Data quality

Two checks run alongside review to surface records worth a closer look: one for where a record was collected, one for the numbers it holds.

Location check

When a record has a GPS point and is bound to a site, CensusIO compares the two and gives a verdict in the Location check column:

On station: the point is within the site’s boundary, or close to its center.
Near: the point is a little way off.
Off station: the point is well away from the expected site.

Open the cell to see the comparison on a map: the collected point with its GPS accuracy ring, the expected site (and its boundary, if it has one), the distance between them, and a plain-language summary. The Off-station filter (and the review map’s location coloring) gathers the worst offenders so you can check them quickly. A record with a GPS point but no expected site shows as Unverifiable.

Imported records

When you enter a record in CensusIO, you pick its site, so the location check has an expected site to compare against. An imported record (from ArcGIS Field Maps, Kobo, ODK, or a spreadsheet) has no such pick. So as each import or sync lands, CensusIO matches a site code or name carried in the imported data against your location tree: when a station field holds a value that matches one site unambiguously, the record is bound to that site and the location check runs, so an imported point that landed far from the site it names is flagged off-station like any other. A record whose data names no known site is left Unverifiable.

It binds to the site the data names, not the nearest site, on purpose: binding to the nearest one would quietly hide a mislabeled record, which is exactly what this check is meant to catch. A record whose data does not name a site you have in your tree (or names one ambiguously) stays Unverifiable rather than being guessed.

Data check

For surveys that record numbers, CensusIO compares each value against the spread of that field across the whole survey type and flags the records that stand out, in the Data check column:

Check (a warning): a value toward the edge of the distribution.
Outlier (suspect): an extreme value worth verifying.

Hover a flag to see which field and why (for example, a weight that is far above the median). For fish, the check also includes relative weight (Wr): it flags a weight that is implausible for the recorded length and species, which catches a transposed or mistyped measurement that a simple range check would miss. The Data outliers filter gathers these records.

When a field is marked unique across records (a marker, buoy, or tag number), the Data check also flags a value that ended up on more than one record as an Outlier. The web app and imports block a duplicate as you save, but records synced from field devices are never blocked, so this flag is how a collision that arrived by sync is surfaced for review.

Recomputing the checks

The checks are computed from your data, so they can be refreshed after you edit sites. An admin can choose Recheck locations or Check data to recompute them across the survey type. Recheck locations also (re)binds records to the site their data names, so run it to bring in records imported before this matching existed, or after you add or move sites in the location tree.