The first time you run Analyze Document on a chapter, the screen fills with colour and a row of percentages. Some are red, some are white, and it is not obvious what they mean or what to do about them. This guide explains each number, where the guide levels come from, what to do about a high or a low reading, why the tool is a prompt and not a verdict, and how to make it part of a routine. It is written for Quillen, but the thinking applies to any tool that counts words.
The short answer
Each chip in the stats bar shows what share of the words in the scene belong to one category, next to a small guide level. A number in red is above the guide, and a number in white is at or below it. The guide levels are rules of thumb for general prose. They tell you where to look, not what to change. Treat a red chip as "check these underlined words", not "this writing is bad".
What the stats bar shows
Click Analyze Document in the editor toolbar. A bar appears above the page, with one chip per category: a coloured dot, the category name, the percentage of your words in that category, and a small "<" figure showing the guide level. Every instance is underlined on the page in the same colour. Fiction books show nine chips. Nonfiction books swap two of them, as the table below shows.
| Category | Colour | Guide | What it counts | Applies to |
|---|---|---|---|---|
| Filter | Red | 4% | saw, heard, felt, noticed, realized, wondered, knew and similar | Fiction |
| Passive | Purple | 3% | a form of 'to be' followed by a word ending -ed or -en | Both |
| Vague | Blue | 8% | forms of 'to be', very, really, just, good, bad, nice, big, small | Both |
| Crutch | Green | 4% | a little bit, seemed to, kind of, sort of, quite, almost, slightly | Both |
| POV | Orange | 2% | telling phrases such as took a sip, walked across, stood up | Fiction |
| Adverb | Pink | 2% | -ly words, with an exception list | Both |
| Abstract | Amber | 1.5% | something, everything, nothing, thing, stuff, way, lot, someone | Both |
| Cliché | Indigo | 0.5% | common stock phrases | Both |
| Conjunction | Grey | 1% | sentences starting with and, but, so, yet, because, or | Both |
| Hedging | Teal | 1.5% | arguably, perhaps, it could be argued | Nonfiction |
| Wordiness | Brown | 1.5% | due to the fact that, in order to, at this point in time | Nonfiction |
Reading a chip: three examples
Here is what the numbers looked like on a short, deliberately flawed sample scene of about 140 words:
- Filter 6.6% against 4%. Over the guide. The scene has filter verbs (felt, realized, noticed, wondered, saw, heard, knew) in nearly every sentence. Worth a look.
- Passive 1.5% against 3%. Under the guide. No action needed, though the tool only matches words ending in -ed or -en, so it misses passives with irregular verbs such as "was found".
- Cliché 5.1% against 0.5%. Far over the guide. On a sample this short, a few phrases (heart raced, breath caught, dead of night) add up to a large share. In a real chapter this number is usually small.
Why the numbers move with scene length
Percentages on a short selection swing wildly: one cliché in a 100-word scene is 1% of the words, but in a 2,000-word scene it is 0.05%. Analyze a whole scene or chapter before drawing conclusions. A tiny excerpt will often look worse than the same writing in context.
What the guide levels are, and are not
The levels are starting points taken from common self-editing advice, not scientific norms. Different books deserve different readings:
- A hard-boiled thriller may run high on short sentences starting with "and" or "but".
- A literary novel in a distant narrator's voice may run high on filter words and still be deliberate.
- Dialogue-heavy scenes may run high on vague words, because people speak vaguely.
- A nonfiction book argues differently from a novel, which is why the app's nonfiction mode counts hedging and wordiness instead of filter and POV.
Limits of a word-matching tool
Analyze Document finds patterns of words. It does not understand your story. Three things follow:
| Limit | Example | What to do |
|---|---|---|
| False positives | 'Was tired' flagged as passive because 'tired' looks like a participle | Ignore the highlight |
| False negatives | 'Was found' not flagged because 'found' has no -ed or -en | Read for passives yourself |
| No sense of intent | A deliberate cliché in a character's dialogue is still flagged | Decide on purpose |
| No sense of quality | A clean page can still be dull | Use it with reading and listening |
A routine that works
- Finish your structural edit first. Do not analyze a scene that may be cut.
- Analyze one scene or chapter at a time.
- Pick the one or two chips furthest above their guide, and work on those first.
- Use the underlines as prompts: check each, fix those that need it.
- Rerun the analysis once, then stop. Do not chase zero.
- Finish with a listening edit.
Common mistakes
| Mistake | Problem | Fix |
|---|---|---|
| Chasing every number to zero | Flat, strained prose | Use the guide as a prompt, not a target |
| Analyzing a short excerpt | Percentages swing wildly | Analyze whole scenes or chapters |
| Editing the numbers instead of the writing | Swapping words to dodge the list | Fix the sentence, not the statistic |
| Running it before the structural edit | Polishing text that gets cut | Use it in the sentence pass |
| Treating a clean score as a good chapter | A tool cannot judge story | Read and listen too |
Do it in Quillen
- Open a scene in the editor.
- Click Analyze Document. The toolbar button changes to Clear Highlights, and the stats bar appears.
- Read the chips. Each shows the category, your percentage and the guide level. Red is over the guide.
- Look at the page. Every flagged word is underlined in its category's colour. Overlapping underlines mean a word belongs to two categories, as "was sealed" does in Passive and Vague.
- Edit directly on the page. Highlights stay on while you type. Click Analyze Document again if you want fresh numbers.
- Click Clear Highlights to return to a clean page.
- Switch book type if needed. A nonfiction book shows Hedging and Wordiness in place of Filter and POV.
The categories each have their own guide: filter words, passive voice, adverbs, clichés, vague words and telling.
Checklist
- Analyzed a whole scene or chapter, not a snippet.
- Looked first at the chips furthest above their guide.
- Checked each underline before changing it.
- Left deliberate choices alone.
- Reran once, did not chase zero.
- Read or listened to the result.
Common questions
What do the red numbers in Analyze Document mean?
A red number means that category's share of your words is above the guide level. It is a prompt to look at the underlined words, not a sign the writing is bad.
Where do the guide levels come from?
They are rules of thumb based on common self-editing advice, not scientific norms. Different genres and voices can legitimately run higher or lower.
Why do my percentages look so high on a short scene?
Percentages swing on short text: one cliché in a 100-word scene is 1% of the words. Analyze a whole scene or chapter.
Does Analyze Document change my text?
No. It underlines words and counts them. You make every edit.
What is different for nonfiction books?
Nonfiction books count Hedging and Wordiness instead of Filter and POV. The other categories are the same.