In re Frozen Potato Products Antitrust Litigation, No. 1:24-cv-11801 (N.D. Ill. Oct. 23, 2026), full opinion (PDF)
Parties negotiating an ESI protocol often dispute whether TAR may be used on top of search terms rather than as a substitute. An October 23 order in a consolidated antitrust action in the Northern District of Illinois took up one version of that dispute. The TAR paragraph at issue lets the parties agree on TAR "Instead of relying on a Traditional Search Methodology". The defendants wanted the words "or in addition to" added to that paragraph. Read naturally, the added words would have expressly allowed TAR to supplement, as well as replace, a traditional search methodology. The magistrate judge entered the plaintiffs' version, which omits those words.
The same order permits sampling of images, videos, and similar documents for which, in the order's words, "text-based search technologies are fundamentally ineffective".
What happened
Stacking is the practitioners' term for running TAR only on the documents that search terms have already returned. The order neither defines stacking nor decides whether the practice is permitted. Its stipulated text contemplates a traditional search methodology and/or a TAR methodology. The parties may also modify the order by agreement. Stacking that the parties agree on therefore appears to remain available.
Stipulated TAR provisions also call for null set testing. Once enough documents returned by TAR have been reviewed, the responding party draws a random sample from the documents TAR did not return. It shares the method and results with the requesting party.
The court's analysis
The court entered the plaintiffs' version "to minimize any confusion over whether responsive materials are inadvertently withheld". It noted that the parties had addressed the pros and cons of stacking in great detail. The court considered the defendants' argument for adding "or in addition to" unpersuasive. The order gives no further explanation.
For images, video, and similar documents, the court agreed with the defendants that sampling is more efficient and proportional than individual review, citing Rule 26(b)(1). Once those documents are identified as potentially relevant, the entered provision requires them to be reviewed or sampled for responsiveness. If the responding party considers review overly burdensome, the parties must meet and confer on proportionate methods.
Why it matters
A responding party that wants to run TAR only on search-term hits should describe that workflow expressly in its proposed protocol. It should be ready to answer the court's concern about confusion over whether responsive materials are inadvertently withheld. Under the order's terms, null set testing samples the documents TAR did not return. Documents that search terms culled before TAR ran may fall outside that sample. A stacking proposal should therefore say how the responding party will test the documents its search terms excluded. A requesting party can ask for testing of those excluded documents while the protocol is still being negotiated.
A responding party holding large volumes of images or video can propose sampling in the protocol draft as the proportional alternative to individual review. The entered provision sets no sampling method. Parties who want the method settled in advance should write it into the protocol.
The full opinion is available as a PDF.
