Short answer: A WooCommerce store selling stationery and kids' learning toys in Bangladesh came to me with three problems at once: the owner couldn't edit the site without breaking it, the host had suspended the account for CPU overuse, and full backups with customer data were publicly downloadable. Over a few focused phases I replaced the page builder with a lightweight theme the owner can manage, fitted the checkout to Bangladeshi buyers, fixed caching and bot load, moved the store to a better host behind Cloudflare, and ran a full security audit. Every change was tested on the live site.
The client and the starting point
The client runs an online stationery shop in Bangladesh: school supplies, art materials and learning toys, with a catalogue of a few hundred products, built on WordPress and WooCommerce. Most customers order from their phones and pay by Cash on Delivery or bKash.
By the time the work started, the owner was frustrated with the website itself. What they wanted was simple:
- a site their own staff could update without breaking the layout;
- a checkout that doesn't force an email and makes bKash orders arrive with the payment details;
- a site that stays online, after the host suspended it.
When I looked under the hood, I found more:
| Area | What I found |
|---|---|
| Editing | Built with Elementor and a Pro add-on; header, footer and homepage were builder templates |
| Checkout | Email required; bKash Send Money details not enforced |
| Shop page | A Load More button with no crawlable URLs for later products |
| Order tracking | Only the default form, which needs the billing email |
| Performance | Cache plugin installed, but page cache switched off; a stray drop-in from an old plugin |
| Bot traffic | Crawlers hitting ?add-to-cart= links, which can't be cached |
| Hosting | Shared cPanel plan; account suspended for CPU and memory limits |
| Security | A 2.2 GB site archive, two full-site migration backups, a theme zip and a notes file, all downloadable from the web |
| Maintenance | Several plugins out of date; SSL between Cloudflare and server not strict; TLS 1.0 allowed |
My approach: fix what hurts, in an order that's safe
A store that takes orders every day can't be switched off for a rebuild. I set three rules:
- Work on a local copy first. Every change was built and tested on a local WordPress install before it touched the live site.
- One area per phase. Header, then homepage, then footer, then blog and pages, then checkout. Each phase was checked on desktop and mobile before the next began.
- Always keep a way back. Back up before every upload, deactivate before deleting, move files instead of deleting them.
Phase 1: A theme the owner can actually manage
I replaced Elementor with a small child theme that uses standard WordPress screens:
- Header and mega menu from Appearance > Menus. A menu item with sub-items becomes a mega menu automatically.
- Homepage built from "Home Sections" in the dashboard: product rows (newest, best selling, on sale, handpicked), shop-by cards, promo banners and a brand list. Each can be reordered or switched off.
- Footer text and links in the Customizer.
- A help page inside wp-admin that answers the questions staff actually ask.
One-click importers rebuilt the old menu and homepage, so go-live didn't need hours of manual setup. Elementor stayed installed but inactive for a few days as a safety net, then was removed.
Details: Rebuilding a WooCommerce store without Elementor.
Phase 2: A checkout built for Bangladeshi buyers
- Email optional, moved to the end of the form.
- Cash on Delivery first.
- bKash Send Money with server-side checks: the order can't be placed without a valid bKash number and a TrxID.
- Order tracking without login: order number plus phone number, no personal data shown, and 15 attempts per 10 minutes per visitor.
- Numbered pagination on real URLs instead of Load More, with AJAX kept only for filters.
Details: the checkout changes, order tracking without login and Load More vs pagination.
Phase 3: Ending the hosting suspension
The host reopened the site for 60 minutes and asked for the cause to be fixed. It wasn't a "heavy" site. It was an uncached site under bot traffic:
- Removed the stray
advanced-cache.phpleft by an old plugin. - Turned on W3 Total Cache's page cache with Disk: Enhanced, so cached pages are served without starting PHP.
- Excluded cart, checkout, my-account and order tracking, plus WooCommerce cart and session cookies.
- Blocked bots from add-to-cart links with
robots.txt, a403for bot user agents and Cloudflare's Bot Fight Mode. - Moved DNS to Cloudflare's free plan, with only the website records proxied.
Then I tested the three states that matter: anonymous visitors got cached pages, shoppers with items in the cart got fresh pages, and checkout was never cached.
Details: the suspension fix, WooCommerce cache exclusions, stopping bots and the Cloudflare setup.
Full-funnel campaign management, from setup to scaling spend profitably.
Phase 4: Moving to a better host
The old plan had little headroom left, so the owner bought a plan on a LiteSpeed server in Singapore. I:
- took separate file and database backups and checked that no orders came in after the database dump;
- uploaded a clean copy of the site, without old archives and caches;
- re-imported the original database so all URLs were already correct;
- connected the domain, switched DNS in Cloudflare and got the host's SSL certificate issued;
- tested every key page, add to cart, checkout with both payment methods, and order emails.
One mistake: I switched the root record before the new server had a certificate while www was still proxied with Full (strict), and the site showed a Cloudflare 525 error until the certificate was issued a few minutes later. I wrote up how to avoid that in the migration post.
Phase 5: The hidden problems and the security audit
The day after the move, a deeper check found what a visual test missed:
- the page cache was off again, its drop-in missing and the exclusions gone;
- an older theme file had come back, without two security fixes;
- the upload archive and the database dump were in the public folder, and Cloudflare had cached a copy of the dump.
All three were fixed the same day. Then I ran a full audit:
- WordPress core checksums verified; 18 plugins verified against WordPress.org;
- no PHP files in uploads, no backdoor patterns;
- login, xmlrpc, user listing and directory listing blocked;
- four outdated plugins updated one at a time after a database backup;
- the migration plugin and the old page builder removed;
- theme and plugin file editing disabled in
wp-config.php; - Cloudflare set to Full (strict) with TLS 1.2 minimum.
Details: what to check after a migration, exposed backup files and the audit checklist.
Results
What changed, all verified on the live store:
| Before | After |
|---|---|
| Page builder; edits broke layouts | Menus, Home Sections and Customizer; no builder |
| Email required at checkout | Email optional; phone is the key detail |
| bKash orders without payment details | Order can't be placed without a bKash number and TrxID |
| Tracking needed an email | Tracking with order number and phone |
| Load More, products hidden from crawlers | Numbered pages on real URLs |
| Page cache off, account suspended | Page cache on, tested for visitors, shoppers and checkout |
| Shared cPanel plan | LiteSpeed server in Singapore behind Cloudflare |
| Backups with customer data downloadable | Removed from the web, purged from Cloudflare |
| Outdated and unused plugins | Updated; unused plugins deleted |
| Cloudflare SSL "Full", TLS 1.0 | Full (strict), TLS 1.2 |
After the move, cached pages through Cloudflare returned in roughly 0.1 to 0.4 seconds in my own tests from a regular connection (homepage around 0.3 s, shop around 0.2 s). I don't have reliable measurements from before the work, so I won't claim a "before vs after" speed percentage.
What I'd tell any shop owner from this project
- "Installed" isn't "working." The cache plugin was installed the whole time. Test the feature, not the plugin list.
- Most hosting suspensions are fixable without an upgrade. Check caching and bot traffic before you pay for more CPU.
- Backups are the easiest leak. Archives in the web folder had customer data in them, twice. Keep backups out of
public_html. - A migration isn't finished when the homepage loads. Settings, file versions and leftover files need checking.
- Give the owner tools they can't break. Structured sections beat a free-form builder for a store with a fixed design.
What's still on the owner's list
Two steps only the owner should do: change the admin password and turn on two-factor authentication in Wordfence. After that, enabling HSTS in Cloudflare is the next improvement.
FAQ
How long did the work take?
The theme rebuild was prepared on a local copy first. From the day the new theme went live, the suspension fix, Cloudflare setup, hosting move and security audit took about three days.
Was the store offline during the work?
No, apart from a few minutes during the DNS switch, when Cloudflare showed a 525 error until the new server's SSL certificate was issued. Everything else was done on a local copy first or applied live with a backup in place.
Did the owner need a bigger hosting plan?
Not to end the suspension. Fixing the cache and bot traffic did that. The move to a new host was a planned upgrade for more headroom afterwards.
Can the owner manage the site without a developer now?
Yes, for day-to-day changes: menus, homepage sections, banners, footer, products and pages all use standard WordPress screens, with a help page in the dashboard.
The store's name is kept private; all technical details are from the live project. If your WooCommerce store has similar problems, let's talk through my web development service.
Clean, accurate tracking infrastructure that reports numbers you can trust.

