FREE TOOL

Outlier & Anomaly Detector

Paste your daily metrics and instantly spot statistical outliers with possible explanations.

Input Your Data

Paste daily metrics as date,value per line (e.g., 2025-01-15,1234) or just values one per line. CSV with headers also works.



Method:

What Are Outliers in Analytics Data?

An outlier is a data point that significantly deviates from the expected pattern. In marketing analytics, outliers can signal real events (a viral post, a bot attack, a tracking failure) or natural variance. The key is distinguishing signal from noise.

This tool uses statistical methods to flag data points that fall outside expected bounds, then suggests possible causes based on the deviation pattern. All calculations happen in your browser \u2014 no data is sent to any server.

How Outlier Detection Works

IQR MethodUses the interquartile range (Q3-Q1). Points beyond 1.5x IQR from quartiles are flagged. Robust against extreme values.
Z-Score MethodMeasures how many standard deviations a point is from the mean. 2 SD catches ~5% of normal data; 3 SD catches ~0.3%.
Modified Z-Score (MAD)Uses Median Absolute Deviation instead of standard deviation. More robust than Z-Score for non-normal distributions.
Context MattersStatistics flag candidates. You decide whether a spike is a tracking bug to fix or a campaign win to celebrate.

Best Practices

\u2713 DO
  • Use at least 14 days of data for meaningful statistical bounds
  • Check if outliers coincide with known events (launches, outages, holidays)
  • Try multiple detection methods to cross-validate findings
  • Investigate high outliers AND low outliers \u2014 drops often matter more
  • Run detection weekly as part of your analytics QA routine
\u2717 DON’T
  • Automatically remove outliers without understanding why they occurred
  • Use fewer than 7 data points \u2014 statistics need minimum sample size
  • Ignore seasonality (weekday/weekend patterns will create false positives)
  • Assume all outliers are errors \u2014 some are genuine performance changes
  • Run outlier detection on already-filtered or aggregated data

Frequently Asked Questions

Start with IQR \u2014 it’s the most robust for marketing data which often has non-normal distributions. Use Z-Score when your data is roughly bell-shaped (like daily session counts). Use Modified Z-Score (MAD) when you suspect your data has fat tails or extreme skew.

Minimum 7 for basic detection, 14+ for reliable results, 30+ for high confidence. Daily data for one month is ideal. Weekly data works but needs 3+ months of history to be meaningful.

B2B sites typically see 40-60% traffic drops on weekends. If you analyze a full week, every Saturday and Sunday will flag as a low outlier. Solution: analyze weekdays and weekends separately, or use 7-day rolling averages.

In practice, they’re used interchangeably. Technically, an outlier is a statistical deviation from the distribution, while an anomaly implies something unusual happened. All anomalies are outliers, but not all outliers are anomalies \u2014 some are just natural variance.

This tool is designed for point anomalies (single-day spikes or drops). For gradual trend shifts, you’d need change-point detection algorithms. However, if a gradual decline accelerates past the statistical bounds, this tool will catch the acceleration point.

Your data never leaves your browser. All statistical calculations happen client-side in JavaScript. Nothing is sent to any server, stored in any database, or shared with anyone. You can verify this by checking the network tab in your browser’s developer tools.