Olivers
E-commerceA florist's online shop, rebuilt off a rented Shopify platform and onto infrastructure Olivers owns outright.

Olivers sells online and takes bookings for weddings and workshops — two very different jobs that a general-purpose platform like Shopify was never built to do together.
Running it all through rented software meant every change to the range, the delivery rules or the booking pages went through a template not built for a florist.
Same-day delivery around Newport is a real part of what Olivers sells, and it is not a shipping option a general template understands. It is a postcode question with a cut-off attached, and the customer needs it answered on the product page rather than discovered at the checkout.
Weddings and workshops are not accessories to a bouquet order either. They arrive as consultations and ticketed sessions carrying dates, capacity and deposits, and they need their own records rather than a bouquet's with the fields renamed.
A florist's catalogue also moves. Prices follow the season and stems come in and out of range, so whatever a change cost to make was going to be paid every week rather than once.
A ground-up store rebuild — catalogue, bookings, weddings and workshops — on infrastructure Olivers owns outright instead of renting.
Delivery runs on three named rails rather than one shipping table: same-day inside the Newport postcodes, next-day across South Wales, next-day to the rest of the UK, each with its own cut-off and its own fee. A postcode checker sits on the product page and in the cart, so the answer arrives while the customer is still choosing.
The checkout treats the browser as something to be checked rather than believed. The server prices every line from its own table and ignores whatever price the page sent, recomputes the delivery fee from the postcode, recalculates the total and mints the order reference itself. A total that disagrees with the server's by more than a penny is refused, and the security rules let nothing but the server create an order record in the first place.
Two things were deliberately not built in. MAST did not bolt a card processor onto the shop at the start: Olivers already takes card across the counter, so the online rails are the ones the business was actually using, with a card processor wired in behind a switch to be turned on when card-not-present demand paid for it. The same reasoning ruled out the obvious subscription add-on, which would have tied the shop straight back to the platform it was leaving. Recurring orders were left as a contact-to-arrange page with the online wiring deliberately unbuilt, because at launch the business arranges them in person and paying for rails nobody uses is the mistake the whole rebuild was avoiding.
An AI-assisted operations layer drives product and content changes end to end, so the shop can be updated without a developer on hand.
That layer works to named routines rather than to open-ended instructions. A new bouquet, a price change, a retirement, a product going live, a workshop launch and a marketing pack each have one, and running any of them fans the work out across everything it touches — the catalogue, the photography, the printed pack, the marketing and the live site — because adding a product rarely ends at the product. Each run opens a ledger of what it still owes, the ledger survives a crash, and it does not close until the checks pass.
The checks are the part that makes the rest safe to run. They compare the shop's prices against the server's, the quoted delivery fee against the charged one, whether every product carries an image, whether the live product page actually renders and whether the checkout completes. Alongside them sits a fixed list of surfaces automation does not touch at all: payments, sign-in, the security rules, the legal pages, the checkout and the canonical catalogue. Those move on a person's say-so.
What it actually looks like, running.


A small team runs the whole shop — catalogue, bookings and content — without a developer on hand, and can take on more without buying new software.
The cost of a change came down, which matters more here than any single feature: a florist changes prices and range with the season, and the routines carry each change through the photography, the print packs, the marketing and the live site rather than leaving that trail to somebody's memory.
The shop also sits on ground Olivers controls. The storefront is a static export served from a CDN with the dynamic work on functions, and the customer data lives in a UK region — chosen for residency as much as for the shorter round trip.
Half an hour to see what an owned storefront looks like for your business.
Start a conversation