Why this article exists
When a user downloads elevation data from the Elevation tab, Portal offers the option to exclude outlier data. This article explains what "outlier" means, how Portal decides a reading is one, and what excluding them does and does not change. Use it to answer customer questions and to keep our own explanation consistent.
Short answer
An outlier is a single elevation reading that sits far enough away from the rest of the readings that it's almost certainly noise — a bad measurement, not real slab condition. Portal identifies outliers using the interquartile range (IQR) with Tukey's 1.5×IQR rule, and can drop them so one bad value doesn't distort the results.
How Portal defines an outlier
Portal does not treat "outside the interquartile range" literally — that would flag the entire top and bottom quarters of the data (half of every dataset), which is not the intent. Instead it uses the standard statistical method built on the IQR:
| Term | What it is |
|---|---|
| Q1 | 25th percentile — 25% of readings fall below it |
| Q3 | 75th percentile — 75% of readings fall below it |
| IQR | Q3 − Q1 (the spread of the middle 50% of readings) |
| Lower fence | Q1 − (1.5 × IQR) |
| Upper fence | Q3 + (1.5 × IQR) |
A reading is an outlier if it falls below the lower fence or above the upper fence. Everything between the fences is kept. This is why the release notes describe filtering as removing "a small number of extreme points" — in typical data, only a handful of readings ever land outside the fences.
Worked example
A FieldPrinter measures a slab and reports each reading as a delta from the selected reference (high / low / mid point of the scan).
Across the measured area, the readings produce Q1 = −1/16" (−0.063 in) and Q3 = +3/32" (+0.094 in) — a floor that's mostly flat with typical slab variation.
- IQR = 0.094 − (−0.063) = 5/32" (0.157 in)
- Lower fence = −0.063 − (1.5 × 0.157) = −19/64" (−0.298 in)
- Upper fence = 0.094 + (1.5 × 0.157) = +21/64" (+0.329 in)
A reading of +3/16" (+0.188 in) — a real high spot a finisher would want to know about — is kept, because it's inside the fences. A reading of +1 1/8" (+1.125 in) is flagged as an outlier and dropped: no slab is a full inch proud over a single grid square, so it's almost certainly the FieldPrinter passing over rebar, a chair, debris, or a curb edge rather than the slab itself.
Outliers are computed per dataset, and point elevation data uses the same computation as area/heatmap data.
What the "exclude outlier data" option does
| Behavior | Detail |
|---|---|
| Where it appears | An Exclude outliers checkbox at the download step in the Elevation tab, available for both the floor (area/heatmap) and point elevation tabs. |
| What it removes | Only readings that fall outside the 1.5×IQR fences. |
| What it keeps | All readings between the fences — i.e., all normal slab variation, including genuine high and low spots. |
| Effect on the file | Excluded rows are deleted from the downloaded file. The raw data is not lost — to get the full unfiltered dataset, download the file again with Exclude outliers unchecked. |
| Effect on the heatmap | Removing extremes tightens the heatmap's color range so real variation reads more clearly instead of being washed out by one bad value. |