About this demo
GoodCane Drops is a prototype built for a CODED paid gig. It is not GoodCane's shop and nothing here is a real order.
Try as a returning customer
Demo Customer A is a made-up customer who has ordered before. Their private link opens "Same again?" in one tap.
Open as Demo Customer Ahttps://goodcane.vercel.app/c/DemoCustomerA-GoodCaneDrops-sampleLink-0001
Shared by everyone trying the demo. If someone uses "Delete my data now" or the owner's "Send new link" on this customer, the link stops working until the demo data is reset.
What is real and what is simulated
- Payments
- Simulated. You choose success, failed or no response. No money is taken and no payment provider is used.
- Messages
- Never sent by the app. Reminders wait in a queue; the owner opens a prefilled WhatsApp chat by hand.
- Data
- Made up: drinks, prices, delivery areas, customers and past orders. Anything you type here is demo data too.
- Real
- The ordering, reordering, private link, reminder consent and demand view all work, end to end, on a real database.
What we assumed
No contact has been made with GoodCane. These are stated, not checked.
To confirm with GoodCane before any pilot
- How orders arrive today (WhatsApp catalogue, calls, messages) and in what share.
- What "Delivery on the 3rd day" means and which flavours it covers.
- Whether the WhatsApp catalogue is up to date.
- Whether bottles are frozen, and how long they keep once thawed.
- Which problem matters most: waste, selling out, workload or cash tied up.
- How many bottles can be made in a day.
- Shop hours, delivery areas and fees, and other pickup options.
- How payment is taken today.
- Current prices (seen publicly: 330ml 35, 500ml 45, 2L 180, 4L 260 cedis).
- Which flavours are always on and which are seasonal.
- The exact wording and evidence for product claims.
- Shop restocks and frozen bagged cane are business sales and out of scope.
- Which payment provider GoodCane would approve for a pilot with real payments.
- GoodCane's registration with the Data Protection Commission and its record keeping.
To test in a pilot
- Most buyers are adults aged around 22 and over.
- Comments asking for location and price are orders being lost.
- Customer review videos drive discovery and trust.
- "Bought more than one" means buying again, not just a bigger order.
- Customers will order before a drop is made if they trust the process.
- Customers will opt in to reminders, and prefer WhatsApp.
- Finding GoodCane on TikTok and ordering on WhatsApp feel disconnected.
- Orders arrive in a predictable pattern before each cut-off.
- The owner will enter batch numbers consistently after each drop.
- The owner will work a short daily queue and send reminders by hand.
- Walk-in demand is steady enough to plan a buffer for.
- A rising share of pre-orders means more predictable demand, not just fewer walk-ins.
Your data
Only what is needed. No email, date of birth, precise location, advertising identifiers or trackers.
First name
Always. To hand over your order and greet you.
Phone number
Always. For your orders and, if you opt in, reminders.
Order history
Automatic. For "Same again?" and the demand view.
Reminder channel and how often
Only if you choose. To remind you before a drop.
Delivery address
Delivery orders only. To deliver.
Consent choices
Automatic. Proof of what you agreed to.
If you place no order and change no reminder setting for 6 months, your name, phone, address and links are deleted. Your past orders stay in GoodCane's numbers without your name or phone.
You can delete your details at any time from Manage ("Delete my data now"). Your order history is only ever shown through your private link or on the phone you ordered from, never because someone typed a phone number.
Data protection
This demo was designed with the principles of Ghana's Data Protection Act, 2012 (Act 843) in mind. GoodCane, as the data controller, would need to confirm its registration with the Data Protection Commission and its compliance. This is not legal advice.
About the forecast
The owner's demand forecast is simple arithmetic anyone can check, with no AI. In this demo it learns from made-up past drops, so it shows how the method works, not how accurate it would be for GoodCane. Real accuracy can only come from a pilot.