Peers
Group each fund with genuinely similar funds and spot outliers.
Peers automatically groups each fund with a cluster of genuinely similar funds, based on how they perform, what they hold, and how they're structured. Peer behavior then becomes a yardstick for spotting outliers. It answers: is this fund behaving in line with others like it, or is it an outlier?
How it works
It groups each portfolio with a set of genuinely similar funds, then checks whether a fund behaves unlike its group. From a large dataset of thousands of portfolios it builds features describing past performance (return, volatility, drawdown), composition and turnover (exposure by country, sector, asset class, rating), and static traits (fees, strategy), then reduces the data and clusters it to assign each portfolio to a peer group. Clusters are rebuilt periodically while each portfolio is re-evaluated daily, and a fund is flagged when its returns or exposures fall outside the normal range derived from its peers.
What you get
- A peer-group assignment for each fund, refreshed as the fund evolves.
- Peer groups built from performance, composition and turnover, and static traits such as fees and strategy (proprietary, not an off-the-shelf classification).
- Daily expected ranges for fund metrics derived from peer behavior.
- Anomaly flags when a fund's returns or exposures fall outside its peers' normal range.
How to use it in Spark
The peer groups and expected ranges arrive as data in your workspace. Compare a fund against its peers on a canvas, and let a workflow or the Agent flag funds that fall outside the peer range.