Waze Highlights Inconsistencies in Local Data
The mapping and navigation app Waze has been getting a lot of good press lately, not to mention a big spike in downloads. Apple CEO Tim Cook suggested that users unhappy with Apple Maps should try it out, and even the Queen of the Internet has offered her enthusiastic endorsement. Wanting to know what all the fuss was about, I’ve been playing with Waze over the last few days, and I’m impressed by its unique feature set and its crowdsourced approach to gathering and sharing local traffic data. I’m pleased to see that Waze is already fulfilling at least some of the potential for community-sourced local content that I touched on in my post last week.
But that’s not why I’m writing about Waze. Because I’m the kind of person who wants to know how local business content is sourced and distributed, one of the first things I tried to do in the app was to search for local businesses. Interestingly, Waze has taken a federated approach. They do not have their own listings, but instead present separate results from Yelp, Google, YP, Foursquare, and Bing.
Searching for the same business category across all five of these services neatly highlights some of the discrepancies and commonplaces of local data. Below is the result of my own informal study. Because of an issue with Foursquare discussed below, I thought it was only fair to compare the results on the providers’ own apps with those on Waze, though for the most part they are consistent.
Several findings in this set of results are worth highlighting.
- No provider except Yelp achieved 100% relevance.
- All providers, with the exception of Yelp, list in their top results from one to three once-prominent local grocery stores that are now closed. One store has been closed for about three years.
- Top results in both Google and YP included the same wrong address for a store that is now closed.
- Top results in Google and Bing include linked but irrelevant businesses inside a grocery store (Walmart Vision Center, Albertsons Photo Finishing).
- Yelp results are the most relevant and up to date — and the fewest.
- Foursquare results in Waze are notably irrelevant, but this seems to be an issue with Waze’s Foursquare implementation. The same search on the Foursquare app yielded a short list of targeted, relevant results second only to Yelp.
I was particularly unforgiving here in my assessment of whether a search result was relevant to the search term. My assumption is that no one entering the phrase “grocery store” will be happy to see a 7-Eleven or a liquor store unless there are no other results available. In fact, results for “grocery store” also included gas stations, bakeries, water treatment facilities, and food distributors. These “outside of category” results, along with the aforementioned vision and photo centers linked to grocery stores, were very common across all providers except Yelp and Foursquare.
While in desktop search, users may have become acclimated to wading through a certain number of irrelevant results to find what they’re looking for, the situation is different for mobile. With mobile search, the bar for relevancy is being set higher than ever. Mobile users are ready to take action almost immediately and will gravitate toward services that answer their queries effectively and quickly. There are signs in this limited sample that search providers will need to step up their game in order to satisfy mobile users.
It’s notable that crowdsourced content from Yelp and Foursquare is so much stronger in these results than content from directories. Like Waze, Yelp and Foursquare are demonstrating that an active and engaged user base can dramatically improve data quality and user experience. But the lesson is not so much that directories should abandon their current approach and submit to the wisdom of crowds. Clearly, the approach of Yelp and Foursquare is better suited to popular search categories like grocery stores that inspire stronger engagement among users. If local search is also to service the long tail, it must find better ways of delivering targeted, relevant content through algorithms rather than people.
The frequency of closed businesses here speaks to the need for faster routing of changes through the publication process. Aside from that, the clearest takeaway is that search providers need to better understand the intent behind keyword searches.
As anyone knows who has dealt with the inner workings of local data, one of the greatest challenges in passing content from one platform to another, or combining data feeds from multiple sources, is the multitude of standards for categorizing businesses. There are two governmental standards, SIC and NAICS; the former is weighted toward the production and sale of goods while the latter is said to be more oriented towards services. NAICS “replaced” SIC in 1997 according to the Census Bureau, but in reality both are still commonly used.
Both of these systems categorize businesses like a bureaucrat would, with headings like “Delicatessens (except grocery store, restaurants)” and “Grocery stores, with or without fresh meat, retail.” But they are the closest thing we have to a common standard, so one or the other frequently acts as a lingua franca when companies share local data. Most search providers create their own methods of categorization for use by consumers, however, so that ordinary language queries like “grocery stores” will be well supported. The problem is that the more user-friendly category sets are proprietary, while local search data is sourced from all over the place. Listings can be mapped well or poorly by category from one system to another.
Simply put, there are too many different ways to categorize businesses, and none of them represents a unified standard for online search. Such a standard if widely implemented would make all businesses categorized as grocery stores line up neatly with each other and would provide a significant boost in overall relevancy.
Damian Rollison has served as VP of Product for Universal Business Listing since 2010. He holds degrees in English from the University of California, Berkeley, and the University of Virginia, where he did graduate work at the Institute for Advanced Technology in the Humanities. Damian’s articles on emerging technology have also appeared in VentureBeat.