Aggregators and resellers of hotspot access are likely to rise to more prominence as roaming becomes de rigeur, but how many locations do they offer, anyway? Three companies dominate hotspot aggregation and reselling: iPass, GoRemote (formerly GRIC), and Boingo Wireless. The former two work almost entirely with large corporations, offering a combination of dial-up, wired Ethernet (in hotel rooms and elsehwere), and Wi-Fi hotspots. Boingo resells just hotspot access. None of the three build infrastructure; Boingo does offer turnkey hotspots for venues that want to be Boingo-only locations.
A question has been raised many times over the last several months about how hotspot operators and aggregators count their locations. Even companies that don't resell, like SBC, have adopted terminology that isn't entirely clear. SBC talks about 20,000 access points and 6,000 locations over a few years--but why mention access points or individual pieces of hardware at all?
To produce a comprehensive list and a spot check of counts across each aggregator, I downloaded the free client software either directly from the company, in Boingo and iPass's case, or through a reseller that provides up-to-date listings, in the case of GoRemote. A few days ago, I updated the listings for all the software. I was able to extract the directory information for iPass and GoRemote; it's stored in plain text in a clearly labeled file. Boingo uses a database structure that's password protected, and so I turned to their Web site's location finder to get accurate results.
First, let's look at how each company states their current pool of hotspots venues. The tricky starting point is that many hotel locations that are aggregated by Wi-Fi-only Boingo are, in fact, mostly or entirely Ethernet based. Newer or revamped installations typically feature Wi-Fi in common areas and Ethernet to the room, although more hotels are switching to or choosing all Wi-Fi. So you can't entirely split out wired from wireless locations.
iPass states that worldwide they have over 10,000 Wi-Fi hotspot and Ethernet broadband locations. Boingo Wireless notes that they have 6,000 total locations under contract of which 3,300 are currently available to their users worldwide. GoRemote says it connects to 7,800 Wi-Fi hotspots in 45 countries and territories and 1,391 Hotel Ethernet locations in 27 countries.
Next, I took all of GoRemote and iPass's information and loaded into a flat database fielded by their particular values so that I could examine apples-to-apples information. I used Boingo's Web site and reviewed their entries and address listings to come up with their totals. I started by spotchecking three cities, two of which I had some familiarity with: Seattle, New York (all boroughs), and Houston.
In Seattle, Boingo listed 26 locations, all of which were unique. iPass had 160 entries that represented 149 unique venues. GoRemote, on the other hand, had 35 entries which represented seven unique physical locations.
Drilling down, I found that the Seattle-Tacoma International Airport (Seatac) was listed as a single location by Boingo, four locations by iPass, and 26 locations by GoRemote. Consistently, Boingo had the most conservative, unique count; iPass had an overlap of the same address of about two to three percent for non-airport locations; and GoRemote listed every access point in a given venue, floor by floor and concourse by concourse.
In an interview with iPass a few weeks ago, the company explained that in cases where different network names were used across a facility, they tended to have multiple entries. This information is provided by companies like Wayport, and inflates their raw counts. I found examining all of iPass's airport listings that they often provided two to eight entries for each airport. But other listings were generally clean.
I contacted GoRemote a few weeks ago to get an explanation of how they differentiated between access points (individual hardware installations) and hotspots (individual venues). Their press contact promised to get back to me, but did not. A follow-up was unanswered.
A side note: this site operates as an independent marketing and editorial partner with Jiwire, a hotspot directory. They were not involved in the preparation of this article. Jiwire currently counts a hotspots as a unique provider at a unique vendor, but that will ultimately change on a timetable they have yet to announce.
In New York City, Boingo provided 62 unique listings with no overlap; iPass had 258 entries which represented 247 unique locations, with LaGuardia providing most of the overlap; and GoRemote had 61 listings, but only 21 locations, with the Warwick Hotel accounting for 37 separate entries. I confirmed in their client software that I was seeing these multiple locations for the Warwick, too.
In Houston, GoRemote showed 17 listings and 14 unique locations; iPass had 106 unique listings and locations; and Boingo had 41 unique listings and locations.
It was tedious, but I also built a comparison nationwide in the U.S. -- tedious, because I needed to examine thousands of records (sorted, fortunately) to determine where overlap was occurring. Boingo's records were cleanest with 2665 entries in the U.S., and at most, one to two percent overlapped with each other. Airports were again the culprit for multiple entries for a single venue.
iPass listed 6829 locations, and after excluding airports, which represent no more than two percent of their listings, the general overlap rate was under three percent. Finally, GoRemote had 1361 listings, and 843 unique locations, a remarkably higher rate for this sub-sample. GoRemote doesn't appear to be differentiating in their marketing between the necessary additional entries required for roaming across a venue, and the unique number of locations that a purchasing decision might be made on.
It's clear that for maximum transparency to the user base--including those making purchasing decisions--aggregators should work even harder to collapse their listings into unique venues rather than provide multiple numbers or multiple entries. The decisions required by software to handle many network names in a single venue should be entirely, not partially, hidden from users. Boingo manages to achieve this; there's no reason why all resellers cannot.
When we try to interpret the scale of this industry, having an accurate count of unique numbers offers a better gauge of the relative scale of aggregated networks, and a better tool for customers making decisions.