Creating Validations

Last updated: 2026-03-23

To create a TAR validation test, open the Case Menu, pick a tag, name the test, choose L1 and/or L2, and set your statistical parameters. Hintyr calculates the sample size automatically and starts the grading workflow right away.

TAR Validation dialog - Create tab

Configure a new validation test with all options filled in

Select Tag

Validation Test Name

Validation Type

Confidence Level

Margin of Error

Target Recall

Statistical Method

Accessing the TAR validation dialog

Open the Case Menu and select TAR Validation. The dialog opens with two tabs: Create Validation Test and Continue Test. Select the Create Validation Test tab to configure a new validation.

Selecting a tag for validation

Use the Select Tag dropdown to choose which tag you want to validate. The dropdown lists all tags in the case along with the number of documents in each. The selected tag determines the population from which L1 samples are drawn (documents in the tag) and from which L2 samples are drawn (documents not in the tag).

Make sure your selected tag has enough documents for meaningful statistical sampling. For more on creating and managing tags, see tags.

Naming the TAR validation test

Enter a descriptive name in the Validation Test Name field. This name shows up in the Continue tab and the Grading Panel header, so pick something that clearly identifies the test's purpose. For example, "Responsiveness Validation - Production Set 1" or "Privilege Tag QC Round 2."

Choosing the TAR validation type

Under Validation Type, check one or both options:

  • TAR 1.0 - Control Set Validation - Draws a random sample from documents in the selected tag (responsive documents) to measure precision.
  • TAR 2.0 - Elusion Testing - Draws a random sample from documents not in the selected tag (the discard pile) to measure recall by checking for missed responsive documents.

You need at least one type selected to proceed. When both are checked, Hintyr creates a combined test with samples from both populations presented together in the Grading Panel.

Statistical settings for sample-size calculation

The Settings section lets you configure the statistical parameters for sample-size calculation:

  • Confidence Level - The probability that the sample results reflect the true population metrics. Accepts values from 90% to 99%. The default is 95%.
  • Margin of Error - The acceptable range above and below the measured value. Accepts values from 1% to 10%. The default is 5%.
  • Target Recall - Shown only when TAR 2.0 (Elusion Testing) is selected. The minimum recall percentage your review must achieve to pass validation. Accepts values from 50% to 95%. The default is 75%.

Higher confidence and tighter margins of error mean larger sample sizes and more documents to grade. The defaults of 95% confidence and 5% margin of error are standard for most legal review scenarios.

Starting the TAR validation test

Once you've selected a tag, entered a name, chosen at least one validation type, and configured your statistical settings, click Start Test. Hintyr calculates the sample size, draws random samples, and opens the Grading Panel so you can start grading right away. For details on the grading workflow, see grading samples.

Frequently asked questions

Can I change the settings after starting a test?
No. Statistical parameters are locked once the test begins because changing them would invalidate the sample size calculation. If you need different settings, create a new validation test.
What happens if my tag does not have enough documents?
Hintyr will alert you if the tag population is too small for the requested confidence level and margin of error. You can either lower your statistical requirements or wait until more documents have been tagged.
Can I assign other users to help grade?
Yes. You can optionally assign users to a validation test. Assigned users can access the test from the Continue tab and contribute to the grading work.
What confidence level should I use?
A 95% confidence level with a 5% margin of error is the most common configuration in legal review and is consistent with the proportionality standards of FRCP Rule 26(b)(1). For high-stakes matters where courts require greater certainty, you may increase confidence to 99% at the cost of a larger sample size.

Related articles