Merge pull request DataDog#788 from DataDog/stephen/intervals

StephenKappel · web-flow · commit 37818ad45179 · 2016-06-22T16:23:03.000-04:00
[anomalies] include time intervals
diff --git a/content/guides/anomalies.md b/content/guides/anomalies.md
@@ -81,11 +81,11 @@ At different zoom levels, the same query can result in time series with very dif
 
 The width of the grey band that is drawn by our anomaly detection algorithm is, in part, based on the noisiness of the time series in the plot. The band must be wide enough that ordinary noise is mostly inside the band and doesn't appear as anomalous. Unfortunately, when the band is wide enough to include ordinary noise, it might also be wide enough to hide some anomalies, especially when viewing short time windows.
 
-Here's a concrete example to illustrate. The `app.requests` metric is noisy but has a constant average value of 8. On one day, there is a 10-minute anomalous period, starting a 9:00, during which the metric has an average value of 10. The chart below shows this series in a graph with a one-day time window.
+Here's a concrete example to illustrate. The `app.requests` metric is noisy but has a constant average value of 8. On one day, there is a 10-minute anomalous period, starting a 9:00, during which the metric has an average value of 10. The chart below shows this series in a graph with a one-day time window; each point in the graph summarizes 5 minutes.
 
 <img src="/static/images/anomalies/disappearing_day.png" style="width:500px; border:1px solid #777777"/>
 
-The grey band here makes sense; it is wide enough to capture the noise in the time series. Yet, it is narrow enough that the anomaly at 9:00 stands out clearly. This next chart shows a zoomed-in view of a half-hour time window that includes the 10-minute anomaly.
+The grey band here makes sense; it is wide enough to capture the noise in the time series. Yet, it is narrow enough that the anomaly at 9:00 stands out clearly. This next chart shows a zoomed-in view of a half-hour time window that includes the 10-minute anomaly; each point in the graph summarizes 10 seconds.
 
 <img src="/static/images/anomalies/disappearing_half_hour.png" style="width:500px; border:1px solid #777777"/>
 
