A biotechnology company engaged Sandline to support a complex, high-stakes litigation matter requiring defensible document review at scale. With growing volumes of electronically stored information (ESI) and increasing pressure to manage outside counsel spend, the client sought a long-term partner capable of delivering both consistency and efficiency throughout the lifecycle of the matter.
This engagement involved multiple phases of discovery and document review over time. Sandline had already been conducting ongoing review workflows for the matter, supporting responsiveness and quality control efforts as the case progressed.
As the review entered a later phase and approached final production, the client faced a familiar challenge:
- Large remaining document populations still required assessment
- Manual attorney review would significantly increase cost
- Timelines demanded accelerated workflows
- Defensibility and recall could not be compromised
The client needed a way to reduce remaining review volume while maintaining rigorous legal and statistical standards.
Building on the foundation of prior review work already completed in the matter, Sandline introduced Continuous Active Learning (CAL), a form of technology-assisted review, to support defensible culling, efficient workflows and drive cost savings.
Sandline paired advanced machine learning with attorney-led review consulting and oversight. The CAL model was iteratively trained using reviewer coding decisions accumulated throughout the review effort, allowing it to rapidly distinguish relevant from non-relevant documents across the eligible population.
CAL was used specifically to:
- Identify documents highly unlikely to be relevant
- Exclude low-value material from manual attorney review
- Preserve defensibility through transparent statistical validation
- Drive measurable cost savings in a late-stage review phase
Throughout the CAL process, Sandline continuously monitored model health, training stability, and performance indicators. Once training was complete, Sandline conducted a statistical validation against the full TAR-eligible universe of 23,597 documents.
Validation metrics were reported with:
- Point estimates (most likely outcome)
- Confidence intervals (conservative defensible range)
This approach ensured transparency and provided the client with clear, court-defensible risk boundaries when using TAR for culling.
Review Volume Reduction
Total TAR-eligible universe in this review set: 23,597 documents
Documents excluded from attorney review: 7,866
Percentage of review avoided: 33.3%
By eliminating thousands of documents from manual responsiveness review, Sandline delivered meaningful cost savings during a critical phase of the matter.
Elusion Rate (Primary Risk Measure)
0.8% (confidence interval: 0.32% – 1.47%)
The elusion rate estimates the percentage of relevant documents that may remain in the set predicted as non-relevant. An elusion rate well below 2% is widely considered best-practice for defensible culling—making this result exceptional.
Precision
99.7% (99.61% – 99.86%)
The model was extremely accurate when identifying relevant documents.
Recall
99.3% (98.61% – 99.69%)
Nearly all relevant content in the dataset was captured.
Richness
40.2%
Despite a relatively high proportion of relevant material—typically a challenge for effective culling—the model maintained outstanding performance.
By integrating CAL into an ongoing, multi-phase review effort, Sandline enabled the client to:
- Reduce late-stage review volume by one-third
- Avoid unnecessary attorney hours and outside counsel costs
- Maintain statistically validated recall above 99%
- Implement a transparent, defensible TAR workflow
- Accelerate production readiness while protecting quality
Sandline’s CAL-driven approach delivered meaningful efficiency and cost savings for a large biotech company during a later phase of a long-running litigation review. With low estimated risk, exceptional accuracy, and a 33% reduction in review volume, this engagement demonstrates how Sandline helps legal teams escape the “quality rollercoaster” and maintain consistent control over legal spend across the full matter lifecycle.
