The growing use of location data is one of the most exciting elements of digital media. Mobile ad spend has jumped 72% in the financial year to June 2016, topping $1.9 billion according to the August IAB/PwC Online Advertising Expenditure Report. With location marketing contributing to that uplift, the market is clearly responding to new opportunities to connect with people.
Spend aside, research from Pubmatic shows integrating location data into campaigns can increase eCPM by up to 275%. So with spend growing and evidence of effectiveness mounting, it’s time to tackle one of location marketing’s biggest uncertainties - location data accuracy.
Left unchecked, inaccurate or improperly used location data can undermine campaign efficacy. Based on evidence from Blis’ platform, around 80% of location marketing spend might be being wasted, depending on the objectives of the campaign. If the marketplace is to tackle this issue, suppliers, clients and agencies need to realise that not all data is created equal, and get serious about scrutinising it.
Location as a proxy to defining an audience lives and dies on accuracy. In the current market, accuracy can be as much as hundreds of kilometres off. Even 10 to 20 metres can be problematic. What if I want to build an audience of gym bunnies, but am actually seeing impressions from a burger-lover at the fast-food outlet next door?
These accuracy problems might be OK if you’re simply aiming for real-time proximity targeting, say, reaching people in Sydney, but they start to create problems when you want to build a predefined audience. The real, and significantly more scalable, opportunity for marketers is in audience targeting. To be more specific, using location habits as a proxy to define an audience that can be reached anywhere at any time.
Say for example you see impressions at around 3pm on a weekday in a popular cooking app. You might assume they originate from parents sitting in a car waiting to pick their children up from school. But imagine that impression included location data that placed the user in the Qantas business lounge. You can now imagine that user as a business traveller, and their audience profile changes entirely. Taking historical and real time location habits to build audience pools is a hugely important part of location data, and inaccuracy is poisoning the well.
Looking solely at mobile, agencies and clients should take a closer interest in the chain of data that leads to an ad being bought. I’m not saying mobile specialists are being dishonest, most simply haven’t been independently validating what they’re being told by publishers and exchanges.
During campaign work for the federal election, I had an agency partner tell us that one of our competitors could ‘see’ three or four times more inventory that we could. While I bit my tongue at the time, reason for this is simple. That competitor wasn’t vetting their data properly, and was looking at inaccurate location signals. Accuracy isn’t just about woolly data, it’s about understanding what kind of data is and isn’t useful to your campaign objectives.
In our platform dashboard, I recently ran a test against every biddable impression that we saw within AU/NZ on a typical given day. Only 16% of impressions passed our internal verification algorithms. Let's take a look at the top three reasons for failure.
Centroids
GPS data is the most accurate location data on offer. Where GPS location data is missing, publishers often partner with 3rd party suppliers to approximate user location. This often results in locations being conveniently approximated right in the centre of the CBD. Centroids can be wildly inaccurate, and leave advertisers bidding for, and buying, CBD impressions that could actually be closer to Penrith. It’s worth noting centroids are fine if you’re building a city of state campaign, but they’re useless if you’re trying to build an audience pool, as advertisers are increasingly looking to do.
Uniques
Publishers fail this test when they send a suspiciously large volume of data from a single latitude and longitude (precise point on a map). For example one app may send over one million bids from an exact spot in the very centre of New York. This is not realistic human behaviour and should be scrutinised.
Not enough data
Methods are in place to measure publishers against a minimal threshold of bids needed in order to make a confident decision on whether a publisher’s data is accurate. Without enough bids for a publisher (less than 100 daily impressions) best practice errs on the side of caution and eliminates these publishers from the mix.
So what does all this mean?
When publishers fudge the data that informs the bidding and buying process, and mobile specialists don’t check that data closely for inaccuracies, advertisers’ campaigns are compromised.
Location feeds the complete picture of the user, and greatly enhances the overall picture of who they are. Recurring residences, CBD activity, shopping malls, all of these data points mesh to complete the consumer profile, and continually augment that picture in real-time.
Campaigns running on inaccurate data are more normal today than many would like to admit. If we want to work in a self-regulating industry, closer scrutiny of data sets is paramount. With such large and amounts of money changing hands, can we really accept anything less? Especially as location spend continues to grow.
Rupert Pay,
APAC VP,
Blis