The Browser Doesn’t Matter So Long As It Goes to Google

The whole Google/Firefox issue is an interesting one: Google is the default search engine in Firefox because it pays to be there. The three-year deal expired in November 2011. Would they renew? Some thought no. They were wrong.

Not only has Google renewed the deal whereby it effectively bankrolls Firefox, but it’s the first time that it’s continued the deal after launching its own browser, and the first time it’s done so after Chrome is actually has as many users, according to some measures, as Firefox.

On top of that, there are reports from AllThingsD that the deal is worth $300 million a year, more than three times what they were paying under the previous arrangement. What gives?

Several theories:

We’re Partners

The official version is that Google and Firefox are buddies, after the same thing: the betterment of the web [ReadWriteWeb].

Bidding War

One is that Microsoft and possibly Yahoo! were after the deal. Makes sense: Microsoft is desperate to gain market share for bing, while Yahoo! is, well, desperate.


Another theory has it that Google is basically after eyeballs, and doesn’t care how it gets them. Paying for them by getting to be the default search brings oodles of traffic. This is definitely true. I reckon that Firefox had as many as 500 million users in 2010. If 90% of those users don’t switch their default search that’s worth a lot of money to Google, and as ExtremeTech has pointed out, makes Firefox the biggest single source of traffic to Google (I calculate they paid 20 cents per user, whether or not they actually use Google.)


There are other theories. One is that Google is worried about antitrust issues [David Ulevitch, Twitter feed, via paris lemon] and therefore wants there to be a competitor about. This argument has some merit: expect Google Chrome/Chrome OS and Android to converge more and more, which is bound to attract the attention of regulators.

There’s no question that Google benefits any which way this goes.

  • It’s clear that Microsoft has failed to dislodge Google as the search engine of choice: While its market share in the U.S. is around 15% [WinRumors, quoting comScore] globally it’s tiny: less than 4% on desktop browsers, 1% on mobile devices [both from NetMarketShare]. In other words, Google doesn’t need to worry that Internet Explorer shifting traffic to bing. While in decline IE is still the most popular browser at about 40% [StatCounter].
  • Google doesn’t really care what browser people use. It would prefer they use Chrome, but as long as the browser points to Google, who cares (as Deng Xiao Ping said, who cares what colour the cat is, as long as it catches mice?). Which is why Google are just as happy to do a deal with Apple (6%) and with Opera (2%). In fact, the only browser that doesn’t have Google as its default search engine is IE. (Apple talked about cutting a deal with Microsoft last year [Daring Fireball], but it was probably a negotiating tactic. DF says he reckons the Google/Safari deal was worth $2 million a month.

Finally, then, if the new figures are true–that Google is now paying $300 million a year for the Firefox traffic–is that money well spent? Well, it’s not easy to calculate. But let’s assume that Firefox traffic continues to fall at its present rate. So in 2012 it accounts for only 21% of the market. Likely number of Internet users in 2012? Anyone’s guess, but probably about 2.4 billion? (It was 2.1 billion in March 2011, according to Internet World Stats.)

So Firefox potentially should be able to bring at least 440 million users to the table. So that’s $0.68 per user. Quite a bit more than what it’s currently shelling out–but less than what it’s paying Opera, according to my very rough calculations. Opera said it received $41 in ‘Desktop revenue’, the bulk of which it says comes from ‘search and commerce’. Assuming all of that, for the sake of argument, is money from Google for search, then using their official figure of 51 million desktop users in 2010, Opera was getting $0.80 per user from Google. (I realise that might be inflated given the ‘commerce’ component.)

That would seem to suggest that actually Google was getting users from Firefox pretty cheaply. Even if my calculations for Opera are a tad high, the new deal with Google, valuing a user at about 65 cents, doesn’t seem overly expensive. We don’t know how much Google pays Apple, but the $2 million a month means they’re the cheapest on the block, costing $0.15 per user according to back of the envelope calculations.

Indeed, these are all just back of the envelope calculations, but I reckon they offer a bit of insight into the economics of this part of the game. Remember Google made $9.72 billion in the last quarter [Google corporate pages], and paid out $383 million to “certain distribution partners and others who direct traffic to our website” in that quarter. That’s close to $1.6 billion over a year, putting the $300 million it’s reputed to be committed to paying Firefox every year in perspective.)

A good account of the economics of all this can be found at ExtremeTech.

Dragon Tattoo: Mac or Vaio?

Some folk have suggested that because Sony is behind the new version of the Girl With the Dragon Tattoo all the product placements are Vaio:


This would be a slap in the face of the original novel and the Swedish version, which stuck pretty closely to the text.

But I’m not sure. Here’s another picture from a Guardian piece:


which looks pretty Apple-ish to me.

Inside the Web of Things

This is a slightly longer version of a piece I’ve recorded for the BBC World Service

I’ve long dreamed of an Internet of things, where all the stuff in my life speaks to each other instead of me having to the talking. The vision is relatively simple: each gadget is assigned an Internet address and so can communicate with each other, and with a central hub (my, or my computer, or smartphone, or whatever.)

The most obvious one is electricity. Attach a sensor to your fusebox and then you can see which or your myriad appliances is inflating your electricity bill. Great idea! Well sort of. I found a Singapore-based company that was selling them, and asked to try one out. It was a nice, sleek device that promised to connect to my computer via WiFi and give me a breakdown of my electricity consumption. Woohoo.

Only it never worked. Turns out the device needed to be connected to the junction box by a pro called Ken, who tried a couple of times and then just sort of disappeared. I don’t mean he was electrocuted or vaporized, he just didn’t come back. The owner of the company said he didn’t really sell them anymore. Now the device is sitting in a cupboard.

Turns out that Cisco, Microsoft and Google tried the same thing. The tech website Gigaom reports that all three have abandoned their energy consumption projects. Sleek-looking devices but it turns out folk aren’t really interested in saving money. Or rather, they don’t want to shell out a few hundred bucks to be reminded their power bills are too high.

This might suggest that the Internet of things is dead. But that’d be wrong. The problem is that we’re not thinking straight. We need to come up with ways to apply to the web of things the same principles that made Apple tons of cash. And that means apps.

The Internet of things relies on sensors. Motion sensors which tell whether the device is moving, which direction it’s pointing in, whether it’s vibrating, its rotational angle, its exact position, its orientation. Then there are sensors to measure force, pressure, strain, temperature, humidity and light.

The iPhone has nearly all these. An infrared sensor can tell that your head is next to the phone so it can turn off the screen and stop you cancelling the call with your earlobe. (The new version can even tell how far away you from the phone so it can activate its voice assistant Siri.)

But what makes all this powerful is the ecosystem of third party applications that have been developed for the iPhone. Otherwise it’s just a bunch of sensors. There are 1000s of apps that make use of the iPhone’s sensors–most of them without us really thinking about it.

This is the way the Internet of things needs to go. We need to stop thinking boring things like “power conservation” and just let the market figure it out. Right now I want a sensor that can tell me when the dryer is spinning out of control, which it tends to do, because then it starts moving around the room. Or help me find my keys.

In short, the Internet of things needs to commoditize the sensors and decentralize the apps that make those sensors work. Make it easy for us to figure out what we want to do with all this amazing technology and either give us a simple interface for us to do it ourselves, or make a software kit that lets programmy people to do it for us.

Which is why some people are pretty excited about Twine, a bunch of guys from MIT who are working on a two and a half inch rubber square which connects to WiFi and will let you program it via a very simple interface. Some examples: hang it around your infant’s neck and get it to send you a tweet every time it moves.

It may not be rocket science, but if you’ve got an infant-wandering problem it could be just what you needed.

Podcast: Web of Things

The BBC World Service Business Daily version of my piece on The Web of Things to Come  (The Business Daily podcast is here.)

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

To listen to Business Daily on the radio, tune into BBC World Service at the following times, or click here.

Australasia: Mon-Fri 0141*, 0741

East Asia: Mon-Fri 0041, 1441
South Asia: Tue-Fri 0141*, Mon-Fri 0741
East Africa: Mon-Fri 1941
West Africa: Mon-Fri 1541*
Middle East: Mon-Fri 0141*, 1141*
Europe: Mon-Fri 0741, 2132
Americas: Tue-Fri 0141*, Mon-Fri 0741, 1041, 2132

Thanks to the BBC for allowing me to reproduce it as a podcast.

We’re Not in the Business of Understanding our User


A few years ago I wrote about sometimes your product is useful to people in ways you didn’t know—and that you’d be smart to recognise that and capitalize on itn (What Your Product Does You Might Not Know About, 2007).

One of the examples I cited was ZoneAlarm, a very popular firewall that was bought by Check Point. The point I made with their product was how useful the Windows system tray icon was in that it doubled as a network activity monitor. The logo, in short, would switch to a twin gauge when there was traffic. Really useful: it wasn’t directly related to the actual function of the firewall, but for most people that’s academic. If the firewall’s up and running and traffic is showing through it, everything must be good.

The dual-purpose icon was a confidence-boosting measure, a symbol that the purpose of the product—to keep the network safe—was actually being fulfilled.

Not any more. A message on the ZoneAlarm User Community forum indicates that as of March this year the icon will not double as a network monitor. In response to questions from users a moderator wrote:

Its not going to be fixed in fact its going to be removed from up comming [sic] ZA version 10
So this will be a non issue going forward.
ZoneAlarm is not in the buiness [sic] of showing internet activity.
Forum Moderator

So there you have it. A spellchecker-challenged moderator tells it as it is. Zone Alarm is now just another firewall, with nothing to differentiate it and nothing to offer the user who’s not sure whether everything is good in Internet-land. Somebody who didn’t understand the product and the user saved a few bucks by cutting the one feature that made a difference to the user.

Check Point hasn’t covered itself in glory, it has to be said. I reckon one can directly connect the fall in interest in their product with the purchase by Check Point of Zone Labs in December 2003 (for $200 million). Here’s what a graph of search volume looks like for zonealarm since the time of the purchase. Impressive, eh?


Of course, this also has something to do with the introduction of Windows’ own firewall, which came out with XP SP2 in, er, 2004. So good timing for Zone Labs but not so great for Check Point.

Which is why they should have figured out that the one thing that separated Zone Alarm from other firewalls was the dual purpose icon. So yes, you are in the business of showing Internet activity. Or were.

(PS Another gripe: I tried the Pro version on trial and found that as soon as the trial was over, the firewall closed down. It didn’t revert to the free version; it just left my computer unprotected. “Your computer is unprotected,” it said. Thanks a bunch!)

Podcast: True Video Lies

The BBC World Service Business Daily version of my piece on how video doesn’t always tell the truth. (The Business Daily podcast is here.) The piece it’s drawn from is here

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

To listen to Business Daily on the radio, tune into BBC World Service at the following times, or click here.

Australasia: Mon-Fri 0141*, 0741

East Asia: Mon-Fri 0041, 1441
South Asia: Tue-Fri 0141*, Mon-Fri 0741
East Africa: Mon-Fri 1941
West Africa: Mon-Fri 1541*
Middle East: Mon-Fri 0141*, 1141*
Europe: Mon-Fri 0741, 2132
Americas: Tue-Fri 0141*, Mon-Fri 0741, 1041, 2132

Thanks to the BBC for allowing me to reproduce it as a podcast.

True Video Lies

This is a longer version of a piece I recorded for the BBC World Service.

The other day my wife lost her phone out shopping. We narrowed it down to either the supermarket or the taxi. So we took her shopping receipt to the supermarket and asked to see their CCTV to confirm she still had the phone when she left.

To my surprise they admitted us into their control room. Banks of monitors covering nooks, crannies, whole floors, each checkout line. There they let us scroll through the security video—I kind of took over, because the guy didn’t seem to know how to use it—and we quickly found my wife, emptying her trolley at checkout line 17. Behind her was our daughter in her stroller, not being overly patient. It took us an hour but in the end we established what look liked a pretty clear chain of events. She had the bag containing the phone, which she gave to our daughter to distract her at the checkout. One frame shows the bag falling from her hands onto the floor, unnoticed by my wife.

Then, a few seconds later, the bag is mysteriously whisked off the floor by another shopper. I couldn’t believe someone would so quickly swoop. The CCTV records only a frame a second, so it took us some time to narrow it down to a woman wearing black leggings, a white top and a black belt. Another half hour of checks and we got her face as she bought her groceries at another till. No sign of the phone bag by this time, but I was pretty sure we had our man. Well, woman.

Except I’m not sure we did. What I learned in that control room is that video offers a promise of surveillance that doesn’t lie. It seems to tell us a story, to establish a clear chain of events. But the first thing I noticed was when I walked back out into the supermarket, was that how little of the floor it covered, and how narrow each camera’s perspective was.

For the most part we’ve learned that photos don’t always tell the truth. They can be manipulated; they offer only a snapshot, without context. But what about videos? We now expect to see cameraphone footage in our news bulletins, jerky, grainy recordings taken by unseen hands, raw and often without context.

This is not to say videos are not powerful truth tellers. But we tend to see what we want to see. When a policeman pepper sprays protests at the University of California there is outrage, and it does indeed appear to be unwarranted. But when four of the videos are synchronized together a more complex picture emerges. Not only can one see the incident within context, but also one gets a glimpse of a prior exchange, as the officer explains what he is about to do to one protester, who replies, almost eagerly: “You’re shooting us specifically? No that’s fine, that’s fine.”

This is not to condone what happens next, but this exchange is missing from most of the videos. The two videos that contain the full prelude are, of course, longer, and have been watched much fewer times: 12,658 (15 minutes) and 245,226 times (8 minutes) versus 1,346,781 times (1 minute) for the one that does not  (the other video has since been taken down).

I’m not suggesting that the more popular video has been deliberately edited to convey a different impression, but it’s clearly the version of events that most are going to remember.

We tend to believe video more than photos. They seem harder to doctor, harder to hoodwink us, harder to take out of context. But should we?

It’s true that videos are harder to fake. For now. But even unfaked videos might seem to offer a version of the facts that isn’t the whole story. Allegations that former IMF president  Dominique Strauss-Kahn may have been framed during a sexual encounter at a New York Hotel, for example, have recently been buttressed by an extensive investigation published recently in the New York Review of Books. There’s plenty of questions raised by the article, which assembles cellphone records, door key records, as well as hotel CCTV footage.

The last seems particularly damning. A senior member of the hotel staff is seen high-fiving an unidentified man and then performing what seems to be an extensive dance of celebration shortly after the event. This may well be the case, but I’d caution against relying on the CCTV footage. For one thing, if this person was in any way involved, would they not be smart enough to confine their emotions until they’re out of sight of the cameras they may well have installed themselves?

Back to my case: Later that night we got a call that our phone had been recovered. The police, to whom I had handed over all my CCTV evidence, said I was lucky. A woman had handed it in to the mall’s security people. I sent her a text message to thank her. I didn’t have the heart to ask her whether she had been wearing black trousers and white top.

But I did realise that the narrative I’d constructed and persuaded myself was the right one was just that: a story I’d chosen to see.

Carrier IQ’s Opt-Out Data Collection Patent

ZDNet writes here about an Carrier IQ patent that outlines keylogging and ability to target individual devices . Which is interesting. But Carrier IQ owns a dozen patents, including this one, which to me is much more interesting. This patent indicates what Carrier IQ software could do—not what it does—but it is revealing nonetheless:

A communication device and a data server record and collect events and event-related data to create an activity record. A user of the communication device may request that events and related data be recorded and collected using a configuration option on the communication device or through an interaction with the data server. Data are grouped into data sets and uploaded to the data server either automatically or upon user approval. The data server uses the uploaded data to create an activity record which the user may access through a website. The user uploads additional data which are associated with the activity record. In some instances, the data server embeds a link pointing to the additional data in an entry in the activity record corresponding to an event associated with the additional data.

Basically this patent offers a way for a “user”—which could be either the user of the device or the service—to have a record of everything they do:


While most of the patent is clearly about a product that would create a ‘lifestream’ for the user—where they can access all the things they’ve done with the device, including photos etc, in one tidy presentation, there’s clearly more to it than that. Buried in the patent are indications that it could do all this without the user asking it to. It’s paragraph 0023 which I think is most interesting:

A user of a mobile device requests that events and event-related data be collected by a data server and data collection begins. Alternately, data collection may be a default setting which is turned off only when the device user requests that data collection not occur. In yet another embodiment, a request from a server can initiate, pause, or stop data collection. The mobile device is configured to record events performed by the mobile device as well as event-related data. Typical events that the mobile device records include making or receiving a phone call; sending or receiving a message, including text, audio, photograph, video, email and multimedia messages; recorded voice data, voice messages, taking a photograph; recording the device’s location; receiving and playing an FM or satellite radio broadcast; connecting to an 802.11 or Bluetooth access point; and using other device applications. The data most often related to an event include at least one of: the time, date and location of an event. However, other event-related data include a filename, a mobile device number (MDN) and a contact name. Commonly, the mobile device records events and provides a time, date and location stamp for each event. The events and event-related data can be recorded in sequence and can be stored on the mobile device.

This seems to suggest that

  • basically all activity on the phone can be logged
  • the software can be turned on by default
  • the software can be turned on and off from the server

All this information would be grouped together and uploaded either with the user’s permission or without it:

[0025] The mobile devices may be configured to store one or more data sets and upload the data sets to the data server. In one embodiment, the data sets are uploaded automatically without user intervention, while in other embodiments the mobile device presents a query to the user beforehand. When the mobile device is ready to upload one or more sessions to the data server, a pop-up screen or dialog may appear and present the user with various options. Three such options include (1) delete session, (2) defer and ask again and (3) upload now. The user interface may present the query every time a session is ready to upload, or the user may be permitted to select multiple sessions for deletion, a later reminder or upload all at once. In another embodiments, the uploading of sessions may occur automatically without user intervention. Uploads may also be configured to occur when the user is less likely to be using the device.

This point—about the option to collect such data without the user’s say-so—is confirmed in [0030]:

Although typically the device and the server do not record, upload and collect data unless the user requests it, in other embodiments the communication device and the server automatically record, upload and collect data until the user affirmatively requests otherwise.

And in [0046]:

In embodiments where participation in the data collection services is the default configuration for a mobile device (e.g., an “opt-out” model), it is not necessary to receive a request from a user prior to recording data.

An ‘opt-out’ model is hard to visualize if this is a product that is a user-centric lifestream.

While patents only tell part of the story, there’s no evidence of any such consumer-facing product on Carrier IQ’s website, so one has to assume these capabilities have been, or could be, wrapped into their carrier-centric services. In that sense, I think there’s plenty of interest in here.

Carrier IQ Bits and Pieces

Some background about Carrier IQ before the hullabaloo started.

  • People had found about this before
  • Some in the industry questioned why such an expensive solution for a relatively simple problem
  • Data was available to ‘market researchers’
  • Software was installed on modems too
  • A lot of carriers were involved

This is not new. Several people have pointed this out before. This from December 2010: xda-developers – View Single Post – **warning** you can get your phone to a unrecoverable state:

On whether or not it’s possible for Sprint to dig up data after a complete Odin wipe may be debatable, but I lean toward supporting the “yes, they can” side. Sprint has been, for – as far as I can tell – a while, since the Moment at least, been including Carrier IQ in Android ROMs. Carrier IQ – which you can get more info on here (browse around there) is highly invasive, to the level of being spyware. It tracks signal data, application usage, and much else – its services and libraries are tied deeply into the system, to the point that killing just the client (not the server) will destroy the battery meter.

And this, even earlier, from a potential rival: Carrier IQ: Mobile Service Intelligence ?’s – They point out that Carrier IQ is very expensive, and has raised a lot of money, for something that is supposedly very simple (finding dropout zones). Commenters point out the pitfalls (lower battery life, data in the hands of faceless corporations):

I did not give consent for this and see the use of such software unethical. I can see no positive effect this can have for the end user. I can see many scenarios in which these corporations could heinously profit from it, though.

Back in 2008, it could claim, according to Company 2008: FierceWireless, Fierce 15 – FierceWireless, that

Carrier IQ’s client list includes Sprint and Sierra Wireless. CEO Quinlivan says the firm works with at least seven of the top 10 major OEMs. Look for the firm to increase its scale in the coming year through more vendor and carrier deals.

Huawei is a customer, not only for handsets, but also for modems: Huawei to Embed Network Diagnostic Tools into 3G Modems in 2009 says:

Announcing the partnership, Carrier IQ CEO, Mark Quinlivan, said: “These new cards will make for smoother delivery of Mobile Data services, improvements in Customer Care services, identification of network coverage gaps and increased awareness of actual user behavior.”

This from Sept 2010 Carrier IQ Powers Android Platform with Mobile Service Intelligence makes clear a number of things.

Experience = behavior for Carrier IQ, so this is not just about logging dropouts:

On-device measurement of the mobile user experience is the key to better understanding user behavior and ultimately optimizing product offerings to match market demands.

This data was not just available to the telcos. The press release also includes an unlikely end-user:

Carrier IQ enables mobile operators, device manufacturers, application developers and market researchers to improve their offerings based on direct insight into the customer experience.

As of last year, 12 leading vendors were using Carrier IQ:

Deployed on over 90M devices from 12 leading vendors worldwide, Carrier IQ is the leading provider of Mobile Service Intelligence solutions that use mobile devices to provide detailed metrics in a highly secure environment.