Old Website Rebuild Case Study: From Outdated Site to Conversion-Ready Website

TL;DR #

A precision-parts manufacturer in Jiangsu, exporting to Germany, Italy, and Southeast Asia. The old site was a 2017 dedeCMS build with a broken contact form (silent for six months), 11.4-second loads from Frankfurt, and a mobile experience that was just the desktop view scaled down. We ran a four-week rebuild Sprint without changing the domain. We kept all 23 URLs that had real organic traffic in DACH and 301-mapped them one by one. Scope was narrowed to IA and conversion plumbing first, visual modernization only where strictly required, no full redesign. Six weeks after launch, monthly inquiries went from 2 to 11, average response time dropped from 36 hours to 4, and organic traffic was up 18% over pre-launch baseline. The rest of this post is the play-by-play: what the audit caught, the trade-offs we made, what we did not see coming on launch day, and the open issues we still owe the client.

The client #

Mid-sized Jiangsu manufacturer, around 80 employees, precision parts for European machinery brands. Founder doesn't speak much English. Marketing is one export specialist who also handles trade-show logistics. The 2017 site ran on dedeCMS with a so-called bilingual template. State of play at handover:

• Homepage loaded in 11.4 seconds from Frankfurt, LCP at 8.9 seconds.
• "Mobile" was the desktop layout zoomed down. Buttons were hard to tap.
• The Chinese and English versions were two unrelated dedeCMS installs, no hreflang, inconsistent URL paths.
• The contact form had been silently broken for six months. Expired SMTP credentials. Nobody noticed.
• Nine of twelve product images were 4 MB-plus JPGs.
• The team said "we never did SEO," but Search Console showed 23 URLs ranking on long-tail product queries in Germany. That detail mattered later.

At kickoff the client's brief was "make it look modern." After an hour walking through the numbers above, they reframed it themselves: "Then can we just rebuild the parts customers actually see?" That kind of widening, where a client sees the data and pulls scope toward the real problem, is the moment a project gets a real sponsor instead of a vague approver.

Audit #

We worked through the website renovation audit checklist over three days. Three deliverables came out of it:

1. URL inventory. 23 old URLs with real DACH organic traffic. Non-negotiable to preserve via 301.
2. Content gap list. Of 12 product pages, 4 had zero pageviews in 12 months. Collapse-and-merge candidates. The other 8 needed English rewrites.
3. Funnel diagnosis. Form broken six months. No GA4. No UTM. WhatsApp button in the footer with no clicks. The whole inquiry funnel was a black box.

The judgment call out of the audit was uncomfortable. The client wanted a fresh skin, but the actual business problem was a broken funnel. So we redrew scope as IA plus inquiry plumbing first, with only minimal visual modernization. In hindsight that was the right cut. A full visual rework would not have fit four weeks, and the funnel was bleeding leads every week we delayed.

Goals #

We translated "modern-looking website" into three falsifiable goals on page one of the PRD:

1. Mobile LCP under 2.5 seconds, verified on PageSpeed Insights from a Frankfurt origin.
2. An overseas visitor can name what the company does, who buys from them, and how to make contact, after a five-second test on the homepage and any service page.
3. Within 30 days post-launch, every inquiry is attributable to at least a coarse channel: form, WhatsApp, email, or phone.

The third goal mattered most. The first two are experience metrics. The third is what makes the next iteration possible. Without attribution, every future decision is a guess.

Information architecture #

The old site had 47 pages; 31 had five or fewer pageviews in the past year. We cut hard:

• Down to 18 pages: Home, 4 product family pages (collapsed from the original 12), 4 industry solution pages (new), 3 case studies, About, Team, Manufacturing capabilities, Contact, Blog hub.
• All 23 ranking URLs mapped into the new structure. 17 went 1:1 to a new product family or case page. 6 had no clean semantic match and pointed to the closest parent. We tracked the mapping in a spreadsheet, dropped it into nginx as a rewrite block, and walked every line during QA.
• Added an /industries/ path with pages by buyer industry: automotive, medical, food machinery. This came directly from old GA queries. Plenty of DACH visitors searched by industry, and the old site had no surface for that intent.

The biggest IA trade-off was whether to keep the Chinese site at all. After two rounds of discussion we kept it but starved it: only Home, About, and Contact in Chinese, no products, no cases. The reasoning: 90% of domestic Chinese inquiries came from trade shows and referrals, not the website. Maintaining two parallel content trees would have eaten the export specialist's bandwidth on translation work that did not pay back.

Visual and front-end #

We did not redesign the site. We did fix:

• Above-the-fold hero. Replaced the 2017 factory-gate stock-style photo with a real shop-floor photo plus a single English value prop: "Precision parts for European machinery brands since 2009." In the post-launch five-second test, four of five overseas friends paraphrased that line correctly.
• Navigation. From 11 items down to 5: Industries / Products / Cases / About / Contact. Sticky header on mobile only.
• CTA system. Every product family page and industry page got the same CTA card at the bottom. Left side: a sentence ("Need this part for your line?"). Right side: a WhatsApp button next to a short form link. Simple, repeatable, and it lifted WhatsApp engagement more than I expected.

Front end is WordPress on a lightweight theme (GeneratePress) with custom blocks for the load-bearing pages. No page builder. The export specialist needed to maintain content after handover, page builders carry too steep a learning curve, and the performance penalty is real. For more on the stack reasoning under tight timelines, see How Our Small Team Runs a Four-Week Website Rebuild Sprint.

Conversion plumbing #

This was the bit with the most direct revenue impact. Three pieces:

1. Form. From 9 fields to 5. Submissions go to the business mailbox and Zapier-sync to HubSpot Free. SPF, DKIM, DMARC reset from scratch. Mailgun for delivery. That stack avoids the exact failure mode that broke the old form: a single expired SMTP credential and no one watching.
2. WhatsApp. Configured per the WhatsApp Business setup checklist. Buttons on every product family and industry page, all linking to a wa.me short URL with a pre-filled message that includes a product family code. Sales sees which page the conversation came from in the first message.
3. UTM and analytics. Every form and WhatsApp link tagged with UTMs. GA4 and Search Console reconnected. Four conversion events configured in GA4: form_submit, whatsapp_click, phone_click, email_click.

Field-level conversion patterns used here are documented in the corporate website conversion checklist. We essentially walked that checklist line by line.

SEO preservation #

Highest-pressure piece of the rebuild because the 23 ranking URLs were paying the bills in the DACH region. Approach:

• 301 map. All 23 mapped one-to-many at the nginx layer. We curl-tested every line on staging before launch. Each URL had to return 301 with a correct Location header.
• Content carryover. We didn't just delete the old product copy. The substantive parts moved into the new product family pages. Image alt attributes were rewritten in English (the old site had blank or Chinese alts on most product images).
• Schema. New site ships with Organization, Product, BreadcrumbList, and FAQPage schema. Reference: Google Search Central structured data guidance.
• Sitemap and Search Console. New sitemap.xml submitted on launch day. We left the old sitemap reachable for 30 days so Google could observe the 301 transitions.
• Image weight. All product images converted to WebP. Average file size dropped from 4.2 MB to roughly 180 KB. That single change accounted for most of the LCP improvement. For more on case-study and product imagery as an SEO surface, see Image and Case Study SEO.

On day 7 and day 21, we used Search Console URL Inspection to spot-check 12 of the ranking old URLs. On day 7, 11 of 12 were already recognized as redirected with equity flowing to the new pages. The last one caught up by day 35. Organic traffic dipped 6% in weeks 1–2, recovered to baseline by week 4, and was 18% above baseline by week 6.

Launch checklist #

The last two days of the Sprint weren't development. They were the launch checklist. The 25-item core list:

• All 23 redirects return 301 (curl-verified).
• Mobile LCP under 2.5s on PageSpeed Insights, Frankfurt origin.
• Form submit → CRM sync → email notification all three legs working end-to-end.
• WhatsApp button opens correctly with the pre-filled message on iPhone Safari and Android Chrome.
• All four GA4 conversion events visible in DebugView.
• sitemap.xml submitted to Search Console and Bing Webmaster Tools.
• robots.txt audited so nothing important is accidentally Disallowed.
• hreflang correctly cross-references the English and Chinese versions (yes, even though the Chinese site is only three pages now).
• Email SPF, DKIM, DMARC all pass on mail-tester.com.
• Five-second test: five overseas friends can describe what the company does after a five-second look.

We turned the full version into an internal template. The export specialist ran it themselves and caught one missed redirect (an old blog URL) the day before launch.

First 30 days #

Three things we did not predict:

1. WhatsApp engagement was 3× what we modeled. Sales couldn't keep up. We wrote three English reply templates (price inquiry, needs more info, out-of-scope) on the spot.
2. An Italian buyer never received the auto-confirmation email. Their corporate gateway had blocked one of Mailgun's IP ranges. Switching the sending IP pool fixed it. This kind of thing is invisible in audit and only surfaces from real traffic.
3. One redirected URL had been a PDF brochure download. We had pointed it at the matching product family page, but visitors actually wanted the PDF. We caught it on day 10 from the GA bounce pattern, then stood up a single-purpose page that preserved the download.

Day 30 retrospective produced three numbers: inquiries 2 → 11, response time 36h → 4h (CRM sync plus live WhatsApp), organic +18%. Next phase: one industry hub article per month, four customer cases with real client names, English FAQ to ~20 entries.

Open issues #

Things this project did not solve, in plain language:

• Languages beyond English. The client wants Italian and German next. We pushed back. Invest in English content depth first; multilingual is a content-operations problem, not a translation problem. See Multilingual Site Structure and Hreflang.
• Case study permissions. Two of three case pages still use anonymized clients ("a German automotive supplier"). Half-anonymized cases convert worse than logo-named ones. Getting permission is the next sales-team task.
• CRM still on Free tier. HubSpot Free's automation is limited. We'll revisit when monthly inquiries hit roughly 30 consistently.

FAQ #

Is a four-week Sprint really enough? #

Yes, but only with disciplined scope. We dropped the full visual redesign and limited the work to information architecture, conversion plumbing, and SEO preservation. A full visual system pushes timelines to 8–10 weeks. For the scope-cutting heuristic, see Website Redesign vs Rebuild vs Migration.

What does this kind of rebuild cost? #

This one landed in our middle band. Build, content, and SEO configuration completed inside the four-week Sprint, plus 30 days of light post-launch optimization. The factors that move the number up or down are covered in Website Renovation Cost and Timeline.

Do we have to redirect every old URL? #

Not every URL, but every URL with measurable organic traffic, yes. The cut: list every URL with at least one organic click in the past 12 months, then decide per row whether it's a 1:1 redirect, a merge into a parent page, or a clean 410 (telling Google the page is gone). Skipping this step is how rebuilds visibly lose rankings in week two.

Was the conversion stack expensive? #

No. The full setup here was WordPress + GeneratePress, HubSpot Free, Mailgun (pay-as-you-go, well under $20/month), Cloudflare's free CDN, GA4, and Search Console. Monthly running cost stays under $50. The client's budget went almost entirely into the one-time build and English content work.

Get a focused diagnosis #

If your company website is five or more years old, won't open cleanly on a phone, and you're not sure whether the form even works, send your domain and a Search Console screenshot. We'll run the same audit on it through our website rebuild service and tell you whether to fix the funnel first, preserve SEO first, or rework the visuals, with an honest scope plan that fits a small team's budget and timeline.