What a narrative designer actually does

The distinction that matters

Take a scene where the player learns their ally has been lying. A writer's question is what the ally says, and how it sounds coming from them.

A narrative designer's questions are different, and there are more of them. Does the player find out, or can they miss it entirely? Is it told in dialogue, or discovered in a room, or implied by an item in an inventory? What did the player have to do to get here, and what changes afterwards — a variable, a faction, a door? If they already guessed two hours ago, does the scene know that?

Only one of those questions is answered with sentences. The rest are answered with structure: conditions, state, ordering, and what the systems are allowed to say about each other.

What the job produces

The shape of the story

Linear, branching, hub-and-spoke, or systemic — and, more importantly, where it changes shape. Most games are linear with three branching moments, and pretending otherwise multiplies work that nobody will see.

The graph

Scenes, conditions, consequences, and the state they read and write. This is the actual deliverable on most teams, and it is a structure rather than a document.

The variable contract

What the story is allowed to know about the player, and who else writes to it. The unglamorous half of the job, and the half that breaks builds when it is skipped.

Pacing and gating

What unlocks when, what a player can miss, and what happens if they do things in an order nobody expected — which they will.

Enough prose to test it

Placeholder lines good enough to play through, so the structure can be judged before anyone writes the real dialogue.

Two failures that look like bad writing

The branch nobody sees. A team builds a choice with four outcomes, and telemetry later shows ninety percent of players took the same one. Four outcomes of work, one outcome of value. This is a design failure, not a writing failure, and it is preventable by asking what fraction of players will see a branch before building it.

The story that does not know what happened. A character greets the player warmly in a scene after the player killed their brother, because the scene has no access to that fact. Players read this as bad writing every time. It is a state problem — the narrative and the systems had no shared language.

Both are why the graph and the variable contract are the real deliverables. Prose sits on top of them.

Getting into it

The usual routes in are from writing, from design, or from QA — and the QA route is more common than people expect, because testers spend their days finding exactly the two failures above.

What a portfolio needs is not a screenplay. It needs something playable, however small, where a choice visibly changes a later scene, plus an honest note on what you cut and why. A three-scene piece with real state in it says more about the job than forty pages of dialogue.

Tooling matters less than people fear. Any tool that lets you build a graph with conditions will do. What interviewers look for is whether you thought about what the player can miss.