Precision Matters: The Critical Importance of Decimal Places and Why Five Is the Lowest You Should Go
As we’ve noted in previous posts, GPS is the Gold Standard when it comes to lat/long data. It delivers both precise and accurate information. However, we’ve also made it clear that not everything that professes to be GPS data really is. Sometimes its data gathered from centroids. And sometimes, data that claims to be precise lat/long information isn’t quite precise enough.
Size Matters When It Comes to Decimal Places
Precision can be measured by the number of decimal places in the latitude and longitude provided. The number of decimal places correlates directly to the level of precision. So, as you can see from the table below, a lat/long rounded up to a single decimal place can accurately identify a country or region, whereas rounded up to two could identify a large city or district. But five decimal places can accurately hone in on an individual tree, and six can identify a person.
||Can identify a Country or large Region from another
||Can identify a large city from a neighbouring large city
||Can identify a small village from the next
||Can separate one neighbourhood or street from another
||Can identify an individual street or parcel of land.
||Capacity to distinguish one tree from another.
||Measuring approximately 4 inches wide. Used for structural design & surveyance in engineering.
||Used for precision geographic surveying, representing the practical limit of the use of GPS.
||Conceptually the width of a paper clip. Can be used for charting volcanic movements & tectonic plate shifts.
||Representative of the width of a strand of thread, now in the range of microscopy.
Why does precision matter so much?
With so much precision riding on a single decimal place or two, you can see why it matters so much when it comes to targeting. Rounding up from four decimal places to two means the difference between tracking a person’s behaviour in a store to determining whether or not they’re in a particular neighborhood. Even a difference of 10 meters (as between rows three and four) represents the difference between useful and not useful for Blis. At row four, a consumer may be in the hair salon they visit every week. At row three, they could be across the street buying coffee. The difference may seem small, however in the world of hyper local advertising, it’s significant.
Using Google Maps, you can see for yourself how precision works with respect to decimal places. Type in your address, then look at the URL. I’ve typed in the address of Blis’ New York City office, which is at 19 West 24th Street.
Here’s the URL:
The coordinates are within the URL, right after the actual address I typed in: 40.7428453, -73.9905663. Google sets the standard at seven digits – down to a “practical limit of commercial surveying.” It’s a level of accuracy that is generally greater than most Blis campaigns require: In this example, we’re drilling down to the grommet on a co-worker’s shoe, or to a thumbtack on her desk. Anything beyond this is really impractical, even for our purposes.
If you take those coordinates and put them into a tool like LatLong.net (since Google doesn’t actually let you remove decimal places), you’ll pull up that location. Now, take a digit off each number and search again. Repeat that process, and you’ll see the radius around your location get wider and wider. The accuracy remains, but the precision suffers with every lost decimal place.
It’s easy to see why the decimal places matter – and why Blis has no use for data with less than five decimal places.
Making the Cut
At Blis we have to be particularly careful with lat/long data. It’s the most common place for errors and fraud, but also (fortunately) the easiest for us to test and identify errors.
And again something worth noting, inconsistencies in lat/long data are not always as a result of fraud. A developer could have released an update to their App’s SDK (Software Development Kit), accidentally introducing a bug when querying the phone’s GPS chip. In one case, the error was easy to spot because when the data was visualized, all activity from one single publisher falling exactly along the Equator, clearer not a deliberate mistake. Occasionally lat/long coordinates are erroneous. It happens.
Sometimes, however, the activity is nefarious. After all, bid stream data with lat/long information is worth more that data without it, so there is incentive for fraudsters to add in fake or appended lat/longs. In many cases, decimal places are rounded up or deliberately truncated, so that they appear to comply with the minimal standards Blis and other providers demand.
With our proprietary algorithms and technology, we’re able to accurately scan the data we take in validated that all our lat/long data is accurate. In the cases where it’s not, that data is removed from our pool so that it cannot be used by any of our clients for any purpose – and it can’t spoil any campaign, any attribution analysis, anything at all.
Using location data for behavioral targeting requires both precision and accuracy. Blis is committed to delivering both. For what we do and what our clients expect, we need to be exact.
Tags: Amy Fox, Dubunking data series, lat/long data, Location data