Case study
Eight months to routing that works: what edge cases and an overly talkative knowledge base taught us.
A European B2B technology distributor with around 2,000 employees brings together eight specialist markets and sells through twelve locations in Europe. Enquiries landed in a shared inbox and were forwarded by hand, often only days later. Customers wanted advice beforehand, in their language, and for thousands of products there was nothing but the manufacturer's PDF data sheet.
Entry 1
Starting point
A design engineer in Milan is looking for a sensor for an application in one of the specialist markets. He fills in the form on the website. The enquiry goes to a central inbox, where someone reads it, recognises the market, assigns the country and forwards it to the Italian team for that market. If this happens on a Friday afternoon, or the enquiry is not clear, days go by. In the meantime, a competitor replies.
Sales is organised by region and works in NetSuite. But website enquiries did not arrive there in structured form; they came as forwarded emails that someone had to type in. There was no report on which markets and countries were asking.
Then there was the need for advice before an enquiry. Customers want to know which component fits and which data sheet applies before they fill in a form. Specialists answered these questions by phone, during office hours and not in every language. And many products had no text on the website, because the manufacturer only provided a PDF. The trigger for the rebuild was the website relaunch.
Entry 2
Figures before
| Markets | 8 specialist markets |
|---|---|
| Locations in Europe | 12 |
| Intake | one shared inbox for all website enquiries |
| Forwarding | by hand, often several days before it reached the responsible team |
| Website enquiries | baseline 100 (before the relaunch) |
| Advice | phone and email, during office hours |
| Product data | thousands of products only as PDF data sheets |
| Reporting | none by market or country |
Entry 3
The path
Eight months from project start to go-live, together with the website relaunch. The week figures are rounded.
- Weeks 1 to 4
Analysis of the enquiries
We analysed one quarter's enquiries: where they come from, which market they concern, who is responsible, which details are missing. The result was a matrix of market and country with the responsible team in each cell.
- Weeks 5 to 10
Routing rules and NetSuite
Every enquiry is classified by market and country of origin on arrival and written as a complete lead into the NetSuite pipeline of the responsible team. A dashboard shows who is asking about what, and from where.
- Weeks 8 to 20
Knowledge base and chat
Product documentation, white papers and catalogues form a dedicated knowledge base. A chatbot answers product questions in English, German, French, Dutch and Italian, names the source and hands over to a person when it comes to prices, lead times or a specific project.
- Weeks 12 to 28
Product data from manufacturer documents
The PDF data sheets are turned into structured product data and texts. Every record is checked manually before it goes live. Only what is new gets translated; editorially approved texts stay untouched.
- Weeks 28 to 34
Parallel operation
Routing is live, and the shared inbox still receives a copy of every enquiry. Sales staff from each language area test the chat with their own questions. Whatever is routed or answered wrongly, we correct in the rules.
- Month 8
Go-live and handover
The website goes live with routing, chat and product data. The shared inbox becomes the fallback for enquiries that no rule assigns clearly. Since then we have run it as part of ongoing operation: new markets, new content, new integrations.
Entry 4
Result
Since the relaunch, enquiries via the website have risen by 400 percent, and each one goes straight to the responsible team, with all details in NetSuite. Advice is available around the clock in five languages. Sales sees in the dashboard which markets and countries are asking, and plans accordingly.
To be honest, the figure measures the relaunch as a whole, not just our workflows: a new site, new content, better search visibility. But without automatic routing, five times as many enquiries would have piled up in the shared inbox.
| Metric | before | after |
|---|---|---|
| Website enquiries | 100 | 500 (plus 400 percent) |
| Forwarding | by hand, often days | by rules on arrival |
| Arrival in sales | forwarded email | complete lead in NetSuite |
| Advice | during office hours | around the clock, five languages, with source |
| Product data from PDFs | often missing | generated with AI, checked manually |
| Reporting | none | dashboard by market and country |
Operating mode: the knowledge base sits on our own infrastructure, the chat runs via a cloud model under contract, the enquiries are stored in NetSuite. A person approves product data before it goes live.
Entry 5
What went wrong
- Edge cases in routing. A customer is based in country A, the delivery goes to country B, the design team sits in country C. Our rules only knew the country of origin, and such enquiries ended up with the wrong team. We had to revise the rules twice. The lesson: edge cases are collected from real enquiries before the start, and every rule has a fallback that passes an unclear enquiry to head office with a reason, instead of guessing.
- The knowledge base answered at too much length. The chat answered a short question with three paragraphs from a white paper. Correct, but nobody reads that in a chat window. We switched to short answers with the source and a link to the data sheet. The lesson: answer length is a rule you set, not something that just happens.
- Unit errors in old data sheets. Older PDFs mixed inches and millimetres or used decimal commas and decimal points inconsistently. The first generated records copied this. Today, the check compares every value with a plausible range for each product type and holds back outliers for review.
Entry 6
What we would do differently today
We would make the catalogue of edge cases a fixed part of the analysis. In the first four weeks, we looked very closely at the typical enquiries and too little at the rare ones. It is precisely the rare ones that cause rework in operation.
We would set the answer rules for the chat together with sales before filling the knowledge base: how long, with which source, from what point a person takes over. That would have saved us a round of rework after parallel operation.
And the unit check belongs in every workflow that reads data from documents, from the start. Today it is standard for us, not an afterthought.
Entry 7
Key facts
| Industry | Technical distribution, B2B technology in eight specialist markets, Europe |
|---|---|
| Employees | around 2,000, twelve locations in Europe |
| Systems | NetSuite, WordPress, Google Cloud, Cloudflare, vector database for the knowledge base |
| Services | Sales and communication, Texts and content at scale, system integration, ongoing operation |
| Operating mode | Knowledge base on our own infrastructure; chat via a cloud model under contract; enquiries in NetSuite |
| Duration | eight months from project start to go-live, in ongoing operation ever since |
Read more
Further reading
Handover
The first step is a 30-minute call.
You tell us about the workflow that costs you the most time. We tell you honestly whether AI pays off there and what the next step would be. Whether a workflow analysis follows is up to you.