MATCHA's Project History Analysis Explained

How to use the report

Treat Project History Analysis as evidence of the student's work process, not as AI detection. A short or unusual work history can reflect low effort, an atypical workflow, or misconduct; the report cannot decide which. Interpret the metrics together, in light of the assignment, and give students an opportunity to explain unexpected results.

Submission overview

Submission overview fields
Field	Explanation
Group	The group assigned for reading or grading. This is None unless grading groups are being used.
Integrity	OK means the project passed MATCHA's integrity checks: it was produced by an unaltered copy of MATCHA with no more than minimal interference from automation. A failure means that the project may have been tampered with or that another integrity problem was detected. Open Details for the specific error. If the cause remains unclear, send the original project to technical support for assessment.
History-deleted notice	If this notice appears, the detailed edit and browsing history has been removed. The notice states whether the displayed analysis comes from a retained certified snapshot or whether only the current document remains.

Writing analysis

Unless a row says otherwise, these values combine the main document and notes. Character-source percentages use the current character count as their denominator.

Writing analysis metrics
Metric	Explanation	Reference point
Active writing time	The estimated time spent actively writing. MATCHA subtracts idle gaps, so time spent looking at the text without typing, scrolling, or otherwise interacting with the computer does not count. Some DOCX reports call this Editing time .	Compare with the active work expected for the assignment.
Typed characters	The number of characters entered by typing, as opposed to AI input or pasting from any source. This includes typed characters that were later deleted.	Assignment-dependent; compare with the final length and expected drafting process.
Deleted typed characters	Typed characters that were eventually deleted. This does not count deleted characters that originally came from a paste, including text copied within the document.	Use Character churn below for a length-adjusted comparison.
Pasted from browser	Characters still present that were copied from a source in MATCHA's built-in browser, shown as a count and a share of the current text.	Assignment-dependent; source-based work will normally produce some.
MATCHA AI-generated characters	Characters still present that were inserted by MATCHA's built-in AI assistant, shown as a count and a share of the current text.	0 when AI assistance was off; otherwise compare with the permitted assistance level.
Pasted from system	Characters still present that were pasted through the system clipboard from outside MATCHA's built-in browser, shown as a count and a share of the current text.	Usually 0 when External-Paste Block was enabled; review any listed exceptions.
Current character count	The characters currently remaining in the main document and notes.	Compare cautiously with length requirements: this includes notes and is measured in characters, not words.
Time fragmentation	A measure of nonlinear writing and revision. MATCHA counts adjacent final-text positions whose characters were entered more than one minute apart, then scales the result so a raw fragmented share of 10% or more is 100%. Higher values normally indicate more revisiting and revision.	10–30%, depending on the activity.
Character churn	The percentage of typed characters that were subsequently deleted. Higher values indicate more revision.	Well-edited essays often have at least 30–40% churn.
Offline-certified insertions	Characters entered through an authorized, cryptographically certified offline or outage-writing path. This row appears only when such insertions are present.	Normally absent; if present, reconcile it with an authorized outage.

Typed characters by author

In a project with more than one author, this breakdown attributes typed characters to each author and shows each person's share of all typing. Pasted and AI-generated characters are not counted as typing.

Writing graph

The x-axis is writing time and the y-axis is the text's final location in the main document. A document written from beginning to end without breaks or editing back will look like a straight line from top-left to bottom-right. A well-edited document will show scattered edits throughout the graph. Copy-paste blocks appear as vertical lines. The graph does not include text that ends in the Notes pane, but it reflects the original writing time of content copied from Notes.

Policy violations

This expandable list records events such as loss of focus or a detected overlay during a restricted activity. The description supplies event-specific context. A policy event is not automatically misconduct; consider its duration, the activity rules, accommodations or exemptions, and the student's explanation.

Reading analysis

Reading Analysis is available when Complete Work History was enabled. Recorded visibility and activity support an estimate of reading; they cannot establish comprehension.

Reading analysis metrics
Metric	Explanation	Reference point
Browsing time	The estimated active time in MATCHA's built-in browser, with idle gaps subtracted.	Adapt to the volume and difficulty of the assigned reading.
Referenced works read	The estimated share of material in cited works that was read in MATCHA's built-in browser. This is unavailable when the project has no references or no Complete Work History data.	Higher coverage is expected when students must read the works they cite; adapt to the assignment.
Reading time per word written	Active reading time in seconds divided by the number of words in the main document.	For a typical essay, about 7 seconds per word—roughly two hours per 1,000 words. Adapt this to the assignment.
Automation events	The number of detected automation signals in the built-in browser. This row appears only when automation was detected.	0. A non-zero value warrants contextual review but is not proof of misconduct.
Flagged browsing ticks	Periodic reading samples carrying an automation flag, shown as a count and as a percentage of all samples. This row appears only when automation was detected.	0. Review any non-zero value with the rest of the work history.

Reading graphs and detailed list

In Referenced documents , arc size represents document length and the filled share represents how much MATCHA estimates was read.
In Other documents , arc size represents words read. These arcs are fully filled because their purpose is to compare reading volume, not coverage of an assigned or cited work.
The categories separate read cited works, unread cited works, other read works, and other pages. Select a graph segment or category to inspect it.
The detailed list reports each work's pages, words, estimated words read, percentage read, and active time. Its optional filters can hide short web pages (under 500 words) and documents seen for less than one minute. In review mode, select a work to open the recorded source when it is available.
Certified DOCX reports use a compact list of cited works, non-cited publications, and other pages, and omit documents with less than one minute of reading time.

Reference comparisons and flags

When enough comparable submissions are available, MATCHA compares Time fragmentation, Character churn, Referenced works read, and Reading time per word written with submissions from similar activities across your courses. At least 30 comparable submissions are required.

No lower-tail flag: the value is not unusually low relative to the comparison group.
Yellow lower-tail flag: the value is unusually low and deserves contextual review.
Red lower-tail flag: the value is still farther into the low end of the comparison group.
Comparison unavailable: MATCHA lacks enough comparable submissions, the metric itself is unavailable, or the activity category cannot be established.
The comparison details identify the baseline sample size ( n ), its mean, its sample standard deviation ( SD ), and the value at or below which each flag is applied.

Lower-tail flags are advisory statistical indicators, not proof of misconduct. They should direct attention, not determine a grade by themselves.

Settings on certified DOCX cover pages

A certified DOCX cover page also records the activity configuration under Features . A check mark means the setting was enabled; it does not describe an action taken by the student.

Certified DOCX settings
Setting	Explanation
External-Paste Block	Whether pasting from the system clipboard was blocked for the activity.
Full Work History	The DOCX label for Complete Work History . When enabled, MATCHA records browser reading data for the Reading Analysis.
Locked-Down Mode	Whether the activity required MATCHA to remain fullscreen and in focus.
AI assistance	The permitted level of MATCHA's built-in assistant: Off, Grammar only, Grammar and style, or Full composition. Use the AI-generated character count above to see how much assistant-generated text remains.
Pedagogical mode	When On, the assistant guides the student to make grammar or style corrections instead of applying all corrections automatically.
Notes	Lists exceptions or unusual conditions that actually applied, such as disabled protections, an overlay exemption, an assistive-technology accommodation, or an authorized outage. Review these before interpreting a policy event or an unusual writing history.