Changing a plan is not the same thing as abandoning the outcome; often it is how the outcome becomes realistic. With writing a book, separate commitment to what you want from loyalty to one method. The practical question is whether current behaviour is creating better evidence or merely consuming more time.
The signal that justifies another cycle
Persist when the manuscript bottleneck moves: scenes are drafted, structural issues resolve, revision passes close and useful reader feedback becomes more specific. That is a mechanism-level reason to continue. It does not guarantee the final result, but it shows that another cycle can plausibly produce new information or better odds.
Measure manuscript movement appropriate to the phase, not raw word count alone. This keeps a slow but healthy process from being confused with a dead one. A single disappointing week matters less than a pattern measured with the same yardstick.
The signal that says change something
Change the route when planning and research repeatedly replace writing, or the current structure creates the same unresolved problem across many chapters. Do not protect a failing method by calling every contradiction a “test of faith.” A plan is allowed to be wrong even when the desire behind it remains important.
Run a falsifiable test before deciding
Name the present phase—draft, structural edit, line edit or submission—and do one session that can only move that phase. Write in advance what result would count as encouraging, ambiguous or poor. Precommitting prevents you from moving the goalposts after seeing the result.
Keep the control boundary intact
You can protect writing time, define the next scene or section, finish drafts and revise by bottleneck. However, agents, publishers, reviewers and readers make independent decisions later. Persistence should increase the quality of what you control; it cannot create authority over people, institutions, markets or random events that are not yours to command.
Example: what evidence changes the decision?
Deleting 3,000 weak words can be progress in revision. Adding 3,000 more can be avoidance if the central structure is still broken. The important move is to identify which variable changed the conclusion. “I still want it” and “this method is still working” are different statements.
Avoid the sunk-cost trap
A manuscript can hold sunk costs chapter by chapter. Keeping a scene because it took three days to write is not a reader-focused reason. If the book works better without it, the work was still useful: it helped you discover what the draft needed. Revision is allowed to make previous labour invisible.
Use three thresholds, not one mood
Define phase-specific thresholds. Continue drafting while the missing material is the bottleneck; switch to structural work when the full shape exists; revise locally only after larger problems settle. If beta readers repeatedly identify the same confusion, change the structure rather than polishing sentences around it. Review the current bottleneck, not lifetime word count.
Bottom line
For writing a book, persistence is justified by improving evidence, not by stubbornness. Change the plan when repeated data says fit, method or constraints are wrong. You can remain committed to a meaningful outcome while becoming much less attached to the route that first occurred to you.
A decision note worth keeping
Before the next review, write three sentences in CLEAR Planner: what I am trying to learn, what evidence would change my mind, and what cost I refuse to exceed. For writing a book, make the evidence concrete enough that another person could understand it: usable pages, resolved structural problems, completed revision passes and relevant reader feedback. That note becomes the baseline. At the review, compare the new evidence with the note instead of rewriting the standards after the result arrives.
Before abandoning the book, alter the current craft problem: structure, point of view, chapter order, scope or writing schedule. A reversible manuscript change can expose whether the project is wrong or simply stuck in one form.
One final review question
Before another manuscript cycle, name the exact craft question the next session should answer. If the task is merely 'work on the book,' define it more sharply. Persistence should reduce uncertainty in the draft, not only accumulate time beside it.