The scenario
The brand sells refillable multi-surface and laundry cleaning bundles in Kuwait, direct to consumer, on a Shopify store with KNET, Tabby and a WhatsApp ordering channel that had grown organically alongside the site. Four years in, most of its customer list still lived in the founder's personal WhatsApp, replied to individually whenever someone asked to reorder.
The product itself is the retention opportunity most brands in this category never use properly: it runs out on a schedule a household can predict almost to the week, which means the second order is not a matter of persuasion so much as timing. What the brand had never built was anything that reached out first.
- Customer list size band
- pending client sign-off
- Average order value
- pending client sign-off
- Fulfilment
- Own stock, local courier, next-day across Kuwait
- Team
- One founder, one customer-service assistant
A product that runs out on schedule, and a brand that never reached out first
When we split repeat purchase by the first product bought, the pattern was clear and completely unexploited: a specific bundle at a specific price point brought customers back at close to double the rate of the brand's actual bestseller, which was a larger one-off bundle bought mostly for a big clean rather than ongoing use. Nobody had ever looked at repeat rate by first product before, because nobody had built anything that would act on the answer.
The bigger gap was structural. WhatsApp opt-in existed only as an informal habit — customers who happened to message the founder's personal number stayed in touch, and everyone else disappeared into a Shopify customer list that was never messaged again after the order-confirmation email. There was no flow, no cadence, no segment. A customer whose bottle ran out in six weeks and a customer whose bundle would last three months received exactly the same amount of contact: none, unless they reached out themselves.
The founder had been treating every sale as a one-time transaction to be won again from scratch through paid ads, which is the most expensive way to grow a business whose own product tells you almost exactly when the next order is due. The brand did not have a loyalty problem. It had never built the channel or the flow that would have let a predictable second order actually happen.
What we did — the messages and flows
The losers are here on purpose. A test with only winners was never a test.
The order-confirmation and delivery-update flow
Winner"Your order is confirmed — here's when it lands." then a delivery-day update on the official WhatsApp number
- Format:
- Two-message WhatsApp Business API flow, triggered automatically from Shopify
This is the flow that earns the opt-in in the first place: a customer who gets a genuinely useful delivery update on WhatsApp has a reason to stay opted in for what comes after. It also moved a meaningful share of "where is my order" questions off the founder's personal number, which is what made a broader retention programme possible to run at all.
The refill check-in, timed to the product's actual consumption cycle
Winner"Most households are running low around now — need a refill?"
- Format:
- Single WhatsApp message, sent at the gateway product's average depletion window
This was the single strongest message in the flow, because it is timed to a real physical event rather than an arbitrary calendar date. Sending it a week before the average bottle empties, instead of a generic thirty-day reminder, is what made it read as helpful rather than as a sales push.
The gateway-product-specific welcome flow
Winner"You started with the everyday bundle — here's how to make it last."
- Format:
- Three-message flow specific to customers whose first purchase was the gateway product
Since the cohort read showed this specific bundle already drove the best repeat rate on its own, giving its buyers a dedicated flow rather than the generic one made the pattern stronger instead of just letting it happen by accident. It also gave the brand a specific product to feature more in its acquisition creative, which is a decision the cohort data made rather than a guess.
A generic monthly blast to the entire list
Lost"This month's offers" sent to every contact regardless of last purchase
- Format:
- Single unsegmented WhatsApp broadcast
A reasonable-sounding starting point that lost, and it lost badly: sent to customers who had bought two days earlier alongside customers who had bought four months earlier, it read as noise to both. It carried the highest opt-out rate of anything tested in the engagement, which is the cost of a blast that is not segmented by recency or last product.
The sixty-day lapsed win-back, with one specific reason
Winner"It's probably about time — same bundle, ready when you are."
- Format:
- Single WhatsApp message to the 60-day lapsed segment, named by their specific first product
This outperformed the blanket discount the brand had used before as its only win-back tactic. Naming the specific product and the specific timing gave the message a reason to exist beyond price, and it did not train the segment to wait for a markdown the way a discount-led win-back does.
What we did — the optimizations, in order
Read repeat rate by first product, not as one number
Split repeat purchase rate at thirty, sixty and ninety days by the customer's first product bought, rather than looking at one blended repeat rate for the whole store.
Why: The blended number hid a real gateway product entirely. Once split, one specific bundle stood out as bringing customers back at nearly double the rate of the brand's own bestseller, which changed what the welcome flow and the acquisition creative should both be pointing at.
Move the channel onto the official WhatsApp Business API
Set up WhatsApp opt-in at checkout and again on the order-confirmation page, on the official Business API rather than the founder's personal number, with email kept as the secondary channel.
Why: A personal number cannot carry automated flows, cannot be handed off if the founder is unavailable, and puts the entire customer relationship inside one person's phone. The official channel is what made every flow after this step technically possible.
Build the flows in the order a customer actually experiences them
Order confirmation and delivery update first, then the refill check-in timed to the product's consumption cycle, the review ask, then a replenish or cross-sell message, all written in Kuwaiti dialect.
Why: A formal Arabic template for a delivery update reads like a bank notice and gets ignored the same way. Writing every flow in the dialect the founder already used with customers on his personal number is what kept the same warmth once the channel moved to something automated.
Segment the cadence by last product and recency
Replaced the single monthly blast with two to four broadcasts a month, each sent only to the segment whose timing and last product actually matched the offer.
Why: The unsegmented monthly blast had the highest opt-out rate of anything in the account, because it treated a customer two days into her bundle the same as one four months past it. Segmenting by recency is what turned the channel from noise back into something people wanted to stay subscribed to.
Match the offer to the refill cycle, not a generic discount
Built the refill reminder as a straightforward reorder link at the same price rather than a discount, and reserved actual price offers for the sixty and ninety-day lapsed segments.
Why: A consumable that a household is already planning to reorder does not need a discount to convert; it needs a reminder that arrives at the right moment. Saving the discount for genuinely lapsed customers keeps the everyday refill message from teaching the whole list to wait for a markdown.
Win back sixty and ninety-day lapsed customers with a specific reason
Built two win-back messages, each named by the customer's specific first product and timed to when it would realistically have run out, rather than one blanket discount sent to everyone past sixty days.
Why: The old blanket discount trained the entire list to expect a markdown eventually, whenever it arrived. A specific, product-tied reason to come back does not carry that side effect, and it performed better in the segments where it ran.
Measure revenue per recipient and opt-out rate, and prune
Tracked revenue per recipient and opt-out rate for every flow and broadcast, and killed the generic monthly blast once its opt-out rate was measured against the segmented alternatives.
Why: A message can look like it is selling well in raw revenue while quietly burning the list faster than it is worth. Opt-out rate is the number that catches that before the list is gone, and it is why the monthly blast was retired rather than kept for its short-term numbers.
What changed
The table above carries the numbers, and the shape of the change matters as much as the size: repeat purchase rate moved because the refill check-in and the gateway-product welcome flow both fired at a moment tied to real product consumption rather than an arbitrary calendar date. The generic monthly blast, which the brand had treated as its main retention lever before this, turned out to be actively working against the goal by carrying the account's highest opt-out rate.
Revenue per recipient improved as segmentation replaced the single blast, which is the healthier way for that number to move — fewer messages sent overall, each one closer to something the recipient was already going to want. The founder's personal WhatsApp stopped being the entire customer-service operation once the official channel and its flows were carrying the routine questions.
The most durable change was in what the brand now knows rather than in any single number: the gateway product identified by the cohort read is no longer a guess, and it now shapes what gets featured in acquisition creative as well as which customers get the dedicated welcome flow.
What we would do next
First, add a refill subscription option so the reminder becomes an automatic reorder rather than a message the customer still has to act on, since the consumption cycle is predictable enough to support it.
Second, feature the gateway product more heavily in acquisition creative, since it now has a proven repeat rate the current bestseller does not, and bringing more customers in through it should raise the store's average repeat rate rather than just this cohort's.