Custom software development
Web applications and internal tools, from discovery to a maintained first release.
Every application we build comes with the panels needed to operate it: switch sources and features on or off, review what automated jobs did, watch AI costs, and read traffic statistics that separate people from robots. Built into the application, without third-party trackers.
On our sites a page view counts only when the visitor’s browser confirms it with a small request after the page runs; pages sent but never confirmed are reported as “served, unconfirmed”. On one day on Raftul copiilor, that separated about 2,900 of 3,517 “pages” that were automated speed tests sent with a browser’s name. Robots are counted by name (Googlebot, bingbot and others).
On Cumpără cu AI, clicks to shops are a person’s only when they come from one of our pages in a browser that is not a robot. At one point a crawler began opening nearly every outbound link once a day: 500 to 2,900 clicks daily, against 2 to 84 before; those are now marked as robot clicks, kept for the network but excluded from rankings and reports.
Unique visitors are counted with a hash of address and browser type under a key held in memory and replaced every midnight, so nothing identifies a person and no cookie is needed. The panels also show imports, AI conversations and costs, search problems from real conversations and the configuration of networks, shops, categories and AI models.
What you check daily, what goes wrong, who needs access.
Definitions written down: what counts as a visitor, a click, a success.
Panels inside the application, protected, fast on large tables.
Thresholds that notify you instead of waiting to be checked.
No. A crawler that runs the page and hides its identity can still be counted as a visitor. We separate everything that identifies itself or behaves like a robot and document what the numbers can and cannot tell you.
Our own statistics use no cookies and store only daily totals. Whether your site needs a banner depends on everything else it uses.
Web applications and internal tools, from discovery to a maintained first release.
Controlled AI jobs that classify and enrich data, with validation and a hard budget.
Linux servers with Docker, HTTPS, scheduled jobs, tested backups and safe deploys.
Crawl, indexing, structured data, performance and AI-search access, with prioritised fixes.
Tell us what you need in a few sentences. NeedAI Studio answers personally, in English or Romanian.