MAJOR UPDATE 4/19/2010
I’m sure you all noticed that over the weekend, the tracker changed again. First, let me apologize for not giving anyone forewarning – it shouldn’t have happened as quietly as it did, and the timing (just after it crossed one million) was pretty bad. That said, we are continually trying to make the tracker more accurate, and we ran into an issue with our workaround to placing cookies on iPads – it works, but is not 100% reliable, and often we were seeing the same iPad with multiple cookies.
I’ve put up a post detailing the changes that have been made to the tracker since its launch, when they were made, and why they were made, and we’re reaching out for feedback on how you would use the data we have available to us. We’re also interested in your thoughts on our cookie workaround – how would you do it better than we have? Our goal is to have as accurate as possible a predictor of iPad sales, and to that goal, as with any estimator, we’ll continue to update and improve our methodology, and with your help we can do that more effectively.
Also, from this point on, any and all changes to the tracker will be announced several hours in advance.
This past weekend, the world of computing came to an end.
At least, that’s what it seemed, with everyone remotely interested in tech (and most of those who aren’t) transfixed by Apple’s newest release, the iPad. More than just an oversized iPod Touch, the iPad has been described (by Apple) as “magical”, “revolutionary”, and “thin” (which it is).
So, the Chitika Research team tapped our genius engineer Gui to build us a page that showed just how many iPads were out there browsing the World Wide Web. Here’s how we did it:
– We count how many new, unique iPads we see coming through the Chitika advertising network
– We multiply that by how much of the Internet we see at any given time to figure out how many iPads in total are out there
– We look at where iPad traffic is coming from by state
With that, we’ve learned a few things. For one, as of noon Eastern, we estimate that about 270,000 iPads are live, in use, and surfing the Internet. For another, it looks like California is running away with the title of “iPadest State in America,” with more than double the iPads of any other single state.
We’ve also learned that the WiFi iPad may not be seen as primarily an Internet surfing device. With estimates of over 300,000 iPads sold on release day and perhaps as many as 700,000 sold over the weekend, that only around 40% are showing up on the Internet may imply that people are buying the iPad for the apps, and spending most of their browsing time in the App Store.
In the meantime, Chitika’s iPad Stats tracker will continue to update in real time. Keep an eye on it and see how the iPad is gaining momentum, and where the momentum is coming from (I’m looking at you, California).
UPDATE 4/8/2010 1:25 EST
Well, it appears that our assumptions in calculating iPad sales have been way off. That’s the bad news.
The good news is that Steve Jobs has given us (and the rest of the world) a second data point – about 450,000 iPads sold through end-of-day yesterday. With this, we’re able to get the proper assumptions – mostly, we need to refine our assumptions on how many unique IP addresses the average iPad hits our network from.
We should have the tracker updated with the second data point and correct calculations soon. Obviously, our 700k number is fairly far off. Thanks for the patience.
UPDATE 4/8/2010 2:00 pm
So it looks like an iPad coming through our system will hit from an average of 2.73 different unique IP addresses. As a fix, we’ll be implementing this as a modifier to the growth numbers. Should have an accurate count up soon.
UPDATE 4/8/2010 3:23 pm
We’ve updated the tracker to reflect a better statistical understanding of the IP-address-per-iPad quandary.
UPDATE 4/12/2010 3:16 pm
We’re now able to track iPads by cookie. Obviously, there’s been some re-adjusting over the past day or so; we’re transitioning from IP address tracking (which came with a lot of assumptions) to tracking by cookies, which is a significantly better way of doing what we wanted to with this tracker. Our engineering team has devised a workaround to Safari on the iPad’s rejection of any and all 3rd-party cookies. This should be the last noticeable adjustment of the tracker’s numbers, and I’ll put up a link to the engineers’ explanation of how they managed the workaround in an hour or so.
UPDATE 4/13/2010 5:00 pm
Gui put up his explanation of how to circumvent the iPad’s cookie blocking here.
Interesting stats. Did you compare with the geographies for the Mac, iPod, iPhone a few days after they launched?
Hmm, haven’t done that, but it definitely would make for a fascinating comparison. May have to come back and revisit this when iPhone OS 4.0 comes out – see what locations adopt the new OS vs. a new product.
Any chance you could do something similar when the Ipad is released in Canada? Mapping Canadian IP addresses. Would love to promote this when it’s released on my website.
ipad stas,it seems great.
Hi
Thanks for the interesting stats. It will be interesting to see how this plays out over time and in comparison to Apple’s press releases and financial reports. Have you done any similar stats, say with the iPhone or perhaps a PC netbook or someother device?
Right now your site is indicated that over 414,000 unique iPads are on the internet and have been found by your tracking.
Using Apple’s report of 300,000 iPads being sold on Saturday as a given, is your reporting indicating that about 114,000 iPads have been sold since the first day of sales, i.e. Saturday (414,000 – 300,000 = 114,000)?
One person asked about Canadian tracking. I assume your system can detect iPad usage worldwide as iPads sales begin in many countries later this year. Is that correct?
Thanks.
i am sure many persons will be visiting your site many times over the next several months and I am sure the numbers will be used many different ways.
I obviously have to get more persons in Rhode Island to buy iPads but since we are the smallest state with less than 1,000,000 person I don’t think our numbers will be very great. Although with Brown University, RI School of Design, Providence College, Univ of Rhode Island, Roger Williams University, Bryant University to name a few perhaps student usage of iPads will increase our per capita iPad population
“Roughsailing”
Bristol, RI
@Roughsailing, thanks for the questions. Unfortunately, I don’t have any similar stats for the iPhone launch; I’m planning on doing something similar when iPhone OS 4.0 hits.
As for the report of sales, keep in mind that our network can see new iPads that are browsing the web with Safari. What I take from the numbers we see versus the numbers Apple reports is that quite a few iPad users are pretty much sticking to Apps for their usage.
We’ve actually seen a good deal of iPads coming through yesterday and today; in fact (and I’ll have a bit more detail up later), the biggest single hour for new iPads was last night, edging out Saturday night. I’d say the iPad started slow, but it looks like it’s going to roll pretty well.
As for international, yes we can and will track iPad usage from different countries when they’re launched.
And go go Rhode Island – we’re actually just up 95 -> 495 from you, in central Mass. I’d imagine a lot of RISD students will be checking out the iPad pretty soon.
Daniel, very nice of you to share these stats. Thanks!
I have a question. You said, “We multiply [the count of new unique iPads] by how much of the Internet we see at any given time to figure out how many iPads in total are out there”
Would you be willing to share a bit about this multiplying factor, how do you derive it, what its value is (roughly), does it vary much over time, etc?
And a couple of suggestions which shouldn’t be too much trouble:
1. It would be great if you could keep a tally of daily new sightings, and not just the current day’s new iPads. I know it can be deduced from the little gray bar chart there, but showing the actual numbers (or even rounded to the nearest thousand) would be really helpful.
2. Doubling the vertical size of the graph would make it much easier to get a feel for those evening surges and late-night stalls. The graph looks a bit squashed as it is now.
Again, thanks for making this available.
Hello there,
could you also publish the raw data of the number of unique ipad that you have seen go through your network ?
I understand that you are exptrapolating it with your network reach, but raw data helps in understanding hiw many ate there for sure.
About the iPhone numbers. Could you also run some stats on how many iPhones and iPod touches hit your network.
I am assuming that clearing the JavaScript etc will make the smae ipad to be counted twice ? This is not very common, but just a thought.
I’d be curious to plot a non-linear trend through the daily uniques so far to come up with a forecast for one month out from launch. Have you guys looked at that? Thanks.
Can the map of the US States be changed to a map with more info if you click or touch a particular state? Then have the data as data points inside the state, or dots of various sizes for number of ipads in various locations in that state, such as clusters on cities?
Are the sales clustered around locations with Apple Stores or Best Buy locations?
Will be great to compare this information with the first week sales of the 3G version, see if that is clustered more closely with areas of good AT&T service. Are the WiFi sales more evenly distributed in non-AT&T areas?
How do you know each iPad you count is unique? What are you using to determine uniqueness? How can you be so sure you don’t keep counting the same ones?
I believe, like cell phones, iPads have a unique number for each.
Interesting data. Can you add a geographic distribution beyond the US? I have seen reports of people who are bringing iPad abroad quite similar to how it was done with the iPhone 1st generation.
For sure there are many iPads being used oversea WiFis.(sold in the USA then went out)
Why those are not being counted ?
You say the see these ipads when they browse the web using safari. My question how do you account for ipads that are being used at the various apple stores browsing the web?
I am also interested in Stu’s question of how you determine that an iPad is a new unique instance.
Also, can you distinguish between the iPad and the iPad Simulator in the SDK?
How do you factor in all the in store display units that people are using to try out the device ?
wow, that’s cool!
but … as @Stu said before .. wo xan you be sure you don’t keep counting the same ipads more than one time?
The iPad refuses 3rd party cookies by default (ie ones from advertising networks), so if that is what you are using to determine iPad uniqueness, then your numbers are completely wrong as you keep counting the same machines.
It’s highly likely that with Apple about to launch their own advertising network, they have deliberately taken this step to given themselves an advantage over every other advertising network as they will have their own way of determining unique machines. Therefore, advertisers are more likely to go with them than anyone else.
I would even go as far as saying that with iPhone OS 4 they will refuse 3rd party cookies by default, so they give themselves an advertising edge on those devices as well.
@Stu, you’re absolutely right on the 3rd party cookies being refused – we’re determining unique iPads by IP address. That’s also how we avoid stores with several test units – we actually tracked the IP address with the most hits to our network and it was a Best Buy in Minnesota. There’s a few issues with this method, admittedly, but we factor in an iPad’s likelihood to hit from multiple IP addresses as best we can in the extrapolation.
It will be interesting to see what Apple does with iAd; it’s going to be more of an AdMob-style network with ads placed in apps rather than on pages.
@Daniel, at any given time we can see roughly between 10 and 15% of the Internet, so our extrapolation is based on that. We’re working on improving the graphs on the live tracker; one thing I plan to add is a day-by-day bar graph of new iPads seen to replace the cumulative graph that’s there now.
@Austin, good idea, I’ll try to put that together tomorrow
Would you folks mind providing state stats for the less ipadded states?
If you navigate to and from the stats page you will quickly discover that the ‘real time’ counter is faked. I’ve watched it climbing past 571,000 four times in the last few minutes. Viewing the page source reveals that the starting count and the rate of increase are hard-coded in the page. Chitika must be updating these from time to time, but this is hardly ‘real time stats’ or a ‘live tracker’.
Daniel,
if you numbers are approximated using “rough 10% or 15%” of the internet that you can see, wouldn’t then number of iPad users be “roughtly” +/- 25%? So maybe reported number shouldn’t be that exact? Is your number using 10% (lower bound) or 15% (upper bound) or something in between?
Have four of them at one IP address. Are they counted?
Hi there – do you have stats for worldwide usage? Would be interested to see who is using it outside of the US.
Hello,
i purchased 2 ipad from new york, but i am back at home in Canada, montreal Area. could you show us ipad surfing the net (like mine) right now out of USA ? this affect ( very marginal ) the sales in USA. Also, checking on Kijiji, a few are on sale in canada, purchased in US….( toronto, montreal and so on )
thanks a lot for this information
franck
You need to start showing hits to your site. I bet your business improves.
Good job, tho I have to say that there is more uncertainty in your figures than I had hoped.
Still…
As a resident of Orygun, I also would like to see a listing of pads per capita, as we are also at a slight disadvantage to our southern neighbor.
Thanks
No es más fácil esperar a que Apple dé sus propias estadísticas de ventas reales en lugar de andar perdiendo el tiempo con adivinanzas ridículas? Creo que alguien debe ser despedido ASAP!
Daniel
Regarding Austin’s idea, I actually have been using your data while acknowledging Apple’s press release of 300,000 units as a starting point. So for the few days that have passed since Saturday. I visited your site (not precise since I grabbed data at whatever time I could late in the evening or early in the morning) and used that data as a daily number and plotted that. I then calculated a sales rate (units per day) for each day. I then made CY2010 projections based on a selected percentage of the current sales rate (assumes that the current rate will drop).
I did this to answer an iPad critic’s comment on an aapl financial discussion group. He had proudly announced using your website that the sales rate had gone down since Saturday. I pointed out that obviously 300,000 units sold per day was not sustainable but at 50% of the current daily rate (which as I see it is over 60,000), Apple will sell over 8 million iPads (I think that is still not realistic, however. It is probably too high but with only 4+ days of data … ).
I think the state chart should be sort by “Pop. per iPad” instead of be sort by “% of iPads”
Of course the bigger states sell more iPads. What is more interesting is what percentage of the population is buying iPads.
Quick FYI for everyone interested, we’re working on improving the stat tracker today – should have more accurate results and more telling graphs by end of day today. We keep on trying to hone the accuracy of the number – it’ll never be perfect, but it’s getting better.
Any suggestions for more info that’s not being displayed is welcome! (Sneak preview: some of our data shows that Google has about 98% of the iPad search engine breakdown)
Best Buy have sold out ipads, it expects to have inventory replenished by Sunday.
But looking at the your stats, the speed of # of IPAD sold is not decreased.
This is kind of weird.
@Jeff, good point, but Apple stores are still holding strong with enough in stock; if someone wants an iPad and finds Best Buy sold out, they can still hit the Apple store that’s likely half a mile away.
Regardless, I need to check in on the growth rate and make sure there’s no bugs.
Please note: we just did a quick adjustment – our growth rate this morning turned out to be somewhat inaccurate on the high side. You may notice estimated total sales dropped from just over 700k to approximately 680k. This should be a more accurate estimate. We’ll continue to improve our accuracy as the day goes on.
@ JEFF – you say:
“But looking at the your stats, the speed of # of IPAD sold is not decreased.
This is kind of weird.”
Perhaps not. Think of it, iPad is a totally new device. While some people have faith in Apple and are willing to jump in, most people probably have taken a “wait and see” attitude. So they waited and they saw, and they like what they saw.
So as people actually see these things our in the wild, they are deciding they want one. It could be a mushrooming growth cycle.
Just a thought.
I bought one iPad and used it in at least 5 different wifi hot spots outside my own home( all with different IP address). Does that mean my iPad is counted 6 times according ton you method?
Another issue here – I think rather important.
You have on your counter page:
– 147,760 new iPads were seen today.
It is very important that people NOT think of this as iPads SOLD today. We like to think of the customer as getting their new toy and tearing it open in the mall food court and jumping online while their coffee gets cold.
Yet many do not even open it for some time. For example, if purchased as a gift, or people preoccupied with other things. Again, many people might be most interested in the books and not much in browsing.
The point is, that if someone waits for 3 days before they finally show up on a site that is monitored, then they show up in statistics for day 3 not day 1. Additionally, if you are counting them as a “missed customer” on day 1 but they show up later, then they are being counted twice.
Again – just a thought.
Still, overall it is an interesting project, and I am sure that it provides some insight into the relative sales growth, even if actual numbers may be skewed.
Thanks again.
“The first day we sold 300,000 iPads, and I want to update you — as of today we’ve sold about 450,000.” -Steve Jobs
oops…
What’s a 50% overestimation among friends?
Hey, this is great – until now we had the static day-1 300,000 number to compare with. Now with two distinct data points, we can see where our estimation is wrong, and overhaul our assumptions. Initial thought is that we vastly underestimated the # of unique IPs each iPad would see this quickly. Need to update (obviously)
@Jz, yeah, your iPad would have been counted 6x in that scenario. As we’ve discovered (the hard way), people with iPads are using them in about double the locations we had initially assumed.
Unfortunately, that’s a problem with assumptions, particularly when you’ve got a single data point to compare with. We had to assume two figures in our initial calculations: the % of the Internet we were seeing at any given time (which we’re pretty comfortable with); and the average # of unique IP addresses each iPad sees.
In very short order, we’ll have fixed that second assumption in the algorithm. Thanks for the patience!
On the 20th of the month, Apple has their quarterly financial conference. I am certain you will get a third data point at that time regarding iPad sales.
The number of IP addresses each iPad logs in as will be a function of how long the iPad has been in use.
My older laptops have been online with thousands of different IP addresses.
My iPad has been on about 5 IPs so far. By this time next year it will have been online with a hundred or more.
How do you discriminate between my iPad’s 10th IP and a new iPad’s first IP?
wow!! this is amazing.. can you compare the sale of iPad against sale of PCs.. just curious, if the buyers are opting for iPad instead of a PC..
I was sure that CA would lead, but not by this much ..
Why aren’t you guys going against cookie count ? I find it strange you are counting them by unique ip.
thanks for this great map, but can i request to add a worldwide map? I bought online an ipad from US and i am using here in japan as i couldnt wait and it wasnt that expensive, i love to see how many more people are using the ipad in japan, asia and around the world.
Did you take into consideration how many IP addresses were from outerspace to determine which alien life-forms purchased iPads too?
Quacks.
Daniel, thanks for collecting these stats for the community.
I may be misunderstanding how your stats are gathered, but as you can only see 10%-15% of the internet, it would not be surprising if it took a few days before a new iPad hit your stats. However over several days, as the iPad user browses different sites, I would expect nearly every iPad to be recorded on your network.
Would it be better then to live with a lag of a few days but not multiply your raw stats by anything – otherwise after a week when the initial 300,000 iPads eventually are recorded by your stats – you will be reporting over 2 million iPads…
If you could give us raw stats or explain why your methodology needs a multiplier it would be more useful.
Curious if you have tracked the Nexus One (Google Phone)? After 3 months, sales numbers are sketchy to non-existent.
Is there any chance you’d just publish the counter of IP addresses? People could interpret the results however they wanted, but at least it would be a concrete number.
It seems that your numbers are still probably 10%+ high. Apple annouce 450k sold as of yesterday morning, it is very unlikely they sold an additional 118k iPads in 24 hours.
it’s pretty obvious after Steve Job’s revealed the true IPad numbers sold that your algorithm and IPad identification system is flawed. I suggest you take down this website since it’s been proven your business can’t accurately count unique IPad ident. numbers. Don’t waste our time or prove Steve’s numbers wrong.
It seems like your estimations will continue to overestimate. The average unique IP’s number should probably be a function of time, as people own their iPad for longer time, they get to bring it to more different places.
After having my iphone for a week I had probably been to 2-3 different LANs, but after 1 year it’s seen 50+ different LANs.
You got no credibility with these wild numbers.
None, nada, zilch
Sorry, but I think this is ridiculous. Your numbers will be off. The longer people have iPads, the more they will be used and the more they will be used with different WiFi networks. To say that there will only be 2.73 IP addresses per iPad is very very very stupid. I know it might be your best estimate for now, but that doesn’t make it valid. Your counter is currently showing just under 695 thousand, I’d put my money that the actual number is closer to 550.
This is pretty cool. Thanks for doing it.
Quick thought: I’m assuming that when SJ says, ‘We’ve sold 450K iPads,’ he’s counting ones which have been ordered but haven’t been shipped, and ones which have been shipped but haven’t arrived yet. Has Apple released any data on what % of sales are brick&mortar vs UPS? That should allow a way to adjust the numbers for, say, 3-5 days lagtime (before you’d see them) on the orders but none on the walk-ups.
You fail to take into account that Ipads are vanity goods, mobile vanity goods. Those are shown at work, visiting friends, relatives, Starbucks etc etc. Hint: The unique IP / Ipad number grows each day, eventually Ipads have been seen everywhere and used every dynamically assigned IP available.
I would like to second two earlier requests if at all possible:
1) The ability to see state break downs ordered by ascending people per iPod.
2) The ability to see a list of how many were sold each day since launch day – I’m fine with you back adjusting the earlier estimates.
Remember, it’s a pretty hard multiplier factor to guess how long people wait to go online with their iPods and how many IP addresses they use and that could well shift over time. Will you be able to track the iPods that will use cellular networks for browsing?
Thanks,
– Jeff
I think you may have a problem with your methodology here. Suppose Apple sells out of inventory and so there are zero sales each day. I suspect your numbers would still be going up by 50,000 to 100,000 per day. This is because you calculated the average number of unique IPs per actual iPad at 2.73 based on the two data points. Which was valid at the time. But that number is constantly increasing. For instance, when all those average people who use their iPad on 2.73 IPs go visit a new airport or coffee shop, then the average number rises to 3.73.
But you DO have interesting data, and it is still valid in, say, a relative way for how many people in each state and country are using an ipad. It is useful in other ways as well. I agree with the person who asked you to publish the raw counter of unique IP addresses per day so everyone can see the data for themselves. That would be great! Thanks.
I think you should change your state listing order.
Right now, it appears that all you are doing is showing the order of the states with the highest population. For example, last night when I checked last your top ten states are states that are in the top 15 most populous states. Seven of your top 10 are also in the top 10 most populous states.
I would suggest you use your other statistic – population/iPad – in each state to show the top states. The District of Columbia that shows up as the top (non) state (with 357 persons/iPad) and Hawaii in second place (with 531 persons/iPad). California the most populous state is actually third and New York (3rd most population) is actually 10th.
Curious but more interesting and not just reflecting population.
Perhaps you can relate to number of Apple Stores in a state?
Unfortunately your occasional substantial lowering of the projected sales impacts the functionality of the count.
This has now become my favorite website, eclipsing even MacRumors. But this morning you did a major readjustment, dropping the sales from 900K to 600K right as I was waiting for it to pass the one million mark. Yet you made no comments of this on your website.
Please update your website with such points – it’s still really exciting, and I think it will be a lot more exciting when we do the count down to one million (which I assumed would be early tomorrow based on your prior estimates.)
Also, as always I would love it if you’d show the estimates sales for all days since launch, obviously including back adjustments.
Thanks for a great source of entertainment,
– Jeff
Would it be possible to extend the iPads vs. time chart further into the past (with the revised IP-address-per-iPad parameter implemented)? As interesting as it is to see the daily increase and plateau periods, the growth of the user base over the iPads full time in public is more relevant to most people.
Thanks for putting this together!
The substantial lowering of the projected sales tend to impact the functionality of the count.
Wow a 140+ decrease from this mornings numbers – did your system hiccup or were you off all weekend?
I noticed you adjusted your iPad tally again today. Late Sunday night it was just under 850,000 and this morning it read 695,000. Care to enlighten your curious followers as to your strategy?
Yes indeed, we had to readjust our numbers again. The good news is that this should be the last time: our engineering team has managed a workaround to the cookie-squashing, so we’re actually seeing each iPad by cookie. No more IP address assumptions (@John, your point about IP addresses per iPad increasing over time is absolutely correct, and happened a lot faster than we anticipated), and our engineers should have a post up on their blog about how they pulled off the workaround in an hour or so. I’ll update with that when it’s up.
how many times are you going to rejigger your numbers before you actually get a useful alorithm. It’s nice to see the numbers but it’s pure speculation. You should take down the site.
Daniel – I have been trying to follow these numbers from the very beginning, but the constant readjustment is really throwing the numbers way off each time. Can you please publish *exactly* (step by step) how you are counting the unique iPads? That might help us to understand the numbers a bit better. Thank you.
Guys, I *LOVE* this site! Don’t listen to people complaining about the accuracy. The most you’ve been off was about a factor of 2 (compare this to more traditional methods of estimating sales), and the relative changes are still important. I also believe that you’d continue to be able to refine your estimates, such as the rate of average growth in IPs/iPad. And now that you have cookie support, you can even display the average IP addresses used per iPad.
In any case, glad this is the last major adjustment, and keep up the good work. 🙂 Now that you finally have accurate counting nailed, how about enhancing the GUI?
Thanks,
– Jeff
I second Jeff. I like the site. It’s a fun experiment, and I appreciate you guys updating the figures as you sort out better methodology. Nicely done!
This is stupid. Look the counter carefully. Why is it counting 3 seconds then waiting for 1 or 2 seconds. How could you explain this thing. It’s a simple JavaScript setTimeout and setInterval exercise.
And also, as a developer, I look at the sites communications using my Firebug for any server and I saw nothing. I mean there is no Ajax call to any server but the counter is still counting.
This is a stupid and cheap trick.
I also second (third?) Jeff. This is good stuff that no one else is doing, and if it takes a while to get it more exact, no problem. Appreciate the effort that goes into it… and bet it brings a lot more people to your site.
Where is the new counter? I don’t see it on this link?
With 38 millions population and 145000 ipads in california, we have 260 population per ipad and not 584 as stated in your ipad stat webpage
Keep up the good work, guys. Greatly appreciated. Thanks a lot.
Keep up the good work. It’s really interesting to see the evolution of your attempt to count iPads. If Apple announces the millionth iPad, it’ll be interesting to see how close your count matches.
You are still way off. We already know that the Ipad sales dropped from initial 300k first day to 30k / day for the succeeding 5 days. It’s probable that they are currently somewhere between 10k – 20k range now, not in 40k as your counting indicates, probably about 550k-600k ipads sold, rather than 760k you’re counting.
@sam remember that the 300k number included the online pre-sale, so the declination is a bit overstated.
@ ChartsChartsCharts
I don’t think what you pointed out is the most relevant thing to note. The fact is that we know the sales dropped to 30k / day and it’s quite a stretch of imagination that the sales would have risen back to over 60k / day as suggested by the tally. Most likely the sales are well below 30k / day at this point.
I’m thinking that the count of iPads coming on line is a bit “lumpy.” Obviously a huge number came online on the 3rd, but supply is somewhat constrained after that. Even so, by the following Wednesday, 3 business days, SJ claims another 150K, so thats 50K per day. But remember he’s counting shipments to BB as sales. Then yesterday was the 12th when Apple was promising to make a second shipment of pre-sales. So we’ve had 5 more sales days, counting Saturday, plus the Apple shipments on the 12th. Its not unreasonable to think there are as many as 750K total in customer’s hands. There’s been speculation that SJ would announce the millionth shipment. I expect to hear it next week, but wouldn’t be surprised if they hit it this week. Personally, I’m still waiting for my 3G version later this month.
Oops! Apple says only 500 K shipped in the first week. So if you are seeing anything more than 10-15 K devices per day you are off.
Q : If you switched to a cookie method midway, how would you know if you had counted a certain ipad already, or if it is new!
yeah by then end of the day today the number should be some where around 552k
Since it looks like they are now selling 13k a day
For all the complaints about your estimator, your numbers are too conservative.
I wouldn’t put much relevance in what Jobs reports, Apple is notorious for under reporting just about everything with regard to Apple until earnings report. I expect Jobs to report 1,000,000 iPads sold AFTER its earnings report on April 20. Apple is going to be deluged with questions about iPad sales during the conference call, and they will want a different venue for that type of announcement.
I’m pleased to see your use of cookies to determine number of iPads in use. It takes a lot of the wiggle out your assumptions
@Gregg Thurman,
It’s a SEC violation for Apple and Steve Jobs to say ANYTHING that is not true — that affects the stock price of Apple shares materially.
Apple and Steve Jobs don’t have to say anything until their quarterly earnings report. But if they say anything prior to the quarterly earnings report — those information better be accurate. Annoucing 450000 ipads sold in the first 5 days and 500000 ipads sold in the first week — affects Apple share price — therefore these information better be the truth.
So now that you guys have setting cookies in safari figured out, can you guys reset the counter?
So apple sent our a press release saying that sold more than 500,000 in the first week. I assume they are counting sales for the first 7 days. On Thursday they said they sold 450,000 up to the end of day Wednesday. So that would mean they sold another 50k in 2 days (Saturday April 3 was first day, end of day Friday April 9th would be the end of a week)so looks like they are averaging around 25k a day. So the today should be around 625k (500k + (5days x 25k)) your number says 787 right now. I think its way off your should reset and let the cookies work.
Apple said 500k this morning
You site is close to 800k this afternoon.
That is a pretty big discrepancy.
Do you think Apple is sandbagging by that much, or do you think your model has a couple of hundred thousand iPads worth of error in it?
I wish I could think of a way to help reduce the error in the model, but I don’t know enough about it to offer helpful suggestions, and I might not be able to in any event.
827,563 iPads = 307,000,000 people . 1 iPad = x people. 307,000,000 / 827,563 = 371 people per iPad. That is lower than any state. The only exception is DC (pop 600,000). Are the state breakdowns hopelessly out of date?
are you guys going to fix the counter? is is way off
This page needs to be put out of its misery – it’s a complete and utter joke!
@Altavistagoogle, I’ve got to check, but I suspect when we changed over the main counter to use cookies rather than IP addresses the state counter kept plugging along with the old style… will get that fixed.
@Ron, not sure. I doubt Apple would put out disingenuous numbers, so I’ve got to assume that our method is imperfect. It’s an estimate, so unfortunately we’re going to be off no matter what. I think we’re as accurate as you can get without actually tracking the sales at all Apple Stores and Best Buys.
@Pete, we’re trying to do something of a rolling reset transition from IP to cookie-based.
Cont you have the number of ipads that your recorded before you started to use cookies? just clear that number off of the current total.
@Pete- Way Off ? How do you know? I believe it’s reasonably close.
Your math is flawed.
1) 450K and 500K numbers are just milestones, not actual data points at a specific time. SJ said “Over.” We have no idea how much over. Last Sept @ iPod event, SJ said Apple has sold over 30 iPhones. Two weeks later the quarter ended and Apple reported almost 34M iPhones sold with 7.3 in the quarter. So when SJ said over 30M it was actually around 33M.
2) Calculating a run rate between two unknown data points, and then multiplying… magnifies the error. It’s very sensitive. Sales could have been 550K which would imply 50K per day, or 250K change which would be close to counter’s estimate. It could have been 480K / 500K, or 10K per day, 50K change…. Impossible to tell… So, that method gets us nowhere.
3) Even if we had exact data points, the run rate on Thursday & Friday wouldn’t remain constant over the weekend. It would be much much higher, perhaps 100K per day, we don’t know.
We do know that iPad was the feature of Best Buy’s weekly ad last Sunday, and on home page of its site. Store manager I talk to, said received a large shipment of units which were sold in no time. Sales were very strong, and have remained to be, when the store has stock.
He was out yesterday, pinged me back saying he received a drop shipment late in the afternoon. Retail availability has been in constant flux. It’s likely that there was minimal stock for those two days between announcements resulting in unit sales under-representing actual demand. There are just a number of flaws that make it impossible to do a back of the envelope calculation to arrive at what sales actually were yesterday vs the ticker.
@Turley Muller
The difference is very small. Steve Jobs announced the 30 million iphone mark three weeks away from the end of the quarter. 3 weeks means 1.7 million iphones for that particular quarter. Channel inventory also went up by almost 600K iphones for that quarter. Count them all together — the exact sale number at the Steve Jobs announcement would have been 31.4 million iphones, a 4.6% under statement.
A 4.6% difference means instead of 500K ipads, it’s 523K ipad.
How do you decide what state to put an iPad in? For example, many Virginians are using their iPad at home in Virginia and at work in DC.
I see another small drop. We were over 888K and we are now back to 887k? are you correcting something?
I’m likin’ the rhythm the counter is kicking out right now (April 15, 1:30 p.m. central time). Two quick counts, then a pause, then three quick counts, then a pause, then another three quick counts. Then the whole rhythmic sequence repeats itself exactly. Been doing this for a while now. It’ll shift to a new rhthym eventually and stick to that for a few hours. It’s really remarkable how Apple has been able to train iPad purchasers to go online is synchronized fashion. Guess the reality distortion field is more powerful than anybody thought. Someone should check Apple stock purchases for similar rhythms.
Look at the script they use: (refresh only seems to change the iCount initialization value)
google.setOnLoadCallback(drawAdoptionByHour);
/*
tRateC: 83730.56,
tRateP: 83524.48,
yRateC: 77379.2,
yRateP: 77152,
dRateDiff: 21.119999999995,
upCountRate: 0.57244444444445
*/
var upRate = 0.57244444444445;
var iCount = 843344.88888889; /* Last updated [1271357701] */
function cup() {
var iForm = ”;
iCount += upRate;
iForm += parseInt(iCount);
var rgx = /(\d+)(\d{3})/;
while (rgx.test(iForm)) { iForm = iForm.replace(rgx, ‘$1’ + ‘,’ + ‘$2’); }
document.getElementById(‘hpCount’).innerHTML = iForm;
}
var cupt = setInterval(“cup()”,1000);
I think this is really cool. I read over the comment and I agree with Chitika Research while results are off I don’t think difference is more than 10%. Sure we can’t tell exact number, also we don’t know how many preorders have been taken for iPad 3G already since those units are not on line just yet.
Sure we can try to make assumption in regard to Data reported by Apple selling over 450K units and then over 500K but we don’t know how much over those numbers. Also I bet they sell more units on Friday through Sunday vs Mon through Thursday.
Let’s wait and see what Apple has to say on 20th. Numbers will be revised upward or downward. Also I bet when iPad 3G ships numbers will have to be revised upward.
Great idea on digging the search patterns and produce an estimator.
I don’t follow your math.
If 19% of the 951,874 iPads sold have gone to Californians, then Californians have purchased 180,804 iPads. As the population of California is 36,961,664, your “Pop. per iPad” should state 204 people per iPad, not 454.
What am I missing?
Gr@w\!x
A question about population per iPad. As the number of iPads sold approaches 1M, wouldn’t we expect population per iPad in US overall approaches about 300 per iPad (with a total of 300M population)?
Can’t wait to celebrate when the Chitika counter hits 1,000,000. Should be about 5:30 p.m. central time. Got some bubbly. Then we can start making bets about when the iPad’s real sales will hit one million. I’ll go out on a limb and say May 10 (and Apple will announce 700,000 iPads sold during their April 20 conference call).
As the national counter creeps towards 1 million, the math get’s really easy and the discrepancy between national and state iPad/population ratios is clearly evident, even without using basic algebra.
Hm. If I take the 2006 census populations per state and divide by population-per-ipad, I get a total of 436,220 as of now… factor of 2.3 ?
1,018,225 was about the highest Saturday night, April 17, 2010; then reset to about 720,000, at about 11:15 p.m. CST.
Yesterday you were at 1 million and now you are some 700k.
Can you please explain why you changed it.
Regarding reducing the count 1 Million+ by 300K last night, are you now counting only units with the “new cookie” and assuming the most sold units have the new cookie?
Well, I’ve been watching with some interest, reading the posts, mostly wondering why people need to be so rude. Yesterday I watched the counter climb over 1M. Today I look in on it again and it’s back to 750K, which, in and of itself, doesn’t disturb me at all. What I don’t understand is why the sudden huge drop is not accompanied by any explanation whatsoever.
Pretty significant revision down today in the number. Think it was over 1mm when I went to bed last night. Any explanation out there?
It’s irresponsible to without any explanation, alter your methodology to such a degree that the number drops by 300,000, particularly when the change came shortly after reaching the magical one million mark. You owe it everyone who has been tracking your work for weeks. Credibility is an easy thing to lose. Once lost, it’s very, very difficult to ever regain. You’ve got a tough road ahead, better get to it.
The 3g’s should be shipping this week to arrive on time. If they ship Tuesday and Apple counts shipped pre-orders as sold expect to hit 1m shortly.
So it is possible Jobs will be able to make the 1m announcement as part of the conference call. However I would expect Apple to save it as a separate press release. Tuesday will have enough good news with great numbers. Then save the 1m announcement as a way to focus press coverage on the arrival of the 3g going into the weekend.
darn! another major correction without any explanation whatsoever. even an ardent follower like me is getting a teeny weeny bit pissed now!
I’ve seen too many resets that drop the running total as much as 200,000 units overnight to have any more confidence in your effort.
Friday evening it appeared that Apple was poised to see the one millionth iPad overnight. Sunday morning your total has reset (once again) to less than 750,000. Your margin of error is way to high (even with your tweaks) for one to have any confidence in your results.
I applaud you on your effort but think your internet sample is far too small to be of any use.
The resets have finally been too much for me as well. For a solid week now I’ve watched the count start near 900K, and then every single day drop by at least 100K. I stop watching for 6 days, and when I come back, it’s STILL in the 700Ks! What’s the point of watching this if it always says the same numbers? You guys have really hit the “half decaffeinated” coffee point and are pleasing no one – it still doesn’t match the real sales values, but since you reset it every day there’s no trend to follow either. At this point, it’s just a random counter counting up from whatever you guys set it back to.
If you had made a post each time you changed the count with an explanation, I’d still be following you, but at this point, even I say this has become utterly useless.
Here’s a more reliable way to track iPad usage, IMO. And doing some ratios based on what Apple has disclosed, a more consistent estimate (though still rough) of iPads in the wild can be derived.
http://www.netmarketshare.com/operating-system-market-share.aspx?qprid=42&qptimeframe=D&qpcustom=iPad&qpsp=4108&qpnp=17&sample=20
I think the link on my last comment will not advance in time automatically. You can change one of those parameters for it to advance, but here’s a link that will advance on its own:
http://marketshare.hitslink.com/report.aspx?sample=17&qprid=42&qpdt=1&qpct=4&qpcustom=iPad
Sorry for the triple comment, but I think most readers would be interested in this comparison:
http://marketshare.hitslink.com/report.aspx?sample=20&qprid=42&qpdt=1&qpct=4&qpcustom=iPad,iPod,Android%202.1
I live in Asia, and therefore I am able to see the counter at the very end of the day in USA (when most people are sleeping). Right now it says, that more than 70k iPads are sold today (on a Sunday). Even though I wish it was true, that sounds very unlikely in my ears.
According to your notes above, I think that you still haven’t fixed the largest error in your assumptions, namely that you can just normalize the hits according to what fraction of the internet you see at any given time. This model assumes that there is no correlation between the types of sites hit by iPad users and the sites that you “see”. So I think you need to be using a different multiplier.
I can tell you have yet to get this right, because your rate is still moving along pretty much like it has been, and yet the rate has been shown to be at least a factor of two too high. (You keep having to reset your counter downward by a significant amount… but the rate doesn’t change too.) The good news is that since this is a linear model, and since you’ve gotten at least two official counts from Apple in the past, you could “back out” the appropriate multiplier and get rid of this offending assumption.
Go ahead and do the calculation to fix the multiplier, and I’ll wager you get a more accurate result going forward.
Thompson
Guys I’ve told you before, each Ipod continues to pop up in your stats several times as they are moved and showed around.
The trick is not in cookie tricks. What you need to do is analyze your data. You need to find out how often each device is counted. So first day 5 devices is sold. Those generate 8 counts. Next day 5 more device are sold, generating again 8 counts. In addition those that were sold day one generate 4 more counts.
So what you need to crack is how many new counts each device generates first day, second day etc. It’s probable that the number of new counts will drop close to zero after few days (when the device is taken to all the places a person frequents).
Simple as that 🙂
Hi,
You guys are doing a great job! And fun too:)
Thanks for giving us a lot of fun with this counter thing.
I presume you have some Ipads and your adjusted the counter by tracking them already.
Cheers:)
This is obviously not a lab I would want to have do my market research.
I followed closely the first time it went over 800,000 and then to a million.
I have lost interest and now check in periodically just to see the reset.
I assume a lot of comments are censored. Doubt if you will see this one.
It would be great with a plot showing how many new numbers there has been each day for lets say the last 30 days. This graph would probably also be corrected if similar corrections like we saw a few days ago occured. This would give an indication of whether sales are slowing down or not.
No fair!! You list states’ ranking without taking into account their population. Obviously California is going to be first; it’s by far the largest state in the country. If you take into account what percentage of the population of each state purchased an iPad, the ranking would be more like this:
1. District of Columbia
2. Hawaii
3. California
4. Minnesota
5. Alaska
6. Massachusetts
7. Colorado
8. New Hampshire
9. Nevada
10. New York
11. Washington
12. New Jersey
13. Oregon
14. Arizona
15. Connecticut
Oh, and Wyoming would still be in last place.
911,425.
Is this going to reset tonight, back to 723,000?
touché 🙂
People! Chill already, jeez. This isn’t life or death — it’s pretty much just for our collective amusement. The count keeps changing because they keep trying to refine their methodology. Wouldn’t you rather that, than they just leave it where it was and let it be that much farther off? If you’re that annoyed about it, don’t keep checking the site. No need to get all huffy and rude, c’mon.
FWIW, I think Apple is going to announce 1,000,000 units sold this next week. The announce will come one day after they receive the data, and that will be one day after they hit the milestone.
Whatever, if Apple announces 1,000,000 units sold prior to shipping the WiFi + 3G model I think they will sell 3,000,000 for the quarter.
FWIW, ITCWRTE.
(For what it’s worth, I think Chitika will reset this evening.)
Almost there… I think it will get reset a few more times until Apple announces the 1M mark and then then Chitika will also be there…. what a coincidence..
It’s at 1 million now with no reset. I estimate 875,000 sold (based on 500,000 first week & 25k a day after)
it’s very interesting that apple declined to state ipad sales during their earnings call last week. Ether sales went down or the are
waiting for the 3g pre order sales to bring them over 1 million.
Many assume that Apple will announce 1M sold. I don’t think that’s necessarily the case.
3G sales are starting tomorrow, and I’m eager to see how this will impact the counter….
Plus, the ipad availibility is still very low
Everyday single day the “Today so far” bar indicator shows more units than the previous one. Although I don’t doubt how well the ipad sells, I don’t see this has the actual reality. Am I missing something?
Why not Turley?
I think they are waiting until Monday or Tuesday to announce two million
humm, I have the feeling my message has been deleted.
I would like to understand why the bar chart showing “today so far” is always, always, higher the previous day.
That would mean that each day Apple is selling more than the previous day.
That doens’t fit the low availibility and the upcomming of the 3G.
Again, the number must be wrong….
almo
“I would like to understand why the bar chart showing “today so far” is always, always, higher [than] the previous day.”
“It’s higher each day because these guys are counting the same devices over and over and over again…”
In order to grow sales quarter over quarter (which the iPod did until it hit market saturation, and the iPhone is currently doing), sales must increase day over day.
Sales of an anticipated new product goes through a cycle that starts with a surge (early adopters) then trails off until marketing and reviews by the early adopters excite the masses into adopting the new thing. Selling more each day is the sign of a product gaining acceptance from the masses. This phenomenon has absolutely nothing to do with a “flawed” model.
Given Apple’s success with the iPod, followed by the iPhone, I’d be surprised if the iPad didn’t grow sales day over day, even at this early stage in the cycle.
I knew earlier that your algorithm was off to the low side. Now I can see that it is still tracking at a low rate… even though a significant number of new iPads must have come online today. I no longer believe that you have an algorithm at all.
It seems that all you have is a steady state counter (that slows down a little in the wee hours of the night) and all you have done is tweak the rate based on truth data released by Apple.
In short, it is a big scam.
Thompson
actually Bar chart has not updated since 4/26 and so has the figure of how many new ipad seen today.
I think Chitika has milked the ipad publicity all it can and now just is not going to
give any news figures until Apple gives new number.
What a jib.
At 5:00 p.m. tonight, the 3g was available in Apple Stores. The line outside our Apple Store in San Antonio’s North Star Mall (one of two Apple Stores in San Antonio) had a line that some thought was as long as the line when the “wi-fi” only was launched. I’m not sure it was that long, but it was long. 16 GB version sold out in one hour; 32 GB sold out in 90 minutes; still had some 64 GB available for Saturday. I would assume similar story across the nation regarding “wi-fi + 3g” and yet the speed of the counter did not change from earlier today, before the 3g iPad was on sale. Just saying.
A very interesting bit of research you are undertaking.
A suggestion: If the bar label were changed from “today so far” were changed to “to date” it would probably reflect your intent more clearly.
A very interesting bit of research you are undertaking.
A suggestion: If the bar label were changed from “today so far” to “to date” it would probably reflect your intent more clearly.
My Fedex driver said she had already delivered 4 (Friday) and had more on the truck. I’ve raised my iPad units sold estimate to 3,250,000 for the June quarter.
well today they announced 1m ipad sold. which not surprisingly matches the 20-30k / day estimate after the launch i have been telling all along. now you believe?
Official: Apple has sold 1 Million iPads in 28 days http://goo.gl/jxX1
Ok, 1 million on the 28th. That would mean that according to the counter they have sold 830K more in 6 days! Hummm, I want to believe….but….
They might have sold around 400K of the new model.
And also maybe around 200K of the old model.
So 830K in 6 days, so they were at least closer than they used to be.
i think you guys need to adjust your tracking mechanism coefficients or something. apple announced that they sold their 1 millionth ipad sale today (5/3) while your counter almost does twice that (~ 1.8 million last time i checked).
We were a little ahead on the 1 million mark – we called it last Sunday, it actually hit Friday – so when Apple announced 1 million (Friday afternoon) we were at about 1.18. The 3g model is giving us a bit of an issue; if early estimates by others are right, then 300k or so 3g units sold over the weekend. Figure 100k or so WiFi units, and we should be at about 1.5 to 1.6 million (with the 180k we were over Friday).
We’re working on the 3g tracking – it’s coming in a bit too high. We had the WiFi down almost to a science, and will get there with the 3g ones soon.
Incidentally, we do intend to separate the two versions; only issue is the iPad user agent has no indication as to which version it is. We’re trying IP tracking to find the iPads that are coming from AT&T addresses to separate them out – a rough way of splitting them, but a viable first step.
Thanks all.
It would seem that you are 500k to 700k over the real count. I don’t see how you get to 1.5 million from just passing 1 million on Friday. Even best case would be only 1.3 million (assuming 300k over the weekend). It seems more likely that most of that 300k would be included in pre-orders, etc that just pushed the iPad to a million.
A conservative count is about 1000x more useful than than a hopelessly inflated count.
300k estimates you are hearing includes pre orders and what was sold by the end of day Friday. The press release went out today, not friday. They probably only sold 20k additional 3G over the weakened. Your numbers are not even close to accurate. I would estimate by the end of the day today they would have sold 1.15 million wifi & wifi + 3G
Apple announced 1m on Friday and on Tuesday your method calls 2m. Are you nuts?
I’ve sold 1.4M magical iPads as of today.
Sent from my iPad.
I have the feeling you are 200k over the real count.
You statment was good yesterday:
“We were a little ahead on the 1 million mark – we called it last Sunday, it actually hit Friday – so when Apple announced 1 million (Friday afternoon) we were at about 1.18. The 3g model is giving us a bit of an issue; if early estimates by others are right, then 300k or so 3g units sold over the weekend. Figure 100k or so WiFi units, and we should be at about 1.5 to 1.6 million (with the 180k we were over Friday).”
But this 3G counting is messing up with the numbers again. How long do you think it will take to fix it? You should correct it before we pass the 2 millions….
“so when Apple announced 1 million (Friday afternoon) we were at about 1.18. The 3g model is giving us a bit of an issue; if early estimates by others are right, then 300k or so 3g units sold over the weekend. Figure 100k or so WiFi units, and we should be at about 1.5 to 1.6 million (with the 180k we were over Friday).”
It seems you guys have a hard time even with normal estimates. No wonder your counter is so inflated.
Why don’t you drop those 200k over? Drop them, figure it out and reajust. Better than crossing the 2m at the wrong moment.
At this point the estimat is probably a lot more than 200k off. After the initial weekend the Wifi was seeing at 28,500/day. (Actually very good) Let’s assume the initial million doesn’t include any of the 3G iPads and that the sales rate doubles after this weekend.
300k + 26 days @ 28.5k + 300k + 2 wifi days @ 28.5k + 2 mix days @ 57k ==>
1.5 million ….
Note that is likely a very high estimate since it is like some of the 300k 3G were included in the 1 million sold sometime on Friday (preorders etc) and the rate probably isn’t double now. So we are probably looking a correct number being in the 1.3M to 1.5M range.
That isn’t a bad number to be at 32 days from the release date.
Thanks Chitika. This has been an adventure over the last month and this site has made it even more interesting. Right or wrong you are out there for all to see and I like risk takers, whether new ideas or new technologies. Take your lumps if you must but from my view I would rather take a few than just spend my time tossing stones over some wall hoping I’ll be noticed. Make your changes, recalculate and adjust, fine tune the model. That’s what life is about.
I’m 100% with Rocky. At first we all thought that this thing would be perfect but like most things it’s not and it needs to be improved and I don’t see a major problem with that.
ohhhh nooooooo, that’s not a good way out….
Closing the traker, I’m sad! and this was still giving a quite precise idea of the sales…
I’m sorry that you’re closed, but it’s understandable. You’ve done a great job, and I hope that you inspire a new generation of people who use their own web statistics to say clever and insightful things about new technology. You will be missed.
A few words on the why would have been nice….
After a week, not even a goodbye word…
and the counter is still moving…
Its not so good in my thinking
why has the tracker been put to rest??
The ipad mobile traffic is definitely going to increase everyday!
good info !. thanks for sharing.