Data quality
Location data ages quickly, and sources change without notice. That is why verification is a fixed part of every delivery.
- Count reconciliation
- Place check
- Coordinates
- Address structure
- Duplicates
One location, before and after
This is how a record arrives from a store finder – and how it leaves. Seven interventions every location goes through.
-
address Musterstr.12str + number Musterstr. | 12 Street and house number in separate fields
-
place 12345 Musterstadtzip + city 12345 | Musterstadt Postcode and place separated
-
phone +49 (0)123 456 - 789phone 0123/456789 Area code separated by a slash, consistent across all networks
-
homepage WWW.Beispiel.DE/Marktweb www.beispiel.de/markt Lower case throughout, including the path
-
coordinates 10.000000 / 50.000000lat + long 50.000000 | 10.000000 Latitude and longitude were swapped
-
opening Mon to Sat 7.30am – 9pmopenings Mo-Sa 07:30-21:00 Fixed notation, machine-readable
-
fax —fax stays empty What the source does not provide is not estimated
The location shown is invented. The interventions are not – they run across the entire dataset with every delivery.
Six checks before every delivery
What is invisible in a single record only shows in the whole – a missing branch, a duplicated location, a coordinate outside the delivery area.
-
01
Count reconciliation
against the previous deliveryIf the number of locations differs from the previous delivery, the cause is investigated – a rebuilt store finder looks just like a market decline.
-
02
Place and postcode
against 10,939 municipalitiesChecked against the official register of municipalities – the municipality is delivered, not the district.
-
03
Coordinate position
within the delivery areaSwapped or empty coordinates are caught before they end up on a map.
-
04
Address structure
field by fieldStreet and house number separated, suffixes and range details kept.
-
05
Phone numbers
format and lengthConsistent format, extensions kept, head-office numbers do not count as a location number.
-
06
Duplicates
across the whole datasetMatched via normalised street and place keys, not via text comparison.
Traceable, not smoothed over
Every delivery comes with a change count: new, gone, changed.
If a network shrinks, we deliver the smaller number and say why.
- The same field structure in every delivery, across all store networks
- Stable identifiers, so your dataset can be kept up to date
- Empty fields instead of estimated values
- An explanation for conspicuous changes in counts
- Notice if a store network can no longer be collected
An empty cell is honest. A guessed cell does damage no one notices.
Have a dataset checked?
We look at an existing dataset and tell you where the gaps, duplicates and format errors are.