Proline's cosmetics store went from invisible to 10 export buyer enquiries

The situation

The technical faults we fixed behind the growth numbers

• The Greek store was selling in the wrong language

  Within the first week we found the problem the client had not reported, because nobody had looked. Proline ran three storefronts from one e-commerce instance: Greek, Bulgarian, and English. The platform lets the same product appear across storefronts, and per-store content settings had been left inherited rather than set. The result was that a large share of the Greek catalogue, out of roughly 3,500 products, displayed Bulgarian text to Greek shoppers.

  This mattered more than the missing metadata. Google cannot rank a Greek product page that reads in Bulgarian, and a Greek buyer who lands on one does not trust the brand. Fixing it meant exporting the full catalogue, correcting content store by store, and reimporting in batches of 500 so a bad batch stayed small and recoverable.

• Nobody at RaftLabs had shipped this platform before

  We are a React, Node, and AWS team. The client's platform is a decade-old PHP stack with deep configuration surface, and no one internally had production experience with it. There was no colleague to ask whether a given change was the right one.

  We handled this with a discipline rather than a shortcut. The client's production database was cloned to a staging environment first. Each fix followed the same loop: the SEO lead specified the outcome, the engineer researched how the platform expresses it, applied it on staging, confirmed the rendered result, then repeated it on production. Where staging could not answer a question, we brought in an external platform specialist rather than guessing on a live store. We did not attempt anything we could not verify.

• Bulk import and export failed on staging but worked in production

  Editing 3,500 products by hand was never viable, so bulk export and import was the whole content plan. On staging, both failed, and we could not determine why within the time available.

  Rather than burn the pilot on a staging bug, we raised it with the client. Their own development team resolved the export issue on production inside a day. We then ran the content work directly against production in controlled 500-product batches, with the staging environment still carrying each structural change first. That split, structural changes proven on staging and content changes batched on production, kept the risk contained without stalling the engagement.

• The sitemap was pointing Google at the wrong store

  The Greek store's sitemap.xml was corrupted and included URLs belonging to the Bulgarian storefront. We did not expect this, and it undermined everything else: we could correct 3,500 product pages and still hand Google a map of the wrong country's catalogue.

  Regenerating a store-scoped sitemap was straightforward once found. Finding it was the work. It is the kind of failure that produces no error anywhere in the admin interface and no visible symptom on the site.

• Every AI crawler was blocked

  Proline's robots configuration disallowed the AI crawlers. No AI assistant could read the catalogue, so none could recommend the brand to a buyer asking which Greek supplier stocks a particular gel or base coat. For a supplier whose buyers increasingly start with an AI assistant instead of a search box, that is a closed door.

  Opening it was a one-line change and the highest-leverage decision in the pilot. Appearances in Google's AI answers in Greece went from 11 in the month before to 23 in the month after, a 109% increase.

1. Europages, which is where the buyers actually were

   Europages is a B2B supplier directory used across Europe. We built and optimised Proline's profile there in August. Profile impressions went from zero to 421 in the first half of September, a 114% increase over the prior period, and the profile generated 10 genuine buyer enquiries from six countries. These included a 500kg full-portfolio request from a German distributor, a 100-unit reseller order from Belgium, and catalogue and pricing requests from Portugal, France, Spain, and Ecuador.

   This was the outcome nobody predicted. The pilot was scoped around organic search for the Greek consumer store. The demand that arrived was B2B and cross-border.

2. Google Merchant Center, built from scratch

   The catalogue had never been submitted to Google Shopping. We created the account in August and fed the corrected product data into it. From a zero baseline, the store recorded 36 clicks and roughly 1,000 product views by mid-September, with click volume rising 24% between the first and second measurement periods.

3. A Google Business Profile for the manufacturer

   Proline had no local presence in Google at all. A new profile went live in August and recorded 31 interactions and 52 views by mid-September, again from a zero baseline.

4. Ensun, an AI-driven sourcing platform

   Ensun surfaces suppliers to buyers who are sourcing through AI-assisted search. We created the listing on 3 September. In the two weeks that followed, 141 potential buyers viewed it.