Shamdi

Background

Shamdi is a Vietnamese sportswear brand that started in Nha Trang in 2017: gear for the gym, yoga, running and pickleball. Products, prices, stock and retail receipts live in the Nhanh.vn software; shamdi.com is where customers browse, pick a size and pay.

For the CEO, the website is only worth something when two things are both true: shoppers get from browsing to paying without a hitch, and the numbers the CEO reads every day are the real numbers of the website and the shops together.

What was in the way

The online shop and Nhanh.vn are two separate systems. For the CEO to see revenue, stock and customer behaviour as real numbers on one screen, KP3 had to connect them without slowing down or breaking the selling that was already running.

At the same time the buying path still had small faults that cost orders: a search box that returned too little, buttons that did nothing, categories mixed with unrelated products, products with no picture at the top of the page.

What KP3 did

1. Audit first, polish later

   KP3 started with two audits on 30 Oct 2026: one of the shopping interface, one of security. The interface audit listed 49 faults by severity (14 that broke a function or showed wrong data, 29 clearly visible ones, 6 small ones), each with its cause in the code. The security holes found in the buying path and the stock sync were patched before anything else was built.

2. Connect the website to Nhanh.vn so the numbers are real

   An order placed on the site is saved in Shamdi’s own system and pushed to Nhanh.vn through a queue: on an error it retries after 1 minute, 5 minutes, 15 minutes, 1 hour, 3 hours and 12 hours, with no lost order and no duplicate. A bank-transfer order only goes to Nhanh after the money has arrived, so stock is never held for an unpaid order. Stock flows from Nhanh to the site the moment it changes and is reconciled warehouse by warehouse every 30 minutes. Sales from the shops and the site are pulled in every 10 minutes.

3. Redesign the buying path

   A two-column homepage with a best-sellers tab; a big menu that shows categories, sports and real best sellers. Product pages with two-column pictures, a clear size chart and colour swatches taken from real fabric photos. Checkout on one two-column page: the bank-transfer QR code comes first, with the processing timeline and the expected delivery date so the customer feels sure after paying, and an add-on suggestion that takes one tap.

   

4. Measure customer behaviour with real numbers

   Google Tag Manager, GA4 and Microsoft Clarity were added. Viewing a product, adding to the cart and buying are recorded as e-commerce events tied to the customer’s account, and the GA4 numbers are read straight into the admin.

5. An admin for the CEO: one screen, real numbers

   The admin home shows revenue against the same period, today’s to-do list, progress against the month’s goal, the health of the customer groups (new, returning, VIP, at risk of leaving) and suggestions to win customers back. Each card loads on its own: if one source fails only that card reports it and the page never goes blank. Orders are split into tabs by processing step, search is one ⌘K away, and the sales staff have a leaderboard and monthly KPIs.

6. Stock forecast: reorder or transfer

   For every product in every warehouse the average sales rate over 28 days is worked out and sorted into levels: out of stock, urgent (under 7 days left), running low (under 14 days), long-held (over 90 days). Only products that sold at least 2 in 28 days raise an alert, so the list stays clean, and the system suggests a transfer or a reorder for 30 days of sales. With 3 warehouses and 3,086 products the query takes 61 milliseconds. It only suggests; every action is still done in Nhanh.vn.

Results

• 200 million VNDsaved against hiring an outside team to build the internal dashboard alone
• 49faults on the buying path reviewed, the most serious ones first
• 10 minfor sales from the shops and the site to reach the admin
• 3,086 productsforecast automatically across 3 warehouses

The 200 million covers the internal dashboard alone. It does not include the real-time sync between the website and the point-of-sale software, nor the website itself (about 100 million).

Before-and-after conversion rates are not published yet because there is not enough data to compare. KP3 will post them when there is.