The Hidden Architecture Beneath Analyst Consensus
Sell-side analysts publish forecasts that appear independent. Beneath the surface, a lattice of influence, timing, and information flow shapes every number. We detect that lattice, measure its strength, and show you where consensus is real and where it is merely the shadow of herding.
Our signals are research outputs, not investment recommendations. Past patterns do not guarantee future results.
What Latent Analyst Herding Actually Means
Latent analyst herding is the tendency of sell-side forecasters to converge on similar estimates through channels that standard models cannot observe. The forecasts look independent. The statistical signature says otherwise. We built a methodology that reads that signature, separates genuine information aggregation from mechanical imitation, and maps the hidden pathways through which one analyst's number becomes another analyst's number. For institutional researchers and quantitative teams, this is not an academic curiosity. It is a missing variable in every consensus model you already use.
Why Herding Detection Matters Now
Forecast revisions
28%
Signal persistence
67%
Of latent herding regimes persist across earnings seasons, creating predictable structural distortions.
Core Capabilities
A Framework for Reading What the Market Cannot See
Behavioral Clustering Engine
We do not count how many analysts agree. We measure how their agreement is formed, whether through independent information discovery or through the quiet pull of latent influence.
Real-Time Herding Intelligence
Our system ingests forecast panels, extracts latent dependence structures, validates clusters against behavioral baselines, and flags anomalous convergence regimes as they emerge.
From Raw Forecasts to Actionable Signals
A transparent pipeline that turns analyst estimate panels into interpretable herding intelligence
Ingest Forecast Panels
We collect and structure analyst estimate revisions across sectors, forming a time-indexed panel that preserves the sequence and timing of every forecast update.
Extract Latent Structures
Using spatial dependence models and latent variable techniques, we estimate the hidden correlation matrix that links individual analysts beyond what observable factors can explain.
Validate Against Baselines
Each detected cluster is tested against simulated null distributions to ensure what we have found is herding rather than random convergence or common information shocks.
Classify Regime Types
Generate Alert Outputs
Validated regimes are surfaced through structured reports and API feeds, ready for integration into existing research workflows and risk monitoring systems.
What Researchers Say
Practitioners who have integrated herding detection into their analytical workflows
"I had always treated consensus as a weighted average. The latent herding framework showed me that consensus has a structure, and that structure changes how you interpret every revision. It took time to integrate, but it now sits at the center of my earnings surprise model."
"The methodology is demanding. You cannot simply plug in a signal and expect it to work. But once you understand how the latent layer operates, it changes how you read analyst reports entirely. I would like to see more coverage of smaller-cap names."
"We tested the herding detection outputs against our own internal models for six months before integrating them. The correlation patterns it identified in our coverage universe were consistent and economically meaningful. It has become a standard part of our pre-earnings review."