| Product(s): |
WaterSight |
| Version(s): |
10.00. |
| Area: |
Documentation |
Problem Overview
In case the patterns or forecasts don't display any confidence bands at all (no grey shadow just a thin grey line) or in case it starts to be a straight line (see example below), this means WaterSight found no historical data (or very few) available to calculate the expected trend.

Figure 1 - Pattern represented with a straight grey line (and no percentile bands) for the 24th and 25th Jan.
This can happen more frequently when using the last 1 month pattern visualization and for weekend days (as when calculating Saturday or Sunday patterns WaterSight only looks back at past Saturday or Sundays, while when calculating week patterns WaterSight looks back, by default, at all week days, which are much more).
These strange patterns can also negatively impact event generation, by generating false events.
Solution
Suggestions to avoid or fix the issue:
- Make sure that the sensor or zone has some historical data during last months (at least one full week of data is required)
- Look at past events during last months (for the zone or for the sensors belonging to the zone) and check for longer duration events that might have the option "exclude from pattern" set to yes, and update to No (unless they are real burst events or other network interventions). Those periods that cover events with the option "exclude from pattern set to yes" will be ignored for the pattern calculation;
- Go the the alert trigger definition (under Admin >> alerts) and check that the alert for that sensor or zone is activated with the option "Exclude from pattern" set to NO (under Predefined settings);
- Typically is good practice to use historical data of last 2 months or last 3 moths for the alert trigger instead of just last 1 month – this way WaterSight has more historical data available to calculate patterns, even if sensor fails or there are some events with the option of excluding from pattern set to Yes.
See Also
Assuring Cleaned Patterns
Zone Gap Filing: Estimating missing or wrong data
Alerts Administration
Patterns and Demand Forecast calculation
Near Real Time Anomaly Detection
OpenFlows WaterSight TechNotes and FAQ's
WaterSight - Learning Resources Guide