A smartphone user who signs up for Pandora Media Inc.’s downloadable app in San Francisco and then moves to New York could still be getting ads for California restaurants months later.
Therein lies the challenge for advertisers trying to reach a specific mobile-using audience: A lot of locally targeted ads get lost along the way.
Marketers spend as much as $2.4 billion a year on mobile ads aimed at users based on geography, and Pandora is far from the only app whose ads sometimes go awry. About 65 percent of location-targeted mobile ads are materially off from their intended location, according to research from mobile-ad network Thinknear, a unit of Telenav Inc.
“Local mobile ad spending is on the rise,” Catherine Boyle, an analyst at EMarketer Inc., said in a phone interview. “But the definition of local for one advertiser can be very different from another. It’s when you want to get into a one-to-three block radius around a specific location, that’s where the advertiser really needs to do a lot of research.”
That has big implications for developers, advertisers and big Internet companies like Facebook Inc. and Google Inc. that are racing to make money from the growing number of consumers who are shunning traditional computers to spend more time on apps, social media and websites on mobile phones.
Consider the plight of Pandora, which lets advertisers target users by the zip code they use at sign-up. If customers don’t manually change zip-code information when they move, ads will continue to be misdirected.
Only 34 percent of mobile ads using longitude and latitude to target users in a specific location are accurate within 100 meters (328 feet), according to a recent survey of 53 million ad impressions by Thinknear, and 27 percent are off by more than 10,000 meters.
“It’s a buyer-beware sort of market,” Boyle said.
Blue Chip Marketing Worldwide, which helps manage advertising campaigns for the likes of Procter & Gamble Co. and Ricola, does a lot of research. That includes finding an ad network -- the businesses that match marketers with the apps or websites they seek to reach -- capable of placing localized ads.
“It’s definitely challenging,” Sarah VanHeirseele, vice president of digital at Blue Chip, said in an interview. “You have to know who you are working with, and really peel back the layers and understand how it works.”
The reason for the extra research is that everyone estimates location differently -- for instance, based on a zip code or city and state disclosed when signing up for an app. Mobile-app publishers may then provide a central point of that zip code to the ad networks, claiming that as the consumer’s actual location. An advertiser trying to reach a consumer near, say, a Home Depot store may reach him across town instead.
Users on the move create problems. While an app publisher may claim to have the latitude and longitude coordinates equivalent to a one meter by one meter place on a map, that could actually be off by hundreds of meters, Eli Portnoy, general manager at Thinknear, said in an interview. If a publisher used a central point of a zip code provided at signup to pinpoint a user’s location, the customer could actually be currently traveling in a different zip code altogether.
“The industry is moving very, very quickly, and location is very hard,” Portnoy said. “A lot of publishers are not location experts, and they are trying to get location, and they don’t understand how.”
“I don’t want to call it outright fraud, but it definitely isn’t fully honest,” he said.
Not all publishers targeting by location are trying to do so as granularly. Pandora, which doesn’t try to target more narrowly than zip code, says most of its advertisers want to run nationwide, large-scale campaigns and aren’t interested in reaching people currently located on a specific city block.
“I could see it as a valuable narrow niche,” Jack Krawczyk, Pandora’s director of product management, said in an interview.
To bolster its more targeted ads, the online radio company has begun to check where a devices’ IP address is coming from and prompt a user to change their zip code if needed, he said. While not all users actively opt in, Krawczyk said Pandora’s surveys have shown that its locations accuracy is in the high 90s on a percentage basis because people don’t move that frequently.
Temecula Valley Wine Growers Association used Pandora to promote its tasting events to listeners in San Diego County, helping it sell more advanced tickets to two tasting events than in prior years.
Location-targeted campaigns are on the rise, and more than 67 percent of all mobile ads available now contain location data, up from 10 percent just two years ago, according to Thinknear. Publishers have an incentive to try to target their ads because location-specific ads can garner double the regular rates, according to Millennial Media.
To target users more accurately, some agencies and advertisers are starting to bypass the big mobile-ad networks and work with smaller shops specializing in finely-tuned location targeting, said Michael Boland, an analyst at BIA/Kelsey.
“It’s the larger, more broadly defined ad networks that are going about it the lazy way,” he said.
Some ad networks are refining their technology to deliver better-targeted ads. Mobile-ad network Millennial Media uses analytics software to double-check the locations publishers provide by tracking a phone’s IP address and users’ movements over time. Another provider, xAd, places ad requests on a map to figure out if an app’s users appear clustered in the middle of each zip code.
Ad-network PlaceIQ Inc., meanwhile, licenses satellite information and uses cartographers to create maps of businesses, such as university campuses and car dealerships, to target locals more precisely.
Ads that are the most precise in determining a user’s location typically use information from a phone’s GPS, which mobile apps usually ask for permission to tap into. Location can also be triangulated based on nearby cellular towers or Wi-Fi hotspots, or estimated based on additional data the user provides, such as check-ins at places on Foursquare. The GPS data isn’t available if apps don’t collect it, or if users opt not to allow access to that information.
Using these more finely-tuned approaches to targeting can almost double a campaign’s success rate, meaning more people click on the ad, according to consultant Mobilesquared. Location-targeted ads that estimate a user’s location based on, say, a zip code can see click-through rates of 0.5 percent, Nick Lane, chief insight analyst at Mobilesquared, said in an interview. That means an ad gets five clicks for every 1,000 times it’s shown.
If the location is more accurate, such as using a phone’s GPS, 0.8 percent to 0.9 percent of recipients of the ad may click on it, he said.
“It’s huge,” Lane said. “No other media can deliver such a targeted ad to the mass market,” he said of mobile phones.
The rate of clicks for mobile ads that aren’t based on location at all hovers around 0.1 percent.
The danger is that the current flood of poorly targeted mobile ads will turn off advertisers and consumers. Consumers who keep receiving ads that are not relevant to them may pay less attention to all ads thereafter, according to Mike McGuire, an analyst at Gartner Inc.
That said, advertisers are used to a large portion of their ad dollars not generating returns. Single-digit response is considered to be good for most campaigns.
When done right, business owners say the hyperlocal ads help drive traffic.
To lure tobacco customers in its Chicago neighborhood of Ravenswood, Nothing Like It smoke shop tried advertising on mobile phones within a three-block radius. It worked, bringing in two to three walk-ins a day.
The shop had been listed on Yelp Inc. and advertised on Google Maps with little success getting new customers in the door more than a couple of times a quarter, said co-owner Francisco Melendez.
He then turned to a startup, CommuteStream, that places ads in mass-transit apps. Melendez now spends about $5 a day to send his ad to about 250 commuters, and he said it paid off immediately.
Started in February, CommuteStream knows users’ daily commute patterns, has access to data from their phones and can figure out just how much time they have to stop by a merchant and make a purchase before needing to board a bus. The startup can typically determine a user’s location within less than 100 feet, said Samuel Pro, chief executive officer of CommuteStream.
“When you can interact with commuters that are passing by your business on a daily basis, it’s those people who have the best chance of becoming regular customers,” Pro said. “It’s really about driving foot traffic and finding customers that are going to be loyal.”