Methodology1 min read
Data quality · how gaps delays and outliers quietly break strategies
However good the strategy, bad data makes it all wrong. The three most common data problems, what they cause, and how a system should handle them.
By Ruleap Star Team · Published
In a quant system, data is where everything starts. Analysis, signals and risk all rest on the assumption that the data is correct. When that assumption breaks, every layer downstream goes wrong too — often without anyone noticing.
Problem 1: gaps
Data for a period never arrives — a dropped connection, or a source going offline. If the system silently fills in the previous value, indicators look calm while the system is actually reasoning about a market that never existed.
Problem 2: delays
The data arrives, but late. Delays make the system use “past prices” for “present decisions,” which is especially dangerous in fast markets. Timestamp alignment and latency monitoring are basics of the data layer.
Problem 3: outliers
A single data point jumps unreasonably. It might be a source error, or a genuinely extreme move. A system can rarely tell which in the moment, so the more robust approach is to flag it and hold off on decisions that depend on it — rather than trusting it outright or deleting it.
Handling principles
- Flag explicitly, don't patch silently: downstream should know when data is suspect.
- Better to do nothing than to act on bad data: risk control can pause new signals during data anomalies.
- Log everything: which data point, when, and how it was handled should all be traceable later.
This article explains a method and is not investment advice.