Data quality, a major challenge for business management
Each year, 10 to 15% of the data contained in the Customer databases are subject to changes related to, among other things, addresses, telephone or company name changes.
Improving the quality of the client database also requires deduplication actions and better organization of their business data.
It is always possible to use traditional techniques such as “siretisation” of addresses for French customers, optimization of the management of the DUNS number internationally or even the certification of addresses with the help of specialized third-party companies.
But today, there is a much more innovative and efficient approach, the analysis and the recovery of the addresses thanks to the geocoding!!
What does that mean exactly ?
The geolocation of its customers involves the geocoding of addresses, namely the addition of X, Y coordinates in order to be able to visualize on the basis of the map, the positioning of prospects and customers.
The geocoding will reveal ‘visually’ and in a very explicit way the problems (Addresses nonexistent, duplicates, …).
Today, geocoding engines allow geocoding addresses in more than 140 countries
How it works ?
When you run a geocoding on an address database, you get two types of results: non-geocoded addresses and geocoded addresses.
In general, non-geocoded addresses are linked to a coding problem: badly-filled postal code, wrong address.
These errors are easy to correct despite a long and tedious side.
What about the addresses actually considered to be correctly geocoded ?
Geocoders do not all have the same approach; Some will provide contact details even if they didn’t found the address, others will provide a bad geocoding rate but will be very accurate for good results.
Several methods are possible to verify the quality of the geocoding according to the “Match rate” obtained:
- Positioning the points according to their Latitude and Longitude on a graph makes it possible to highlight points that are too offset. Addresses located on the outskirts may be false. This analysis works well on limited areas across a country. It can be refined by removing too eccentric values during the first iteration.
- Work on points with the same location using the previous graph or via a spreadsheet formula. For example, if a large number of points have the same position, the geocoding did not recognize the city or street.
- Check the returned standard address. Via a formula in a spreadsheet, it is possible to compare the input address and the standard address returned by the geocoding application. If the address is set to the number but the number is different from the input, geocoding does not have the expected quality.
It will also be possible to “compete” several geocoding tools to increase the quality of the analyzes.
Errors from the analysis have been identified, what to do ?
It all depends on the data involved and the desired result.
On the expected rendering, the necessary precision must be identified; For example, for the identification of the presence or absence of a tourist office in a city, the positioning in the city is sufficient.
The use of another geocoder makes it possible to improve the zone to be geocoded. Consider whether a common point to these addresses would not mislead them as a different city name from the Postal Code. Lastly, manual repositioning should be considered as a last resort, if necessary, by the operational teams.
Pour approfondir Vandy Berten https://www.smalsresearch.be/comparer-des-geocodeurs/