Level Design: How We Build Levels Players Want to Finish
Level design is where game mechanics meet a player's patience. A perfectly tuned mechanic can still feel terrible if the levels around it ask for the wrong thing at the wrong time. This is how we approach designing, prototyping, and balancing levels so the difficulty curve feels earned rather than punishing.
Start with the verb, not the layout
Before drawing a single room or arena we name the core verb the level teaches — dodge, time, deduce, aim. A level exists to give the player a focused space to practice one idea. In an arcade title like Power of Bombs, an early level should isolate the dodge-and-time verb with very few distractions, so the player builds the muscle before we layer in pressure. Designing layout first, without a verb, produces pretty spaces that teach nothing.
Greybox before art
We build every level twice. The first build is a greybox: untextured shapes, placeholder timings, no polish. Greyboxing is the cheapest place to discover that a jump is impossible or a puzzle is unreadable. We playtest the greybox aggressively because changes here cost minutes, while the same change after the art pass costs days. Only once the greybox plays well do we hand it to artists.
The difficulty curve is a sawtooth, not a ramp
Players burn out on a constant climb. We shape difficulty as a sawtooth: introduce a tough challenge, then deliberately drop to an easier beat so the player exhales and feels powerful with their new skill. Each peak is a little higher than the last. This rhythm — tension, release, tension — is what keeps someone playing one more level at 1am instead of putting the phone down.
We also respect the first-session quit risk. The opening levels carry more forgiveness and clearer signposting than anything later, because we have one chance to convince a new player the game respects their time.
Telemetry closes the loop
Once a level is live, data tells us what playtesting could not. If a single level shows a sharp spike in failures or quits, that is a design bug, not a player problem. We instrument levels to log where players die, how long they take, and where they give up, then re-tune the spots that the numbers flag. Good level design is never finished at launch; it is finished when the curve looks fair in the data.
When the level is a floor, not a screen
Building Up The Beat — the interactive floor game we delivered for a retail activation in Kuwait — made us re-learn level design with the screen taken away. A public installation gets no tutorial: a passer-by has to understand the game from metres away, in the time it takes to walk past, so watching one round has to BE the tutorial. The "level" is really the round structure, and its first job is legibility — seven lit, pressure-sensing circles have to announce what the game wants without a single word on screen being read.
Difficulty in a physical game is tuned for bodies, not thumbs. Reach, stance and reaction speed vary far more between a child and an adult than tap speed varies between two phone players — and you cannot patch a floor after launch the way you patch an app-store build, so the tuning has to be right before the unit ships. The sawtooth still applies, compressed into a couple of minutes: an easy opening beat so the first press feels immediate, a spike that gets the crowd involved, and an ending where the player is beaten by the game rather than confused by it. The spectators are part of the level too — a round that is fun to watch recruits the next player better than any signage.