The scenario
The brand runs occasion abayas and modest sets on a Zid storefront, with KNET, Tabby and cash on delivery live at checkout, alongside a drop calendar announced on Instagram roughly every three weeks. Repeat customers who already know their size checkout on the site directly. First-time customers, who make up most of the traffic a drop announcement sends, almost never do: the product page's only call to action for someone unsure of her size is a plain "order on WhatsApp" button that opens a blank chat.
That blank chat was the whole business model's real checkout page, and nobody had ever looked at it the way they looked at the storefront. The team measured Instagram click-through and Zid conversion rate, both of which looked reasonable in isolation, and never measured what happened to the customer in the gap between the two, where a real share of first-time interest was quietly dying.
- Monthly revenue band
- pending client sign-off
- Average order value
- pending client sign-off
- Fulfilment
- Made-to-order runs, in-house alterations, courier across Kuwait
- Team
- Founder plus two part-time stylists answering WhatsApp during drops
The site was not the checkout. The blank WhatsApp chat was, and nobody had tested it
The product pages were doing their job: strong time on page, a healthy add-to-cart rate for the pieces still in stock, and a click-through to the WhatsApp button that looked, on the surface, like a working funnel. The number nobody had isolated was what share of those WhatsApp clicks actually turned into a completed order rather than a thread that went quiet.
Once we pulled the thread data apart from the site analytics, the pattern was specific. A customer would click through from a drop announcement, open the product page, fail to find any size guide beyond a generic small-medium-large label with no measurements, and land on WhatsApp with a message that read simply "Hi, is this available?" The stylist team, working through a queue of dozens of nearly identical opening messages during a drop, would ask for her height, her preferred length and her size in turn — three separate replies before either side had said anything about the actual piece. By the time sizing was settled, a meaningful share of pieces in her size had already been claimed by someone else, and the reply that came back was that the piece was gone.
None of this showed up as a website problem in the standard sense: the page loaded, the button worked, the checkout accepted KNET and Tabby without issue. The store had built a real checkout for the return customer who already knew her size, and left the first-time customer to restart the entire sizing conversation from a blank message, one piece at a time, on a stock count she could not see.
What we did — the pages and offers we tested
The losers are here on purpose. A test with only winners was never a test.
A real-time stock count on the product page
Winner"3 left in size 54" shown next to the size selector, updating live
- Format:
- Product page, single-variable test against the control
The single biggest source of dead threads was a customer settling her size only to be told the piece was gone. Showing the live count before she ever opens WhatsApp either moves her to message immediately, with urgency that is true rather than assumed, or stops her wasting a thread on a size that is already out, which frees the stylist queue for orders that can actually complete.
A prefilled WhatsApp message carrying the product, size and colour
Winner"Hi, I'd like [piece name] in [size] / [colour]" auto-inserted when she taps the button
- Format:
- WhatsApp click-to-chat link with a structured prefilled template
This removed the three-message sizing back-and-forth entirely for anyone who had already used the on-page size guide before tapping through. The stylist's first reply could go straight to confirming stock and payment instead of re-asking for information the customer had already given the page, which cut the average thread length by more than half.
A size chart built on height and shoulder-to-hem, not S/M/L
Winner"Find your length in one line: enter your height."
- Format:
- Product page module, replacing the generic size label
A small-medium-large label means almost nothing on a made-to-order abaya where length is the real variable. Mapping height directly to a recommended length gave first-time customers enough confidence to message with a size already chosen, and it was the single change most correlated with a shorter, more decisive WhatsApp thread.
A speed pass on the current drop's product pages
WinnerSame photos, compressed, with the size-guide script deferred until scroll
- Format:
- Image compression and script deferral on the highest-traffic drop pages
Drop photography arrived as large uncompressed files, and the size-guide module was loading fully on every visit whether or not a customer scrolled to it. Neither is a page test, both are a floor, and fixing them mattered most in exactly the first ninety seconds after a drop goes live, when the servers and the customer's patience are both under the most strain.
A thirty-minute soft hold on a selected size
Lost"We'll hold your size for 30 minutes while you decide."
- Format:
- Product page button offering a timed hold before WhatsApp handoff
A reasonable idea that created more problems than it solved for a small team. Holds were requested far more often than they converted, real buyers lost pieces to holds that were never followed up, and the stylists ended up doing manual bookkeeping on a promise the storefront had no way to enforce automatically. It was reversed within the same drop cycle.
What we did — the optimizations, in order
Read the funnel past the site, into the WhatsApp thread
We tagged every WhatsApp thread opened from the product page by outcome — completed order, abandoned mid-sizing, or replied to as sold out — and set that against the site's own analytics for the same period.
Why: A funnel that stops measuring at the click to WhatsApp is measuring intent, not orders, and intent was never the brand's problem. This is what turned a healthy-looking click-through number into a specific, addressable leak in the handover itself.
Sit through the WhatsApp threads themselves, not just the recordings
Alongside session recordings on the product pages, we read a sample of real WhatsApp threads with the founder's permission to see exactly where each one stalled or died.
Why: A recording shows a customer scrolling for a size chart that is not there; the thread shows what happens next, which is where the real cost — a piece claimed and lost, three back-and-forth messages, a customer who gives up — actually lives. Neither source alone told the full story.
Fix the floor: speed, the size chart, live stock
Compressed drop photography, deferred the size-guide script until scroll, replaced the generic size label with a height-based chart, and added a live stock count next to the size selector.
Why: These four touch every visitor in the first ninety seconds of a drop, before any test can even run, and a drop that loads slowly or hides its stock count suppresses every downstream fix equally. They came before the page tests, not after them, because the tests would have been unreadable against a shifting floor.
Test the handover itself, one variable at a time
The prefilled WhatsApp message and the thirty-minute hold were each tested on their own against the plain "order on WhatsApp" control, on the same drop pages, with the audience and the offer held identical.
Why: Running both changes at once on a small drop would have made it impossible to tell whether a shorter thread length came from the prefilled message or from the hold, and the two turned out to move the result in opposite directions. Testing separately is the only reason the losing idea could be identified and dropped quickly.
Rebuild the WhatsApp opening message around sizing already given
The stylist team's reply script changed to open with stock and payment confirmation whenever a prefilled message already carried a size, and to fall back to the three-question sizing script only for the shrinking share of threads that still opened blank.
Why: The prefilled message on the page is only half the fix; the human reply on the other end has to actually use the information rather than repeating the old script out of habit. This is the step that turned a shorter incoming message into a genuinely shorter conversation.
Time the alteration check-in and the review ask after the occasion
The post-delivery message moved from a same-day "thank you for your order" to a check-in timed around the occasion the piece was bought for, followed by the review request once she had actually worn it.
Why: A review asked for on delivery day is a review about the packaging, not the fit or the fabric, and fit is the entire trust question this category is bought on. Waiting for the occasion produced fewer but far more specific reviews, which is what a hesitant first-time buyer actually reads before messaging.
Keep a weekly log of dead threads and why
Every WhatsApp thread that stalled or went unanswered during a drop gets logged with a one-line reason, reviewed weekly alongside the page test results.
Why: Without a written log, a two-person stylist team relies on memory during the busiest hours of a drop, which is exactly when patterns like "the size chart still isn't clear enough" get lost. The log is what let the brand tell a genuinely fixed problem from one that had only quieted down for a week.
What changed
The table above carries the numbers once the brand signs them off, and the shape worth flagging now is that the improvement sat almost entirely inside the WhatsApp handover rather than on the site itself: the same traffic, arriving at the same product pages, completed more of the threads it started because it arrived already knowing its size and the piece's real availability.
The live stock count and the height-based size chart did most of the work, and together they attacked the same underlying issue from two directions — one gave the customer confidence to commit to a size before messaging, the other told her honestly whether committing was still worth it. The thirty-minute hold, by contrast, confirmed that a small team cannot manually enforce a promise the storefront itself does not track.
The shorter WhatsApp threads had a second-order effect worth naming: the two stylists could work through a drop's opening rush faster, which meant fewer customers waited long enough to lose patience and message a competitor instead. That is not a line item on the metrics table, but it is very likely part of why the completion rate moved as much as it did.
What we would do next
Bring the structured governorate, area, block and street fields into the Zid checkout for the growing share of repeat customers, so the same address clarity work pays off on the site and not only in the WhatsApp thread.
Second, pressure-test the live stock count and the prefilled message specifically under drop-launch load, since the busiest sixty seconds of a drop are exactly when a stock count needs to be most accurate and most likely to be checked twice before a customer commits.