phase 2: The MVP must build the data foundation first. Predictions only become valuable after 8-12 weeks of purchase data.

Recommended Technology Stack

MVP Stack

Google Cloud Vision (OCR), WhatsApp Business Cloud API, Google Home APIs, Firebase/Supabase (backend), React Native or Flutter (mobile)

Phase 2 Additions

Python ML pipeline (scikit-learn), Gmail API (email parsing), BLE SDK (smart scale integration), custom product matching model

Future Exploration

Edge AI / TensorFlow Lite (camera), custom hardware (smart shelf)

Load Shedding: The SA-Specific Constraint. Every technology choice must account for intermittent power and connectivity. Architecture must support: offline-first mobile app, WhatsApp's store-and-forward model, edge processing for receipt OCR (on-device fallback), and graceful degradation for IoT devices.

Integration Architecture

Pantry connects multiple input surfaces (voice, camera, WhatsApp, manual) to a unified pantry state, then pushes that state out to multiple output surfaces. This architecture prioritises offline resilience, SA retailer compatibility, and low-cost scaling.

System Architecture Overview

Layer	Components
Input Channels	Google Nest Voice, WhatsApp Bot, Receipt Scanner, Mobile App, Smart Scale
API Gateway	Pantry API Gateway (Auth, Rate Limit, Routing) via REST + WebSocket
Core Services	Pantry State (CRUD + sync), NLP Engine, OCR Pipeline, User/Household management, Prediction engine
Data Layer	PostgreSQL (items, users, history), Redis (cache, sync), Object Storage (receipts, photos), SA Product DB
External	Woolworths, Checkers, Pick n Pay (web scraping/future API), Google (Home APIs, Cloud Vision), Meta (WhatsApp API)

Retailer Integration (Phased)

Phase 1: Read-Only (MVP)

Receipt Photo → OCR Engine → Product Matcher → Pantry DB. No retailer integration needed: we read receipts, not APIs. Build SA abbreviation dictionary from 500+ receipts.

Phase 2: Email Parsing

Sixty60/Dash/ASAP emails → Email Parser → Item Extractor → Pantry DB. More reliable than OCR (structured HTML vs. faded thermal paper).

Phase 3: Direct (Partnership)

Pantry App → Pantry API → Retailer API → Delivery. Direct ordering, real-time stock. Revenue share 2-5% on orders. Viable at 50,000+ active households.

Cross-Surface Data Flow

Surface	Sync Method	Latency	Offline Support
Mobile App	WebSocket + local SQLite	< 500ms	Full offline queue, sync on reconnect
Web Dashboard	WebSocket	< 500ms	Read-only cache
Google Nest	Webhook (server push)	1-3 seconds	Nest handles connectivity internally
WhatsApp	Webhook (event-driven)	1-5 seconds	WhatsApp queues messages during offline
Smart Scale	BLE to app, then API	5-30 seconds	Scale buffers readings locally

Infrastructure

Recommended: Google Cloud, Johannesburg region (africa-south1). <20ms latency for SA users, native integration with Cloud Vision and Google Home APIs.

Component	Service	Monthly Cost (ZAR)
API Server	Cloud Run (auto-scaling)	R500 - R2,000
Database	Cloud SQL (PostgreSQL)	R1,200 - R3,500
Cache	Memorystore (Redis)	R800 - R1,500
OCR API	Cloud Vision	R300 - R1,000
WhatsApp API	Meta Cloud API	R0 - R2,000
Total (1,000 households)		R3,000 - R10,500/month

Multi-Surface Strategy

Pantry lives across five surfaces, each optimised for a different moment and context. The key insight: no single surface handles every grocery management scenario. SA households are multi-device, multi-person, and multi-context.

📱 Mobile App

Primary surface. Receipt scanning, barcode scanning, full pantry management, push notifications, offline mode, budget reports, smart scale pairing.

When: Shopping, scanning receipts, on the go

🏠 Google Nest Hub

Kitchen companion. Hands-free voice, ambient display, visual pantry summary, proactive reminders, multi-user voice recognition.

When: Cooking, hands dirty, quick checks

💬 WhatsApp Bot

Universal access layer. Zero app install, zero-rated data, receipt photo processing, voice note transcription, group chat integration.

When: Anywhere, anytime, domestic workers, extended family

💻 Web Dashboard

Planning and analytics. Full spending analytics, price trend charts, store comparison, bulk editing, household settings, data export.

When: Weekly planning, budget review

Sync Matrix

Capability	Mobile	Nest Hub	WhatsApp	Web
Pantry inventory	✅	✅	✅	✅
Shopping list	✅	✅	✅	✅
Receipt scanning	✅	❌	✅	❌
Voice commands	✅	✅	Voice notes	❌
Budget/spending	✅	❌	Summary only	✅
Push notifications	✅	✅	✅	❌
Offline access	✅	❌	Partial	❌

MVP Surface Prioritisation

Priority	Surface	Rationale	Timeline
P0	Mobile App	Primary input surface. Receipt scanning, full management.	MVP launch
P0	WhatsApp Bot	Widest reach in SA. No install needed. Zero-rated data.	MVP launch
P1	Google Nest Hub	Voice-first kitchen experience. Differentiator.	MVP launch
P2	Web Dashboard	Analytics and planning. Less urgent for MVP.	MVP + 4 weeks

Design Principles

Surface-Native, Not Surface-Cloned. Each surface feels native to its platform. The Nest Hub is not a tiny version of the mobile app.

Lowest Common Denominator is WhatsApp. Every core action (add, check, get list) must work via WhatsApp text message.

Notification Routing. One notification, routed to the right surface. "Low stock" to WhatsApp. "Budget exceeded" to mobile push. "Morning summary" to Nest Hub.

Business Model Canvas

Pantry's business model is designed for the South African market, where grocery spending is a major household expense (R4,000-R8,000/month for LSM 7-10) and saving money on groceries is a universal priority.

Value Proposition

Never run out of essentials. Know what you have, what you need, before you need it.

Save money on groceries. Track spending, compare stores, avoid waste and duplicate purchases.

Effortless tracking. Scan receipts, use voice, send a WhatsApp. No manual logging.

Whole household, one system. Parents, domestic workers, family members all contributing.

Subscription Pricing (ZAR)

Pantry Free

R0/month

Manual pantry management
Shopping list (1 list)
WhatsApp bot (basic)
5 receipt scans per month
1 household member

Pantry Plus Most Popular

R49/month

Everything in Free
Unlimited receipt scans
Up to 4 household members
Google Nest Hub integration
Budget tracking and reports
WhatsApp bot (full features)

Pantry Premium

R99/month

Everything in Plus
Unlimited household members
Predictive restocking
One-tap ordering (Sixty60, Dash)
Store price comparison
Smart scale integration
Advanced analytics and insights

Pricing rationale: At R49/month, Pantry Plus costs less than a single Sixty60 delivery fee (R35-R50) and is positioned as a "save more than you pay" proposition. If Pantry saves even R200/month on wasted food and duplicate purchases, the ROI is clear.

Revenue Streams

Stream	Description
Subscriptions (primary)	Free / R49 Plus / R99 Premium tiers
Retailer commissions	2-5% on orders placed via Pantry
Promoted products	Retailers pay for featured placement
Data insights	Anonymised consumption trends sold to FMCG brands
Hardware margin	Branded smart scales at retail markup

Unit Economics

Metric	Year 1	Year 2	Year 3
Households	5,000	25,000	100,000
Paid subscribers	400 (8%)	3,000 (12%)	15,000 (15%)
Monthly revenue	R27,000	R266,000	R1,520,000
Infrastructure cost	R15,000/mo	R60,000/mo	R200,000/mo
Team cost	R350,000/mo	R600,000/mo	R1,200,000/mo
Monthly burn	R338,000	R394,000	Breakeven+

Competitive Moat

SA Product Database. Comprehensive database mapped across retailers with abbreviation dictionaries and price history. Takes months to build, hard to replicate.

Multi-Surface Lock-in. Once a household uses Pantry across mobile + WhatsApp + Nest Hub, switching costs are high. The domestic worker knows the bot. The purchase history is accumulated.

Purchase History Data. After 6 months, Pantry knows a household's complete grocery profile. The longer you use it, the more valuable it becomes.

Retailer Partnerships. Exclusive integration with Woolies Dash and Sixty60 creates a two-sided marketplace effect.

Key Risks

Risk	Severity	Mitigation
Retailers build their own	High	Pantry is retailer-agnostic. Cross-store value is our advantage.
Low conversion to paid	Medium	Commission revenue provides secondary income. Adjust free tier limits.
WhatsApp API changes	Medium	Build abstraction layer. Telegram as backup. Mobile app remains core.
POPIA compliance	Medium	Privacy-by-design. Explicit consent. Data residency in SA (GCP africa-south1).