How the Forage Classification Pages Fit Together
8 min read · Last updated · Page version 2
Where content classification is enabled on a deployment.
The Classification group in the Records menu has seven pages: Overview, Search by Label, Labels, Suggested Labels, Sensitivity Labels, Files to Protect and Content Rules. They look like seven separate tools, but they are one system seen from seven sides. This page explains how a file gets a label, what each page is for and how the pages feed each other.
How a file gets a label
A label says what a file contains. "Bank details", "Payment card" and "Purchase order" are examples. A label has a name, a category, a description and a severity from Routine up to Highest. A label does nothing on its own: creating one puts it on no file.
A rule says when to give a label. For example: "if a file holds a 16-digit number that passes the card number check, with the word card nearby, give it the Payment card label". Every rule gives one label, but several rules can give the same label. A rule is saved in versions, so you can improve it without losing the old wording, and only one version of a rule is active at a time.
Forage checks every file against the active rules. Forage reads each file and keeps its text. A part of Forage called the rule checker then compares every file that has text with every active rule. When a rule matches, the file gets that rule's label, and Forage records which rule version gave it, when, and what it found. A file can carry several labels. The AI assistant never labels a file: it only suggests, and a person has to switch a rule on.
When rules change, files are checked again. Activating or retiring a rule changes the set that files are compared with, so every file is queued to be checked against the new set. The Files checked figure drops at once and then climbs back. Nothing is lost: labels already on files stay until the file is checked again.
The flow, in plain steps:
- Content Rules decide what gets a label. A person writes, tests and switches them on.
- The rule checker compares each file with the active rules. This happens on its own.
- Overview shows how far that checking has got.
- Labels and Search by Label show what was found: which files, where, and why.
- Sensitivity Labels compares the result with Microsoft 365, so files labelled "Bank details" can be checked against your company's Confidential label there. Where labelling has been switched on, it can also set that label on the files.
- Suggested Labels and Files to Protect are where new labels and rules come from. Both use the AI assistant to propose, and both end in a draft that a person tests in Content Rules.
The seven pages at a glance
| Page | What it is for | Go there when |
|---|---|---|
| Overview | A summary of how much checking is done and what was found | You want to know whether checking is running and how much is labelled |
| Search by Label | Finding files by label, with or without words | You need the actual files that carry a label |
| Labels | Managing labels, and seeing each label's files with the reason and strength of each match | You want to create or edit a label, or see why a file has one |
| Suggested Labels | Labels the AI assistant proposes from the kinds of record you hold | You want labels for your own kinds of record, such as purchase orders |
| Sensitivity Labels | Choosing which Microsoft 365 sensitivity label the files of each Forage label should carry, checking what they carry now, and applying it where labelling is switched on | You want to compare Forage's findings with the labels your company publishes in Microsoft 365 |
| Files to Protect | Finding files that look like your protected files but carry no label | You suspect sensitive files are unlabelled in Microsoft 365 |
| Content Rules | Writing, testing and switching on the rules that give labels | You want a label given to files, or a rule labels too much or too little |
Common tasks
Find every file with bank details. Open Search by Label, select the Bank details chip and press Search. Only checked files carry a label, so look at the Overview banner to see how much is checked. For a full list to hand on, open Reports, choose Labels by file, then export a PDF or CSV.
See why a file has a label. Open Labels, select the label, select the file, and read Why it has this label at the top of its details. See what a label or rule found.
Check whether a rule is labelling too much.
- On Labels, open the rule's files and look at the Strength column. Look at the Weak matches first.
- On Content Rules, press Test run and check that the sample finds the right things.
- Press Check accuracy and mark the 25 files Right or Wrong.
- If it labels too much, press Edit as new version. Add a number check, require nearby words, or add words that rule a file out, then Test run again.
- When the new version looks right, activate it. The old version is retired and files are checked again.
Turn a suggested label into a working rule.
- On Suggested Labels, look through a suggestion's example files, especially if it says "Check the examples". Press Edit if the words need tuning, then Accept (or Use existing label).
- On Content Rules, select the new draft rule and press Test run.
- If the sample is right, press Activate and confirm.
- Watch Overview while Forage checks again, then find the files on Labels or Search by Label.
Compare Forage's findings with your Microsoft 365 sensitivity labels. On Sensitivity Labels, press Choose on a Forage label and pick the sensitivity label its files should carry, then press Check, then Show. Look through the dry run and open a few files. Export CSV gives a list for people to work from. Where labelling has been switched on, Apply labels sets the label on the files; see check and apply sensitivity labels.
Act on Files to Protect. Press Look for files and choose the label. Open the High confidence patterns first and answer Protect or Not sensitive on the strongest matches. Press Review a sample so people who know the files give a trustworthy measure. Then export the list for your security team, or press Make a label on a pattern that deserves its own rule, and test that rule on Content Rules.
Words that are easy to mix up
Label, rule and rule version. A label names what a file contains. A rule is the instruction that gives a label. A rule version is one saved wording of that rule. You change how files are found by making a new rule version, not by editing the label.
Forage label, Microsoft 365 sensitivity label and retention label. A Forage label records what Forage found inside a file and lives in Forage. A Microsoft 365 sensitivity label, such as Confidential, is your company's own decision about how a file is handled, and lives in Microsoft 365. A retention label says how long Microsoft 365 must keep a file. A Forage label never decides how long a file is kept.
Active, draft and retired. A rule version is active when files are checked against it now. A draft is saved but has never been active, and labels nothing. A retired version has been stopped, and keeps the labels it gave. A retired label keeps its files, and no rule can use it until it is reinstated.
Checked and searchable. A file can be found by its name once Forage has listed it, and by its words once Forage has read it. A file is checked when Forage has compared its text with the rules active right now, which needs text. Checked falls when rules change, because files must be compared again. Searchable does not fall.
Evidence strength and a rule's accuracy check. These are two separate measures. Evidence strength (Strong, Medium, Weak or Not rated) describes one match in one file, and Forage rates it automatically. A rule's accuracy check is a person's judgement of a whole rule version: a person marks up to 25 random files Right, Wrong or Not sure, and Forage reports the share that were right. Use strength to find doubtful matches. Use the accuracy check to decide whether a rule is good enough to keep. See labels and rules in plain words.
See what each page shows for the rest of the Records section, and product features for how each feature behaves.
Questions go to sales@smikar.com.