Relationship tracking for large casts

The cast creeps up on you. Book one needed five characters; by chapter two hundred there are forty with names, and the question in every planning session shifts from "who is this" to "what does she think of him again". Two failures follow. Either the relations flatten into a generic friendliness, or they contradict themselves: sworn enemies sharing warmth, a betrayal nobody reacted to, trust that reset between chapters.

The fix is to treat relationships as records in their own right, not as adjectives inside character profiles. Three properties do most of the work: direction, history, and shape.

Relationships are directed

"Kai trusts Mia" does not imply "Mia trusts Kai". Unrequited love, one-sided rivalry, loyalty a retainer owes that the lord does not return: half the texture in a relationship web is asymmetry, so record each direction as its own fact, from one character to the other. When a bond is genuinely mutual, that is worth recording too, as its own two-sided entry. Whenever you catch yourself writing "they are friends", ask which one of them would say it out loud first.

Relationships are historical

A relationship is a state that changes over time, so it gets the same rule as every other state: date each change, keep the history. Rival in chapter one, enemy in chapter thirty, ally in chapter seventy: each of those stances was canon for the chapters under it, and the sequence is the arc. A record holding only "ally" cannot tell you whether they were already enemies when the border incident happened, which is exactly the question chapter forty-five is about to ask. Alongside each change, note what caused the turn; the causes are the beats of the arc.

The shape is a graph

Cast relationships form a web, not a list, and the web's shape carries information a list cannot:

• Clusters: who is connected to whom, which camps exist, and who moves between them.
• Bridges: the single character linking two groups. Move or kill her and the camps stop talking, which makes her structurally important however minor she looks.
• Orphans: characters tied to nothing, usually the ones readers cannot place when they reappear.

You do not need artistry to get this. Names and arrows on paper work. What matters is that every named character is on the graph, and every edge carries its history.

Tier the cast

Tracking everything equally is how bibles get abandoned. Tier instead:

• The core (three to seven characters): full relationship records, every change dated, caused and evidenced.
• Recurring characters (roughly the next ten to twenty): a line each in the graph, current stance plus since-when.
• Background names: a name and an affiliation. Do not track feelings for a character who appears twice; inventing depth you will never use is how the record goes stale.

What one record holds

• From and to, and the stance in the characters' own terms. Not "negative" but "blames him for the fire".
• Since when: the chapter of the current stance, and the chapter of every previous one.
• What turned it: the event behind each change. This is the beat list of the arc.
• Evidence: where the current stance was established on the page, so you can check the scene instead of your memory.
• A pressure note: what would plausibly move this relationship next. Half planning, half reminder.

Writing with the web

Two habits convert the record into craft. Before any scene with three or more named characters, read the edges in the room: current stances, last changes, and any asymmetries, which is where the scene's tension can live. And when planning a long arc, read one character's edges in order; that sequence is her relational arc, and if it has no turns, the graph just told you the character is standing still.

The failure modes to watch

• The static web: nothing changes for fifty chapters. The history view shows a flat line where the story promised an arc.
• The unmotivated turn: the stance changed but no cause was recorded, which usually means the reader never saw the turn either. That is the relationship contradiction, one of the classic continuity errors.
• The disappearing bridge: a minor character you moved or killed turns out to have been the only link between two camps, and scenes that need both camps stop being reachable.

How Kanonara approaches this

Relationships are their own record type in Kanonara, drawn as a graph of the cast, and each edge keeps its full history instead of a single current value. When chapters are analyzed, changed relationships come back as proposals with the evidence for the change, and nothing enters the record until the author approves it. See how Kanonara works.