How we build
A small team, a lot of software.
Most of the day-to-day running of our brands is done by automations we built ourselves: orders, customers, marketing, monitoring and infrastructure. Here is what runs on its own, and where to read the details.
01
Automate the repetitive
If a task does not need human judgement, it becomes a system. People spend their time on product, customers and decisions.
02
Own the stack
Our stores, servers, data and workflows are ours. No single vendor can switch off the business or hold its data.
03
AI as a teammate
Models draft, check and summarise. People decide. We use different models for different jobs rather than one tool for everything.
04
Build once, use everywhere
Monitoring, email, analytics and fulfilment tooling are shared across brands, so a new brand starts with all of it on day one.
What runs on its own
The automations behind the brands.
Built and maintained in-house, shared across every brand, and watched by other automations.
- orderCurrency matches shipping country, order cleared
- whatsappOrder confirmation sent, address confirmed
- cartReminder sent for a cart left overnight
- courierInternational parcel checked: margin positive, ship
- supportOrder history attached to a new conversation
- monitorService restarted after a failed health check
Orders and delivery
Every order is sanity-checked
Each new order is compared with where it is shipping. If the currency and the country do not match, the order is flagged, the customer is asked to confirm the address, and the team is told.
Read the write-up →Ship-or-cancel on costly parcels
When the courier asks us to approve an expensive international shipment, the system works out the real margin on that order and recommends shipping or cancelling.
Address confirmed before dispatch
Customers get their order confirmation on WhatsApp with a one-tap button to confirm the delivery address, so fewer parcels come back.
Customers
Abandoned carts followed up
Shoppers who leave something behind get a timely reminder on WhatsApp or email, with the product they picked.
Read the write-up →A journey after delivery
Review requests, referral invites, suggestions and win-back messages go out on a schedule that starts from the day the parcel arrives.
Support with the full picture
When a customer writes in, the agent already sees their orders, live courier status and purchase history next to the conversation.
Marketing
From a form to a video ad
A short brief goes in; a script, visuals and an edited video ad come out, ready for review before anything is published.
Read the write-up →Server-side ad tracking
Purchases are reported to ad platforms from the server as well as the browser, so campaigns optimise on real sales.
Read the write-up →Watching the business
Errors reported as they happen
If a customer hits a broken button or a failed checkout step, the team hears about it in minutes, with the page it happened on.
A morning digest of the site
Every morning a digest lands in Slack with what people did on each page: what they searched for, what they read and where they left.
Systems that repair themselves
A watchdog on the server restarts services that fail, and a second monitor outside it raises the alarm on a phone if the server itself goes quiet.
Infrastructure
Fast pages in every currency
Store pages are cached at the edge, one copy per currency, so a shopper in London and one in Riyadh both get a fast page with the right prices.
Read the write-up →Backups without anyone remembering
Stores, databases and workflows are backed up automatically and copied off the server.
What it runs on
In practice