Combinatorial explosion: when there’s suddenly more data than coffee in the office

In PDP, we hit a classic enterprise pain point: 5 buildings × 25 floors × 4 sections × 8 apartments = 4,000 units. And that’s just the base. For each one there are dozens of notes and inspections, and then the real boss level starts: 6 inspection types × 3 severity levels × 27 tag categories. That gives us a huge state space where, without optimization, the system quickly says: “that’s it, I’m done.”

Here’s what we did to avoid a freeze, like a heavy frontend app on a weak laptop: — batch operations so we wouldn’t hit the system one entity at a time; — smart caching so we wouldn’t recalculate the same thing a hundred times; — pagination with server-side filtering so the client wouldn’t have to pull the whole data ocean over to itself.

Against the backdrop of the hype around AI and Big Data, the old truth still holds: scale doesn’t break the algorithm on paper, it breaks the small things in production. We ran into that, adjusted the architecture, and things became a lot calmer.