Google is going to really hurt eBay, just when they need it the least.
eBay is stuck with a massive Web 1.0 legacy, a walled garden that is struggling to keep sellers. eBay, not being technology-oriented (it's all about community over there) is not going to innovate their way outta this situation.
Let's see, GCheckout offers:
That's all great for rapid uptake. By discounting to AdWords users, Google is buying their way in to payment in a HUGE way.
More than that, they are ready to hook up anyone and everyone to immediate internet commerce. What's the one thing eBay had that facilitated that? Reputation management.
Google will build a portable modern reputation system - i.e. "can I trust this vendor, let me see what everyone who buys with GCheckout says". Google should have the best trust/reputation system for shopping, and it's gonna be global across the internet.
That goes to the heart of eBay's value. Google is already cutting a ton of margin out of eBay's heart, and with a reputation system they'll take away a major thing eBay had done well in its walled garden and Google will democratize it.
Google Base, Froogle, AdWords and AdWords CPA are more of the chess pieces that Google has on the board in this game. GCheckout is the wheel they revolve around. They may be immature, but Google's opening moves are starting to come together.
All of that plus free spells hard times for eBay ahead.
Terry, this is Meg again - are we still OK for $23/sh?
I've been trying to figure out Amazon's Browse Nodes and how their salesrank really works so I put together a very quick and dirty Amazon mash-up. It's Ajax approved, full of JSON-y goodness, and even uses XSL!
Try it at: http://texsy.com/amazonCats.php - remember - it's just a hack I built for research, but it's kinda fun.
If you click on "Baby", you'll see the top selling baby products at Amazon, and you can refine. Some categories don't show any results, however.
Greg Linden puts some estimates up for how many servers (and how much memory) Google has in their world-wide supercomputer cluster. Basically he thinks they have around 450,000 servers and perhaps as much as 4 Petabytes of total RAM.
I think that's right on for the server number and high on the memory - my estimate would be 500k servers, and just under 2 petabytes.
Greg also writes that Google is adding 100k servers a quarter now - but I think that's a typo - because in an earlier post he estimates it would be about 25k new servers per *quarter* - and even that's probably high. I don't think 100k/quarter is realistic.
People were wondering if all of Google's $300M+ quarterly capital expenditures (CAPEX) were going into more servers. But now we know now from Henry Blodget, that a very large chunk of the CAPEX is accounted for in Real Estate, not hardware.
So I think Google might be at 500k total servers now, and yet they are still in a significant server crunch overall - they constantly need to reclaim & consolidate boxes just to keep up with their own demands.
I base the total server number on extrapolation from when they tacitly acknowledged that they had over 100k servers, and on the number of world wide datacenters people know about. The fact that Greg's estimate is in the same ballpark of what I thought is also re-assuring.
As for the total memory, I'd bet the average memory per server is not 8GB - perhaps 4GB would be the baseline. New boxes coming in will probably be 8GB. And of course, every "box" is a dual CPU system.
So the total memory in the Google cluster - across all datacenters - would be more like 2 petabytes, IMHO.
The interesting question for Greg would be: how many boxes will the new 34,000 sq ft. datacenter in The Dalles, Oregon hold? Remember that Google has custom power supplies to minimize power per server. These dual cpu 8GB Xeon boxes don't take 300W when Google builds them.
My guess is that they are building the datacenter to dissipate as much as 200W per sq ft - being Google (aka Design LLC), cutting edge and all. So probably 60k more boxes in Oregon coming in time for the Christmas rush. That alone shows how hard it is to add hundreds of thousands of boxes, however...
I guess up is down now. Cost-per-click is being criticized by the likes of Henry Blodget as the wrong model for the blogosphere. (In other news, Henry has signed up with Federated Media - John Batelle's ad network - mostly CPM...)
Does anyone remember 3 short years ago when CPM ads were all washed up? This CPM comeback (sentiment-wise) is driven by backlash against the success of Google and AdSense in shaping the monetization landscape of the internet.
CPM ads comeback (performance-wise) has happened due to macro ad market trends - a strong advertising market overall, and brand advertisers finding success in CPM buys. Unfortunately, CPM still has two major problems:
Furthermore, CPC and cost-per-acquisition ads (CPA) have built in accountability for direct marketers - you spend more when you make more since you actually can tie spend to revenue.
Upshot: What's old is new again, and the internet is diverse enough to accomodate huge ad spending on many types of media. CPC and CPA have distinct accountability advantages over CPM, so PPC and CPC are not going away, and CPM may be strong but it's not ascendant.
This exchange in the comments at Henry Blodget's blog amuses.
Comments first Posted by: victor | June 21, 2006 at 12:18 PM F**k Second You are becoming a strong opponent Victor Posted by: KING TROLL | June 21, 2006 at 12:24 PM
The internet is a big place. Big enough for the people whose main pleasure in life is getting the first comment in on a blog. Go crazy guys....
Battelle rightly predicted that online ad spending would grow faster than forecasts, and he collects on that prediction today.
Greg Sterling also talks about the offline ad world may suffer more in the next decline in the ad market.
I think there are 3 major factors working here that led to the under-forecasted growth in online ad spending.
In other words, online ads are more effective and the shift in spend will continue until the cost of online reduces the ROI to offline levels. Moreover, as more of the world gets broadband, and spends more time online, reach expands and advertisers go after the bigger audience.
Display advertising on the Internet currently accounts for 6.7 percent of total ad spend. That's $10 billion out of the total $150 billion that's spent on advertising.
Clearly there's a lot of headroom for online to grab $30b more of that total spend.
But many have overlooked the cyclical aspect of the ad market. In 2004, the overall ad world exploded off of several bad years from 2000 - 2003. Coming out of that ad recession, advertisers have spent heavily to make up for the lean years. All that spend has been great for many forms of media. In the online world, that spend enabled Yahoo specifically to make gobs of branded/targeted ad money.
My prediction is that there will be a severe damper on things when the ad market starts to decline again - for whatever reason. Online growth will not escape the next ad recession.
Sterling believes that online will fare better because:
When the next recession hits, companies may shift more dollars out of higher-priced media to the Internet, unless of course prices have reached comparable levels online. Even then the perceived efficiency and measurability of online wins the contest.
Greg is probably right, but to me it depends on the level of the overall slowdown and the part emphasized above. Finally, the fact is that online ads are still new and hard to do for the big spenders, so when budgets get smaller, they might become chary of the online experiments.
The traditional offline world knows well how to hibernate through the bad times. For example, agencies lay off tons of people as one early measure.
The big question is whether all the online money chasing ads will have learned anything from 2000-2003. I'm guessing no.
According to Meg Whitman's keynote (she's the CEO of eBay), at the eBay Developer Conference:
1.3M people make their living selling full-time on eBay
Is it me or does that seem like a lot? Maybe eBay really is all about the community!
I've been trying to buy ads on MSN with the new adCenter, and it's incredibly flaky. I have to open Internet Explorer, of course. But even in IE it has a million little script errors, missing css/xml files, etc. The stuff just doesn't work.
Sometimes the interface doesn't show my ads, sometimes it takes forever to show my keywords. AGGGH!
OTOH, it's the closest web interface I've used that simulates what it was like to use Windows NT. When you have a problem, you just reboot - or in this case, restart the browser, login again and hope.
If anyone has tips for how to reliably use MSN adCenter, please tell me!
Simple fact: If you run any AdWords ads at all, you are INSANE if you don't try Google's fat, phat client AdWords Editor tool.
But here's a tidbit about Google's internal tracking that's pretty interesting, and I haven't seen it mentioned before: Someone at Google wrote a market-based prediction system (think Iowa Electronic Market) and internal people can buy/sell on internal Google milestones.
Here's the Google blog talking about their market-based prediction system...
Questions like Will Orkut have more than 1M users by June 30? can be tracked. In theory, if enough Google employees place their virtual money where their mouths are, the answers could be better than the product manager's powerpoint.
I don't know if that tool gets much use anymore, or if it's any good at predicting outcomes, but it's an interesting approach for a medium to large size company. I wonder how many other companies use anything similar?
Update: Reader Mike Linksvayer commented with a brilliant reference page to companies who are experimenting with prediction markets. Looking at the names, one thing comes to mind, however: Short 'em all and let god sort it out... ah, it's late...
The 5 year stock histories of SBC/AT&T, ComCast, and Verizon are not pretty, but this could be just the time to buy! If they can pay off the Senate, they might have some serious new locked-in revenue for the next several years.
Even if the telcos and cable companies don't win in the Senate, or can't leverage their lobbying influence fast enough, the fact that Ebay, Amazon, Yahoo and Google are all gonna fight the telcos on net regulation seems to set up the opportunity for an actual hedge. I.e. pit the telcos against the technos?
Hmmm, I don't know... maybe it's time to buy the equipment makers again?
MySQL is a great database and it has amazingly good built-in replication which helps for backups, load-balancing, performance and redundancy.
After my post yesterday with my searches in it, I got a couple emails asking how I set up MySQL replication. It's not hard, and there are some great resources out there.
Here are some references to get you started:
Spend a day or two reading those and you'll be able to sprechen MySQL like an expert.
|My Google Searches||Search queries to GotAds?|
In case you are wondering, I did get MySQL replication working, and "Floppy McChokelstein" is an old nickname for Phil Mickelson that I found hilarious. It no longer applies, though.
Batelle pointed to a review about Windows Vista. Reading these previews makes me more and more sure that my next computer will be a Mac. Vista is the most bureaucratically bloated OS of all time. The 'Ultimate' edition beta is over 10GB in size. It's written by bureaucrats for bureaucrats.
Here are some random datapoints on why Vista will suck:
What Vista won't do is make your computer feel fast. And since my 1yr-old 2.8GHz Win XP Dell with 1GB of RAM already feels like mousing through molasses, upgrading to slower is not for me...
So I think the Vista upgrade cycle is gonna be a great thing for Apple.
Hitwise data backs up a prediction I made late last year: Google is getting more dominant. Yahoo and MSN are falling further behind, to put it bluntly.
From Heather Hopkins' blog:
That's just insane dominance.
Furthermore, blogger complains if you try to embed the example code in an iframe. So I've had to host it at another machine and set the src of the iframe to that address. And for that I need another key - for the domain of the other machine... Not exactly easy to use.
In comparison, Yahoo's search APIs are much more complete and much more elegant and useful.
All-in-all - a very early look at something, at least they're not shy about putting out half-baked clams... which is a good sign.
Local search is a big deal for Rosenberg, obviously, and he's big on making the point that Google only tries to show "good ads" - i.e. other "Brand X" search engines would put a lot of bad ads on the "dog parks" search...
The barbie example was about personalization in that it's OK and perhaps expected for Rosenberg to search on that since he has a young daughter.
The [cheese recipes] search was an example of "one-box" refinement with Google Base. Including a box as below:
Google seems to have internal rationalizations working to convince themselves that they are not delivering "annoying", untargeted video ads with their new AdSense based video ad program.
In yesterday's analyst call, Jonathan Rosenberg, SVP of Products at Google, spent a lot of time rationalizing the Video ad experiment.
Proclaiming most other "Brand X" video ads "annoying", he emphasized the "static image" that users must click on to start playing the ads. He mentioned that feature several times in succession and his overall answer was much less confident than his usual short and to the point replies to advertising related questions. A lot more "ums", "ahs" and "uhs" sputtered out as he struggled to justify the relevancy.
He was pressed about how the ads are actually targeted - and the answer is pretty enlightening: - video ads are sold on a CPM basis in the auction so the advertiser is incented to bid on sites that the advertisers believe are relevant!
In other words, video ads are nothing like text ads, where Google determines what is relevant. Video ads are simply site-targeted CPM ads. Obviously, that is a lot less targeted then Google wants.
The fact is, they haven't figured out the model, but they know they need to, so that is why the video experiment is starting even though it's fundamentally against Google's belief in delivering relevant ads.
PS: Eric Schmidt chimed in, with a quick spin summary of the rationalizations:
Of course, all of these were not as self-delusional as Eric's big spiel that competitors such as Microsoft entering the market is good for Google - i.e. the online ad market is growing, so it's not a zero-sum game, competitors validate the market, and more companies move to online spending so keyword prices could just keep going higher, blah, blah, blah. That all may be somewhat true in the expanding market, but it's scary rationalization and it makes you realize that Schmidt has no experience in a declining advertising market...
So just keep telling yourselves all that, guys.
The rest of the call was interesting, but John Battelle already posted a summary list that pretty well covers it. It was interesting that the analysts generally ask the kind of questions that the Google-holic blogosphere, SEOs and SEMs are interested in. I do think the analysts are way too deferential to Google, but I think that's the price of access.
Update: Henry Blodget weighs in with a blog post title that's freakin' brilliant: Real Message of Google Analyst Call: We Care.
He surmises that some people think that Google and Yahoo should be more insulated than others because advertisers (should) know their ROI from their spend at Google and Yahoo.
I think Google and Yahoo are both vulnerable - Yahoo more so because of the brand advertising stable of customers that drives its profits. However, as I've posted on in the past, I think the decline of the mortgage lead market will hurt them both when things slow down.
In other words, I agree with Blodget and I don't think Google or Yahoo are very insulated from the usual advertising cycle. When it comes, it will hurt.