← Back to blog Field notes

Before / After: a B2B search modernized in 14 days

From the field, with real numbers

Eywora Editorial · April 10, 2026 · 9 min read
Before
38%zero result
340 msresponse time
0.9%conversion
After 14 days
9%zero result
52 msresponse time
1.4%conversion

A manufacturing shop with 280,000 products, 38% zero-result rate, 0.9% conversion rate on search traffic. Two weeks later: 9% zero results, 1.4% conversion. Here’s how the switch went, with real numbers, without marketing varnish.

Starting point

The customer is a technical wholesaler in southern Germany. Assortment: screws, nuts, tools, hydraulic components. 280,000 products, three brand worlds in parallel, volume tiers, technical datasheets. Daily revenue in the six-figure range.

Their standard search delivered:

  • 38% zero-result rate on mixed search and category traffic
  • 0.9% conversion on search sessions (industry median: 2.8%)
  • Average response time: 340 ms
  • Manual synonym maintenance by a two-person team, about 8 h per week

Day 0 to 1: setup

API integration into the Shopware backend, full catalog indexing in just under 14 hours (large image volume, technical datasheets co-indexed). Frontend embedded as a widget, no UI changes to the shop. Parallel operation with the standard search activated for A/B testing.

Day 2 to 7: tuning

In the first week, together with the team, we:

  • Imported the top 200 zero-result queries from the last 90 days
  • Activated industry material mappings (V2A/A2/1.4301 and 15 more families)
  • Reviewed compound-word decomposition for “highpressurehose”, “stainlessnut” etc.
  • Configured re-ranking on stock level and margin class

Customer effort: about 4 hours that week, mostly for assortment decisions, not technical work.

Day 8 to 14: live test with 50/50 traffic split

A week of A/B with real traffic. Both searches live, alternating. Result:

MetricStandard searchEywora
Zero-result rate38%9%
Avg. response time340 ms52 ms
CTR on result positions 1–321%38%
Conversion rate on search0.9%1.4%
Avg. basket value (search order)€184€206

Economic impact

Projected against the shop’s daily revenue and search share, this means:

  • +€1.4M additional revenue per year from an improved search funnel
  • ~€12K Eywora license per year (Growth + add-ons)
  • ~6 h per week of maintenance effort saved (synonyms now automatic)

First-year ROI: roughly 100×. In year two, with maintenance savings: higher still.

€12K investment per year €1.4M additional revenue per year
Roughly 100× ROI in year one. Plus 6 h of maintenance saved, every week.

What’s not in the story, but should be

Three points we don’t put in the pitch story:

  • The first 3 days were bumpy. The norm mappings didn’t cover every material code the shop uses. Two days of engineering work, then it fit.
  • One boost rule was too aggressive. We had the margin boost initially at 30%, which caused too much ranking shift. Reduced to 12%, then balanced.
  • There was a short internal veto. The IT department was sceptical about GDPR. After we showed the DPA and the hosting documentation, it was resolved in one meeting.

What you can take away from this story

  1. Setup is fast. 24 hours to 14 days, depending on complexity, is realistic.
  2. Tuning is collaboration. The first days matter: initial configuration, assortment decisions, boost logic.
  3. Impact is measurable. Conversion, zero results, CTR on result positions are the right metrics, not “the search feels better”.
  4. Maintenance effort drops. Most synonym and boost logic happens automatically once conversion signals flow.

If you want a comparable before/after projection for your own shop: request a fit check. 30 minutes, live in a real customer shop, with an honest assessment.

Methodology & sources

The figures cited are based on an analysis of search logs from 50+ B2B shops in manufacturing, wholesale and spare parts (Eywora projects and prior projects of signundsinn GmbH, 2023-2026). Conversion-uplift values are averaged across Eywora customers after at least 90 days of live operation. Response times are measured as the P50 median. Results in an individual shop may differ depending on assortment, data quality and search share.

Published April 10, 2026 · Author: Eywora Editorial

Ready to see what this looks like for you?

Demo
on your catalog.

30 minutes live in a real customer shop. You see Eywora search and add-ons in action, then we talk about your needs.