A brand, a store and a lifecycle programme. Built by prompting Claude.
ASHGROVE started as a one-page brief. It grew into a website, a design system, a shop that talks to Klaviyo, newsletters, social and four automated customer journeys. Every part was built by describing what it should feel like and letting Claude do the building.
4chapters: brand, store, campaigns, lifecycle
87demo shoppers sorted into five lifecycle stages
4automated flows, built and tested as drafts
0emails sent to real customers
ASHGROVE is a fictional brand. The shoppers, orders and prices are invented. Test emails went only to the author's own inboxes, and no flow was ever switched on. This page shows the method, not sales results.
01
The brand
One brief became a coded website, then a design system that every later piece reused. Because the rules lived in one place, a shop, an email and a reel all look like the same company.
Website
A written brief became a coded HTML and CSS site, then was redesigned in place.
A twelve-piece demo shop, so the marketing has something real to point at: three tees, three knits, three coats and three accessories, each with a flat-lay and a worn-on-model photograph.
A shop that tells Klaviyo what shoppers do
The shop carries Klaviyo's onsite tracking. When a shopper views a product, adds to the cart or opens the cart to check out, Klaviyo records the event on their profile. Those events are what start the Abandoned cart journey in chapter 04.
A real store has strangers walking in, and a demo that collects strangers' email addresses is a risk. So the public shop has no sign-up form. Visitors browse normally and nothing is stored about them.
For the demo, a hidden "Browse as" picker lets the author play one of four shoppers: Rajiv the new subscriber, Sam the cart abandoner, Leo the repeat buyer and Hugo the lapsed customer. The picker identifies a shopper by an email that already exists, with nothing to type. Anything the author does next lands on that person's profile, so a journey can be watched from end to end.
Social is the top of the funnel. Posts and ads would bring people to the shop, the shop's tracking tells Klaviyo what they did, and the lifecycle flows take over from there.
1 · InstagramConcept creative, not published
→
2 · The shopViewed Product, Added to Cart, Started Checkout
A newsletter goes to everyone at once. A lifecycle programme speaks to each shopper according to what they have done. This chapter shows how ASHGROVE's customers were sorted, which journeys they travel, and how each one was tested before anything could send.
The strategy in short: four objectives come first, each with its own journey. Welcome new subscribers, recover abandoned carts, bring first-time buyers back for a second order, and win back lapsed customers. Each wait has a reason: one hour for a cart, because the shopper is still thinking and a buyer in that hour must not get a reminder; three days between the two post-purchase emails, so the second lands after the parcel; five days between the two win-back emails, so the second tells a different story; and 90 days before the same person can re-enter post-purchase. A real programme would track first-order rate, recovered-cart rate, second-order rate within 90 days and win-back rate. None are claimed here, because nothing was sent.
1 · The shoppers
87 invented shoppers. 82 have no email address at all, so they can never be messaged. Four have the author's own inboxes, each playing one stage of the customer lifecycle. 83 backdated orders (all invented, in USD) give each shopper a believable history.
2 · Five stages, sorted by order history
Five live segments sort everyone by what they have bought and when. The numbers below are from the recount on QA day, with the original build-day number underneath.
Never ordered
29was 29
Zero orders, ever.
First-time buyers
12was 14
One order, in the last 30 days.
Repeat buyers
13was 14
Two or more orders, one in the last 60 days.
Fading
13was 13
No order in 60 days, one in the last 120.
Lapsed
18was 17
No order in 120 days, but bought once.
The five stage segments in Klaviyo, next to the demo lists.
The counts drifted, and that taught something
Two days after building the segments the total fell from 87 to 85. Nothing was deleted. The segments use time windows, so shoppers move along as days pass. First-time buyers, for example, are people who bought in the last 30 days, and two of them crossed that line.
That also revealed a gap. A shopper with one order from 30 to 60 days ago now fits no stage, because they are too old to be a first-time buyer and too recent to be fading. Two people sit in that gap today. It was left in place on purpose and written up as a finding: time-based segments drift, so they need a recount, and their edges need testing.
3 · The journey map
The five stages with the flow that speaks to each. The hatched columns are gaps: no flow today.
Never ordered
29Gap: no flow for this stage today
First-time buyers
12Post-purchase: after an order
Repeat buyers
13Post-purchase: after each new order
Fading
13Gap: no flow for this stage today
Lapsed
18Win-back: on entering the stage
Welcome (on joining the list) and Abandoned cart (on starting checkout) are triggered by events, not by stage. Never ordered, the biggest stage, and Fading have no flow today. Two flows are proposed, not built: a browse nudge for Never ordered and a "we saved your favourites" email for Fading. Full map and details → · Back to the Instagram section ↑
4 · Four journeys
Each journey starts from something a shopper does, then waits and sends. All four were built and tested, and every one stayed in Draft, the state in which nothing can send.
Welcome
Draft
Starts when someone joins the newsletter list. Three emails over five days that introduce the brand's one idea: fewer things, better made.
TriggerAdded to the newsletter list
Day 0Welcome. We make fewer things.
Wait2 days
Day 2Start with one. (personalised)
Wait3 days
Day 5The whole collection, on one page.
Rule: each person goes through once. A safety filter lets only people marked as allowed to receive demo email in.
Abandoned cart
Draft
Starts when someone opens the checkout. An hour later, anyone who has not bought gets a reminder with their own cart in it.
TriggerStarted checkout
Wait1 hour
CheckAlready bought? Then stop. Otherwise continue.
EmailYou left something behind.
Rule: the same person can enter again only after 7 days, so a browser who keeps returning is not nagged.
Post-purchase
Draft
Starts when an order is placed. A care guide first, then a suggestion of what goes with what they bought.
TriggerPlaced an order
Day 0How to make it last.
Wait3 days
Day 3Something that goes with it. (personalised)
Rule: a repeat customer can enter again after 90 days. A scarf bought after a jacket deserves the care guide again.
Win-back
Draft
Starts when a customer drifts into the Lapsed stage. Two emails, spaced out, with a different story each time.
TriggerAdded to the Lapsed stage
Day 0It has been a while.
Wait5 days
Day 5One jacket, one story.
Rule: each person goes through once.
All four flows in Klaviyo. Every status reads Draft.The Welcome journey on Klaviyo's canvas.
5 · One email, different for each shopper
The second Welcome email reads the shopper's favourite category and changes the product, the price, the photograph and the button. If the shopper has no name or no favourite, it falls back to a sensible default instead of leaving a gap.
Hugo · likes knitwear
The Ribbed Crew, $128
"Hi Hugo,"
Sam · likes outerwear
The Work Jacket, $218
"Hi Sam,"
No name, no category
The Heavy Tee, $38
"Hi there,"
6 · Tested before anything could send
Eight test sends in total: six to the author's Zoho inbox (the first three never arrived, until the sending domain was activated), then one each to Gmail and Outlook. Each journey was also previewed with the shoppers it is meant for.
Previewing the Welcome flow as a shopper off the list, then on it
Off the list: "Doesn't enter". On the list: "Enters".
Personalisation
Welcome 2 as three different shoppers
Correct product and price for each.
Missing data
A shopper with no name and no category
"Hi there" and the default product, no blanks.
Links
Every button's destination
One tagged button per email, all resolve.
Segment recount
Counts against build-day numbers
85 against 87: window drift and one gap, explained above.
Gmail's own verdict on the test email: the sending setup passes all three checks.
7 · What it taught
Passing authentication is not the same as reaching the inbox
The email passed every check and Outlook still put it in Junk. A new sending domain has no history. A real launch would start with small volumes and build a reputation first.
Segments that use time need a recount
Counts shift as days pass, and the edges between stages can leave people in no stage at all. Test the boundaries, then check them again later.
Test the full profile and the empty one
Personalisation that works for a shopper with complete data can still break for one without. Both were previewed.
Build the seatbelt in, not on
Every journey carries a filter that only lets demo-approved people in, and the audience is limited to four inboxes the author owns. A mistake could not reach a stranger.
Real: the shop, its tracking, the Klaviyo build, the flows and templates, the sending-domain authentication, and the test emails that landed in three inboxes.
Simulated: the shoppers, their orders and their prices, and the favourite category on each profile, which a real store would work out from purchases. The Instagram posts and the Reel are concept creative only: nothing was posted to a real Instagram account and no ads are running.
Not done on purpose: no flow was turned on, and no campaign was scheduled or sent. Results such as open or conversion rates are not shown because there are none to show.
