Streek Support

Beter Bij Ons is a digital agency in Sittard that builds websites, apps, webshops and platforms as a long-term partner for its clients. Streek is one of them - a marketplace connecting consumers with local food producers and stores, used daily by three very different roles: consumers, store managers and producers. (Figure 1.0).

Until this project, there was no self-service support at all - every question went through a person. Whatever replaced that had to be clear, structured and role-based: all three roles touch the same screens but need different answers, and that difference shaped everything I built. (Figure 1.1).

There was no structured self-service, so even the most common questions landed with the team first. The same topics came up again and again, responses were slow, and the support team spent hours on work a good FAQ could absorb.

What if users had a clear first place to look before contacting anyone? What if the same screen could answer a consumer and a producer differently? And what if the questions that truly needed a person arrived structured, instead of scattered?

I worked in four steps: research to understand the real problem, design in Figma, validation with users and stakeholders, and a build in the production codebase. Stakeholder interviews, customer journey mapping, competitive analysis of other help centres, user surveys and usability tests all pointed the same way: a small set of topics caused most requests, each role needed different answers on the same screens, and there was no first place to look.

The chatbot deliberately uses no generative AI and no free-text input. It walks a predefined decision tree linked to verified FAQ content - users choose from buttons at every step. That was a design decision: answers stay consistent, nothing can be hallucinated, and the company keeps full control over what users are told. (Figure 3.0).

Users pick their role - consumer, store manager or producer - then a topic: orders, products, payments, returns or account. The chatbot surfaces the matching FAQ answer and asks whether it solved the problem. If not, a short form (name, email, phone, subject, question) creates a Microsoft Teams ticket for the support team, and the user gets a confirmation with a unique ticket reference number. (Figure 3.1).

Built in Streek's production codebase with Next.js, React and TypeScript, the front-end never touches the database directly - every write goes through a thin custom API that validates input and owns the write logic, while Appwrite stores tickets and handles authentication. The chatbot re-uses the FAQ answers, so support content stays consistent across the whole product.