Concept · B2B e-commerce · Phone first, plus web · 2026

Morning Crate: the weekly order,done on the phone before the rush

A reorder app for small cafés: a suggested order, one cart across suppliers, price warnings and delivery check in. A web portal covers the books.

What I’d measurePhone orders, edits before placing, credits recovered.
A short loop of the app in use. Select it to try the real thing. Sample data.

The story in six lines

  1. GoalHelp small café owners place the weekly order without losing a morning to it.
  2. ProblemFive suppliers, five ways to order. Oat milk rose 9% in twelve weeks, unnoticed., go to this part
  3. ApproachMove the order into the gaps of the day: the phone does the job, the web keeps the books., go to this part
  4. DecisionSuggest the order, never place it. Owners know things the numbers don’t., go to this part
  5. LaunchNow a free app for cafés, with no account and no server., go to this part
  6. ResultWhat I’ll measure: orders placed from the phone, and credits recovered., go to this part

Select a line to read that part.

At a glance

Role
Solo: designed and built it
When
A 2026 concept
Platforms
Phone first, plus a web portal
For
Owners of small cafés and shops

Accessibility: Contrast checked, counts read aloud, demos still for reduced motion.

01 · The problem

Small shops buy like big ones, without the tools

A café owner orders from five suppliers every week, each with its own channel, minimum and cutoff.

Most of it happens standing up, so the phone does the job and the web keeps the books.

  • 5

    Five suppliers, five ways to order

    Texts, emails and a PDF form.

  • 9%

    Prices that creep

    Oat milk rose 9% in twelve weeks, unnoticed.

  • $15

    Minimums and cutoffs

    Miss a minimum and delivery costs $15 more.

  • 7 am

    Short at the door

    A short crate becomes a forgotten credit.

The week, before and after

Move the order into the gaps of the day.

The suppliers can’t change. Where the order happens can.

Before · six steps, three channels

  1. Text the dairy
  2. Email the bakery
  3. Fill in a PDF for paper goods
  4. Add up the totals by hand
  5. Wait at the door
  6. Chase the credit on Friday

After · four steps, one app

  1. Open the suggested order
  2. One cart, split per supplier
  3. Tick it in at the door
  4. Credit requested on the spot

Decisions behind it

Four calls that shaped it

  1. The call I made

    Suggest the order, never place it

    Owners know things the numbers don’t, like a Saturday catering job, so the app suggests and the owner decides.

    What it cost: A minute of review a week.

  2. What I dropped

    Paid lines in the suggested order

    Sponsored items stay in the catalog, always labeled. One paid line in the suggestion would break trust.

  3. The constraint

    Short at 7 am

    Check in works for whoever opens, with the credit requested on the spot.

  4. The call I made

    Dark for the night before

    The order goes out the night before, often from a dim back room, so the app follows the phone’s dark setting.

    What it cost: Every color picked twice.

02 · The flows

Every step, and every way off it

The two hardest jobs, screen by screen: the main path on top, the ways off it below.

Drag to move around. Zoom with the buttons, a pinch, or Ctrl and scroll.

From the suggestion to five supplier orders, with a cheaper supplier found on the way.

100%
The steps as text
  1. Today. Friday’s order comes first: how many items are suggested and what they cost. Most suppliers close tonight.
  2. The suggested order. Each line is the gap to par, with the reason in words. Add all, add one, or open a line to see why.
  3. Compare prices. Oat milk is up 9% in 12 weeks. The same case is cheaper at Greenbank: the saving is shown at this café’s usage, not as a list price.
  4. Switched. One tap. It starts with Friday’s order, and the Spend tab counts the saving from here on.
  5. One cart. Split by supplier, each with its cutoff, minimum and day. A quiet afternoon slot is cheaper. The order’s points show before it’s placed.
  6. Placed. One order per supplier, each on its own day. Points land at once, double this month, and the web portal shows the orders straight away.
  7. Orders. Every order from placed to delivered. Until a supplier’s cutoff, it can still change.

Off the main path

  • Something else. What the suggestion missed is a search or a category away. A brand can pay to sit at the top, always marked Sponsored, never in the suggested order.
  • Under a minimum. The cart says how much more gets free delivery, or what the fee is, and offers the one item that gets you over.

Every phone on the map is the working app, held at that step. Sample data.

03 · Small moments

Feedback you can read with one hand

Between customers, feedback has to be quick and plain. Each clip is the working app.

Kettle & Wren

Good morning, Dana

Thursday · No more deliveries today

Friday’s order · most suppliers close tonight

10 items suggested · $1,140.20

Built from what’s on the shelf and how fast it goes.

Rewards · 1,860 points

2× until Nov 30

140 more points and you get $20 off. Every dollar earns 2 points this month.

Running low

Swipe left to add

  • Whole milk3 of 10 on the shelf
  • Oat milk1 of 5 on the shelf
  • Large eggs0 of 2 on the shelf

Swipe to add

One swipe adds a low line to Friday’s order.

Playing on a loop. Tap the phone to take over.

Kettle & Wren

Friday’s order

Suggested from what’s on the shelf and how fast it goes. Change anything.

10 suggested

  • 3 on the shelf7 short of par

  • 1 on the shelf4 short of par

  • 0 on the shelf2 short of par

  • 2 on the shelf5 short of par

  • 1 on the shelf1 short of par

  • 1 on the shelf3 short of par

  • 1 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 1 on the shelf1 short of par

Fill to par

The shelf bar fills toward par.

Playing on a loop. Tap the phone to take over.

Cart

One supplier so far. Add from others and the cart splits itself.

  • Greenbank ProduceOrder by Thu 8 pm · 8 to 10 am

    Add $42 more for free delivery, or pay $10

    • Lemons1 × $38 = $381
    Avocados1 short of par · $56
Items
$38
Delivery
$10
Total
$48

This order earns 76 points, 2× until November 30.

Three ways to pay less

The cart offers the one item that clears a minimum.

Playing on a loop. Tap the phone to take over.

Check in

Northfield Dairy

MC-4471. Tick each line as you unpack it. Anything short or damaged becomes a credit request.

0 of 3 checked

  • Whole milk8 × crate of 6 gallons
  • Oat milk4 × case of 12 quarts
  • Large eggs2 × case of 15 dozen

Checked off at the door

Lines tick off; a short line counts up its credit.

Playing on a loop. Tap the phone to take over.

Up close, slowed down

Four more, close up and slowed down.

Kettle & Wren

Friday’s order

Suggested from what’s on the shelf and how fast it goes. Change anything.

10 suggested

  • 3 on the shelf7 short of par

  • 1 on the shelf4 short of par

  • 0 on the shelf2 short of par

  • 2 on the shelf5 short of par

  • 1 on the shelf1 short of par

  • 1 on the shelf3 short of par

  • 1 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 1 on the shelf1 short of par

Into the crate

Each item drops into the crate.

Playing on a loop, slowed down. Tap the phone to take over.

Kettle & Wren

Friday’s order

Suggested from what’s on the shelf and how fast it goes. Change anything.

10 suggested

  • 3

    3 on the shelf + 3 coming4 short of par

  • 1 on the shelf4 short of par

  • 0 on the shelf2 short of par

  • 2 on the shelf5 short of par

  • 1 on the shelf1 short of par

  • 1 on the shelf3 short of par

  • 1 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 1 on the shelf1 short of par

Hold and slide

Hold and slide: ten packs in one move.

Playing on a loop, slowed down. Tap the phone to take over.

Dairy

Oat milk

Case of 12 quarts · Northfield Dairy

On the shelf
1 of 5
Uses a week
4
Per pack
$34.20

1 on the shelf + 4 comingAt par

Price per pack, 12 weeks

Your supplier against the same pack elsewhere.

Northfield DairyGreenbank Produce
$30$33$35Aug 25Sep 22Oct 20Nov 10$34.20 Northfield$31.40 Greenbank

Greenbank Produce has it for $31.40

At 4 a week, switching saves $11.20 a week. Same pack size, delivered the same days.

4

The price rolls down

Switch supplier and the saving counts up.

Playing on a loop, slowed down. Tap the phone to take over.

Check in

Northfield Dairy

MC-4471. Tick each line as you unpack it. Anything short or damaged becomes a credit request.

0 of 3 checked

  • Whole milk8 × crate of 6 gallons
  • Oat milk4 × case of 12 quarts
  • Large eggs2 × case of 15 dozen

Stamped at the door

Arrived lines get a stamp; short ones a torn tag.

Playing on a loop, slowed down. Tap the phone to take over.

04 · The order

Start from the shelf, then pay less for it

The order suggests the gap to par, with a reason like “sold out early twice last week”.

A creeping price gets caught and switched in one tap. Quiet slots and points cut the rest.

Rewards, quiet slots and sponsored listings are adapted from consumer food delivery.

The working app. Try it.

Kettle & Wren

Friday’s order

Suggested from what’s on the shelf and how fast it goes. Change anything.

10 suggested

  • 3 on the shelf7 short of par

  • 1 on the shelf4 short of par

  • 0 on the shelf2 short of par

  • 2 on the shelf5 short of par

  • 1 on the shelf1 short of par

  • 1 on the shelf3 short of par

  • 1 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 0 on the shelf1 short of par

  • 1 on the shelf1 short of par

Live: add a line and watch its bar fill to the par tick.

05 · At the door

Check it in before the first customer

Whoever opens ticks what came and flags what’s short.

Anything missing becomes a credit request on the spot.

The working app. Try it.

Check in

Northfield Dairy

MC-4471. Tick each line as you unpack it. Anything short or damaged becomes a credit request.

0 of 3 checked

  • Whole milk8 × crate of 6 gallons
  • Oat milk4 × case of 12 quarts
  • Large eggs2 × case of 15 dozen
Live: tick the milk, mark the oat milk short, and request the credit.

06 · Color palette

Two palettes, and one job for each color

I tried two palettes on the same screens. Digital Peach paired a loud peach for actions with deep indigo and cool lavender. The café palette uses a fresh green, light sage, cream and coral.

Peach was bold, but it blurred meaning. The action, the low-stock warning and a chart line were all warm oranges, so “tap here” and “something’s wrong” looked alike. White text on the peach is only 2.5:1, so every button needed dark ink.

The café palette gives each color one job. Green is the only action, coral marks progress and what needs attention, and the charts get their own blue and amber. White on the green passes AA at 4.8:1, and the sage and cream feel like a café counter rather than a software dashboard.

Considered

Digital Peach

  • Peach #FF7E5FActions, with dark text (white is 2.5:1)
  • Indigo #3B3B98The hero card and web sidebar
  • Lavender #F6F6FBThe ground
  • Coral #E8684ALow stock
  • Blue and orange #2A78D6 · #EB6834Charts

Chosen

Café

  • Green #278250The one action color, with white text (4.8:1)
  • Sage #D4ECD9Color blocks and the web sidebar
  • Near white #FAFAF8The ground, with the faintest warmth
  • Coral #E8775AProgress, and what needs attention
  • Sun #F4C24FA small sun in the color blocks
  • Blue and ochre #2A74C9 · #C47A12Charts, kept apart from green and coral
Each palette with the job every color did, and the same button and low-stock tag in both.
Today, the order and Spend in both palettes. Digital Peach: an indigo hero card, peach buttons and blue bars. Café: a light sage block with soft waves, green buttons, coral progress bars and thin blue bars.
The same screens and data in both palettes: Today, the order and Spend.

07 · Light and dark

Made for the night before

The order goes out the night before delivery. The count often happens after close or before opening, in a dim back room.

So the app follows the phone’s dark setting. Green stays the one action, a mint on deep navy: the hour before sunrise.

Every color was picked twice. Text passes WCAG AA in both themes, and the chart colors were checked for color blindness.

The same Today screen in light and dark. Sample data.

08 · Between the main screens

The states nobody screenshots

The states that decide whether an app feels finished.

  1. 01First launch

    Suppliers found from past invoices. Check, don’t type.

  2. 02Loading

    The layout holds while prices load.

  3. 03An empty cart

    Points back to the suggested order.

  4. 04A supplier is down

    One supplier down doesn’t hold up the rest.

  5. 05The order goes through

    Each supplier’s day and window, in one place.

  6. 06On the lock screen

    A live tracker counts down to the door.

Every phone is the working app, held at that state. Sample data.

09 · Also on the web

The same café on a laptop, for the books

On Sunday night, the books get a bigger screen. The portal and phone share one café.

Morning CrateKettle & Wren

Dana, owner · next delivery day Friday

Reorder

Good morning, Dana

Friday’s order: 10 items suggested, about $1,151, from what’s on the shelf and how fast it goes.

ItemOn the shelfWhyPriceOrder
3 of 10 · uses 9 a weekWeekend usage is up 18%$28.80+5%
1 of 5 · uses 4 a weekPrice up 9% in 12 weeks$34.20+9%
0 of 2 · uses 1.5 a weekOut, and brunch is Saturday$42.50-8%
2 of 7 · uses 6 a weekCovers the week to Friday$62steady
1 of 2 · uses 1 a weekBack to par$28steady
1 of 4 · uses 3 a weekSold out early twice last week$54steady
1 of 2 · uses 2 a weekBack to par$46steady
0 of 1 · uses 1 a weekOut since Wednesday$38+15%
0 of 1 · uses 1 a weekPrice down 12%: a good week to stock$56-12%
1 of 2 · uses 1 a weekOne case left$88steady

Price watch

Arriving today

  • Northfield Dairy7 to 9 am · 3 itemsOn its way
  • Old Mill Bakery Supply6 to 8 am · 2 itemsOn its way
The whole portal works: reorder, catalog, prices, orders and spend. Sample data.

10 · Built in code

How the prototype works, and why I built it this way

Ordering is mostly rules, and rules are easier to judge when the total moves.

No outside design system: Morning Crate has a small one of its own. Color tokens set a café palette for light and dark: green for actions, sage and cream for the surfaces, coral for what needs attention, and blue and amber for the charts. Fraunces sets the headings and Plus Jakarta Sans the text, and every control is built here. The phone screens run in the same iOS shell as ontheGo, so the status bar, alerts and screen transitions behave like an iPhone. Charts use a palette checked for color blindness.

codebase for the phone app and the web portal
1codebase for the phone app and the web portal
files, about 2,800 lines of TypeScript
11files, about 2,800 lines of TypeScript
unit tests on the ordering rules
9unit tests on the ordering rules
UI libraries: every control built here
0UI libraries: every control built here

How it fits together

components/crate/model.ts · shortened
// One supplier’s share of the cart
const sub = sum(lines);
const slot = S.quiet && st.slots[sp] === "quiet" ? "quiet" : "usual";
const off = slot === "quiet" && S.quiet ? r2(sub * S.quiet.off) : 0;

// Under the minimum, the delivery fee applies
return { supplier: sp, lines, sub, off, slot,
  fee: sub < S.min ? S.fee : 0,
  short: Math.max(0, r2(S.min - sub)),
  // …the delivery day
};

store.tsx

One café, one order

One state for every screen.

  • Phone appFor whoever opens the café
  • Web portalFor the owner, on Sunday night
  • Flow diagramsPhones seeded by the same steps

Why this stack

  • next 16 · react 19

    Next.js and React

    Hands over as code.

  • strict

    TypeScript

    A wrong field fails the build.

  • tailwind 4

    Tailwind CSS

    Tokens for light and dark.

  • motion 12

    Motion

    Off for reduced motion.

Three changes I’d walk through

  1. 01model.ts

    One café, two screens

    Problem. Phone and web kept separate copies that drifted.

    What I changed. One store and small functions, shared with the flow diagrams.

  2. 021306px → 375px

    A desktop portal on a phone

    Problem. On phones, the portal pushed the page to 1306 pixels.

    What I changed. One grid rule fixed it, found by auditing every page at 375 and 320 pixels.

  3. 03useAutoDemo

    Demos that use the real app

    Problem. Recordings go stale; fake animations can lie.

    What I changed. A script presses the real buttons, and stops when a visitor touches the app.

The rules, checked in code

A bug here costs a café money, so the rules have tests.

Terminal

$ npm test

  • ✔ an order under the supplier minimum pays the delivery fee
  • ✔ reaching the minimum drops the fee
  • ✔ the quiet slot discount applies to that supplier only
  • ✔ a reward applies only once there are enough points
  • ✔ totals add up to $82.21
  • ✔ one cart becomes one order per supplier
  • ✔ placing an empty cart changes nothing
  • ✔ switching supplier and back restores the price
  • ✔ a short item at the door requests the right credit

tests 9 · pass 9 · fail 0

Working with AI

Claude drafted

  • First drafts of components
  • The cart math
  • The tests and the audit script

I decided

  • Read every change before it went in
  • Took out the dim overlay behind the item panel, so the cart stays in view
  • Slowed every loop after watching them
  • Chased down the phone width bug

Types, lint, tests and the phone audit run on every change.

What changes before it ships

  • Real data

    Prices from a feed, orders saved per account.

  • Notifications

    Push and email.

  • A native phone app

    model.ts carries over to React Native as it is; only the screens are rebuilt.

11 · The test I’d run first

Should the order lead with what’s low, or with what got expensive?

The shelf is what owners expect; prices are where the money is. I’d test both.

Where the test sits in the funnel

  1. Step 1Open Today
  2. Step 2Open the order
  3. Step 3Review the lines
  4. Step 4Place before the cutoff

The change is on step 3; step 4 shows it.

A · What’s low firstToday’s design

Fast and familiar: the gaps on the shelf, each with its reason. Price alerts wait below.

B · What crept firstThe challenger

Money first: items whose price rose, with a cheaper supplier one tap away. The shelf gaps follow.

Hypothesis
Leading with crept prices saves money without slowing the order.
Who sees what
Split by café, not by visit.
Main number
Spend per order on the same items.
Guardrails
Time to order, missed cutoffs, shelf gaps.
How long
Six weekly order cycles, so each café places enough orders to compare.
Decision rule
Ship B only if spend drops and no guardrail slips.

Next

What I’d want to learn next, and how I’d measure it

  • Do owners trust the suggested order?

    How often they edit it.

  • Does one cart save time?

    Time to place, against texts and emails.

  • Who checks deliveries in?

    It must work for a new barista.

The numbers I’d track after launch

  • Orders placed from the phone

    Most orders should start and finish there.

  • Lines changed before placing

    Fewer edits means more trust.

  • Credits recovered

    Requested at the door against credited.

What I’d do differently

Sit through a real Sunday order before drawing a screen.

Morning Crate is a concept. Every café, supplier, product and price is made up for illustration.