The Challenge
Legal teams using Microsoft Purview for discovery face more than just speed issues; they encounter defensibility problems. Searches stall, exports time out, and legal counsel must explain delays to opposing parties and the court.
The key question for your team: if the native tool can't reliably find and preserve necessary data, can you still certify your discovery response under Federal Rule of Civil Procedure 26(g)?
This analysis examines technical limits and court decisions that reveal the costs of these constraints.
Environment and Constraints
Purview operates within Microsoft 365, a multi-tenant SaaS platform. Due to the heavy workload of eDiscovery, the platform imposes restrictions to protect other tenants. Throttling, timeouts, and hard caps are intentional design features, not removable through configuration changes.
Microsoft's published limits clarify these constraints:
- Advanced indexing in Purview eDiscovery Premium is capped at 2 GB per hour, meaning a terabyte takes about three weeks to index.
- Exports are limited to 2 GB per hour per mailbox, with daily caps at the tenant level.
- Processes or exports running beyond seven days are automatically canceled.
- Microsoft doesn't provide a service level agreement for search times.
Capacity is another issue. Microsoft has indicated that demand exceeds supply and prioritizes "highest-value opportunities," with Microsoft 365 Copilot at the forefront. The position of eDiscovery in this hierarchy is unclear, but it's a compliance function with no search-time commitment and fixed throughput limits. Plan for the possibility that it's not prioritized.
Court Findings on Native Tooling
Courts have tested reliance on Microsoft 365 tools in several cases.
Search Reliability: In July 2023, Special Master Philip Favro reported in Deal Genius, LLC v. O2COOL, LLC (N.D. Ill.) that Microsoft 365's limitations might prevent reliable search results. It often doesn't fully index data and lacks user validation for search and production results. These findings impact the reasonable inquiry required by Rule 26(g). Microsoft's documentation still shows items exceeding parsing limits marked as partially indexed.
Methodology: In Kim v. Cushman & Wakefield U.S., Inc. (C.D. Cal. Apr. 24, 2026), the court ruled that keyword searches alone are inadequate for Teams data. A toolset relying on basic keyword searches struggles to meet this standard.
Preservation: In Oakley v. MSG Networks, Inc., 792 F. Supp. 3d 394 (S.D.N.Y. 2025), emails intended to be retained through a Purview Legal Hold were lost. The court noted a technical glitch affected numerous matters. Similarly, in FTC v. Match Group, Inc., 2025 WL 46024 (N.D. Tex. Jan. 7, 2025), an Accessioning system failed to capture certain data, resulting in sanctions despite reasonable reliance on the archive.
An alternative architecture, like X1's, indexes Microsoft 365 data in place, bypassing Purview's search index. It supports advanced queries and allows iterative testing and documentation of search methods before collection. Only potentially responsive data is collected and can be uploaded directly into Relativity for review.
Results and Observations
There are no specific before-and-after metrics for deployments, but the record shows:
- MSG avoided sanctions because the court found its use of Purview reasonable. However, it still lost emails and incurred costs for a forensic investigation.
- Match was sanctioned despite reasonable reliance on its archive, incurring costs for additional depositions.
- Service providers report that searches and exports often stall or time out, causing delays and complicating legal proceedings.
The pattern shows reasonableness can protect from sanctions but not from costs. As documented failures accumulate, claiming ignorance becomes harder because Rule 37(e) considers what a party knew or should have known.
Procedural Corrections
From the MSG and Match outcomes, consider these steps:
- Verify holds independently. Don't assume a Purview Legal Hold is self-proving. Confirm preservation for custodians against an independent sample.
- Document custodian hold history. Keep records of when each Records Freeze started and ended, as the new Purview experience doesn't show past holds.
- Validate before relying. Test and document search methodology before collection, including recall checks on Teams data.
- Plan for delays. Courts may sanction late production even with reasonable tool reliance. Build slack into schedules to accommodate the 2 GB per hour cap.
Takeaways for Your Team
- Treat throttling as a design constraint. Calculate indexing and export timelines from Microsoft's limits before committing to a discovery schedule.
- Test native search against Rule 26(g). If you can't validate results or confirm complete indexing, defending your certification is difficult.
- Don't rely on keyword-only searching for Teams. Use more advanced techniques as expected in Kim.
- Keep your own preservation evidence. Don't depend on the platform's interface to reconstruct hold history.
- Account for AI-driven change. Copilot creates new data types that Purview must preserve and produce. Budget for storage and conversion needs.
- Consider a second-tool strategy. Reducing reliance on a single toolset is a defensibility decision, not just a performance one.
Your Records and Information Management program and your discovery program face the same risks. If the platform can't reliably preserve, search, and export, both programs are compromised.
