It always bugs me when otherwise smart people do this in a meaningless effort to prevent automated collection of their email address:

jatbar[at]gmail[dot]com

or this

f8dy(at)diveintomark.org

Anything of the form : BLAH at BLAH DOT BLAH is screaming out: SCRAPE ME! I'M AN EMAIL ADDRESS!

In fact, it might as well be written in day-glo blinking letters for all the spammers perl script's regexes care. See for yourself, try a search on Google...

If you want to let humans see your address, but you don't want scrapers to see it, you probably are best off making it an image, and somewhat CAPTCHA like. But there's no guarantee.

¶ 2/26/2007

Here's a very detailed discussion between Marissa Mayer and Gord Hotchkiss about Google's approach to personalization. Google is slowly but surely pushing personalized search results, mainly because they believe they can increase relevance.

Interesting factoid: currently, Google will only personalize 2 out of the top 10 results, and never the first result.

The actual implementation of personalized search is that as many as two pages of content, that are personalized to you, could be lifted onto the first page and I believe they never displace the first result, because that's a level of relevance that we feel comfortable with. So right now, at least eight of the results on your first page will be generic, vanilla Google results

Later in the interview, Mayer lays out details about how PageRank works for personalization:

Interestingly enough, the reason they were interested in building a faster version of PageRank was because what they wanted to do was be able to build a PageRank for each user. So, based on seed data on which pages were important to you, and what pages you seemed to visit often, re-computing PageRank values based on that. PageRank as an algorithm is very sensitive to the seed pages. And so, they had figured out a way to sort by host and as a result of sorting by host, be able to compute PageRank in a much more computationally efficient way to make it feasible to compute a PageRank per user, or as a vector of values that are different from the base PageRank.

Good stuff for the SEO-ologists. And for the SEM crowd, a tidbit at the end:

And as to ads, I think there are some easy ways to personalize ads that we've known for some time, but we've chosen at this point to focus on personalizing the search results because we wanted to make sure to delivered the end-user value on that, because that's our focus, before we look at personalizing ads

Won't that be fun!

Labels: google, hotchkiss, marissa mayer, personalization, search

¶ 2/23/2007

Pretty interesting to watch as Yahoo's new property MyBlogLog gets attacked by spammers, and one of the people who is simply pointing out the flaws gets banned.

Yahoo's reaction so far hasn't been warm and fuzzy. A distinctive sound of crickets chirping, emanating from the usual suspects. A bit of a mea culpa from the MyBlogLog (MBL) team, but mostly around the technical issues.

Shoemoney and Andy Beal first raised hackles by having a contest to see who could get the most MyBlogLog readers.

Recently Shoemoney posted on how you could spoof any MyBlogLog User, and "surf" with their avatar. This was just one of a long line of possible spammy exploits. The MBL team must have gotten fed up with Shoemoney, cause they banned him.

The upshot is that when Yahoo bought MBL, they massively increased its mind-share - in the minds of spammers - which follows one of my fundamental laws of the internet: Basically, anything successful will be spammed in proportion to its success.

The MBL architecture was not ready for the spamming, and they are now scrambling to revamp it. Meanwhile, many high-profile sites like TechCrunch seem to be pulling the widget.

I'm mildly surprised that Yahoo & specifically Jeremy Zawodny hasn't handled this better - he's been a strong MBL supporter (even exempting MBL from his anti-badge spiel), but he's gone rather quiet when Yahoo needs him most.

Labels: mybloglog, shoemoney, yahoo, zawodny

¶ 2/23/2007

To see the Avg. CPM column, click on a campaign in AdWords. Then you get the "Campaign Summary" screen, showing you the AdGroups under the campaign.

In the top part of the list of ad groups, there is a "customize columns" link. Click on that, and select "Average CPM". The list of ad groups will refresh, and add a column for Average CPM.

What does Average CPM mean? It's your average cost of 1000 impressions for your ad group or search keyword. It depends on the following factors:

• The click-thru rate (CTR)
• The cost per click (CPC)

Google computes the Average CPM by multiplying your CTR * CPC * 1000.

Example:

Ad Group	CTR	CPC	Avg. CPM
AdGroup 1	25%	$0.70	$175.00
AdGroup 2	12.5%	$0.70	$88.75
AdGroup 3	12.5%	$0.35	$43.75

In the table above, you see that with a very high CTR (i.e. 25%), you are grabbing a lot of clicks per 1000 impressions - so your Average CPM is high. If your CPC goes down, or your CTR goes down, your Average CPM will fall as well.

I think the best thing about the Avg. CPM is that it allows you to think like Google.

Google is selling ad space by CPM - even though most people think they are selling by CPC. They are trying to maximize CPM for the pages that they put ads on! That's how they measure monetization.

Google has quality score and ad rank to increase relevance and improve monetization, but those imply that a key monetization metric is page CPM. At the end of the day, they compute overall CPM for all the ads on a search engine result page (SERP).

Understanding Avg CPM, and understanding how Google sees the overall CPM of the ads on the page is instructive and helpful - especially when dealing with a competitive keyphrase.

Labels: adwords, cpm, google

¶ 2/20/2007

Sue Decker sent out a long email to all Yahoos which was promptly leaked and roundly lambasted by Mike Arrington, Paul Kedrosky and others. It's so full of corp-speak obfuscation, that Arrington didn't even try to summarize.

I'll give you one nugget: Steve Mitgang is gone. Who's he? He was ostensibly in charge of the Panama re-vamp. He was the Jim Allchin of Panama.

Panama's development was such a disaster - it was 2 years too slow - that it was very predictable that Yahoo would do some house-cleaning, and regret-shunning (i.e. find someone to take the fall). In fact, I wrote a post saying that Yahoo should clean house at Overture:

Fix search monetization. The Panama rollout is underway. Once that happens, clean out Overture of the people who failed to deliver it 2 years ago. Then make an all out run at Google - develop a way for advertisers to dump entire Google campaigns right into Yahoo. Respond to advertisers better than Google.

So I think this email mainly shows that in their own, convoluted, shuffling-of-the-deck-chairs way, Yahoo is attempting to clean house.

PS: I love that Sue Decker's last name always reminds me of Blade Runner. And "Mitgang" sounds pretty Phillip Dick-ian as well, to me: Decker gets rid of Mitgang - just sounds gritty...

Labels: decker, panama, yahoo

¶ 2/15/2007

Gregg Easterbrook has some stats that guide you on how to predict football scores when doing office pools where you need to pick the final score of some game.

Simple rule-of-thumb for picking the final scores of NFL Football games:

• Do NOT pick 11, 12, 15 and 18 -- these scores rarely occur.
• Good numbers to choose include: 20, 24, 17, 27, 10, 13
• The most common final scores in 2006 were, in order:
  • 20-17, 23-20 and 24-21 (occured 6 times each)
  • 24-10, 27-24 and 17-13 (each occured 4 times)
• Avoid low totals, and high totals. I.e. don't pick 6-3. Don't pick 44-38.

Using those rough guidelines, you have about a 1 in 500 chance of picking the exact final score, according to polymath Easterbrook.

For quick fun, and to double-check, I just did a frequency count on the scores of NFL games from 1990 - 2000 in order to rank the most common final scores.

Here are the most common scores - choose from these:

• 17 occurred 371 times - 9 percent
• 10 occurred 303 times - 8 percent
• 20 occurred 302 times - 8 percent
• 24 occurred 291 times - 7 percent
• 14 occurred 264 times - 7 percent
• 13 occurred 254 times - 6 percent
• 21 occurred 214 times - 5 percent
• 27 occurred 214 times - 5 percent
• 31 occurred 193 times - 5 percent

The least frequent scores are all high numbers above 40. But 11, 12, 18, 25, 32, 33, 36 occur very few times for lower numbers. Never pick a number above 38!

PS. I know this topic isn't seasonal nor very ad related, but when football season rolls around again, I'll know where to find the info - and it'll be indexed by that time.

Labels: betting, football, randomness

¶ 2/14/2007

Microsoft adCenter is having some problems keeping their reports up to date. If you look at the adCenter dashboard and see that your campaigns aren't getting any impressions from yesterday, it's probably because the report system is not up-to-date.

Look very carefully at the bottom left of the dashboard. Right now, as of 9:30am PST, Feb 13, it says (in a tiny gray font):

Performance data last updated on 2/12/2007 7:00:00 AM Pacific Time (GMT - 8)

That's why I don't have many impressions for my campaigns from yesterday! The data is over a day old! I'm glad it's that, and not some editorial action, or billing problem.

I called Microsoft adCenter support and they said "they are working on it". While working, they should make the "last updated" message a lot more prominent when there's a big delay.

Also, why no message on the adCenter blog? They usually put up boring notices of maintenance downtime on that blog, and the reporting has been broken for about 2 days now.

PS: Why does the adCenter blog need a cheesy mediaplex banner ad on top? Is that helping your credibility?

¶ 2/13/2007

I'll pay $50 (via Amazon Gift Certificate) to any person who can parse the following. All you have to do is tell me, in clear, concise terms:What the hell is Microsoft thinking with this Windows Live branding?

Here's Gord Hotchkiss interviewing Microsoft Live Search's Justin Osmer - and I'm italicizing all the cliches that come from muddled thinking:

Q: Why the move in branding from MSN Search to Live Search? What was the thinking behind that?

The thinking behind Windows Live in general is that it's a suite of services that are available across the Microsoft platform of products. So MSN has the ability to give you your email, get to search and so on, but those things are Windows Live branded experiences and services. MSN still has search today, but it's powered by Windows Live Search. Same as if you go to Microsoft.com today, it's powered by Windows Live Search. So, it's somewhat intentional in setting up in Windows Live a collection of things to allow people to get all the information they care about in one spot. With one log in you get your contact list, a melding of your IM contacts and your mail contacts, you've got your Windows Live personalized page that you can set up to get your own RSS feeds, you've got your search results there, you can build in your search macros, you've got all sorts of things that you can do to make it even that much more uniquely yours.

MSN continues to be a very strong brand and a place where people love to go get content, read content, review content and get immersed in that content. Windows Live is more about you cherry picking what's important to you, pulling that down and having that in one spot. And then the services that underlie Windows Live will then power the experiences on these other properties.

Was that answer supposed to make things clearer? Strunk & White would have a field day with that answer.

I'm not sure why normal people would need to log in, build search macros, or want to meld their contacts? Do real users think : "I'm going to MSN to get some content"?

Moreover, if MSN "continues to be a very strong brand", then why dilute it with Windows Live? From that answer, the best I can infer is that Windows Live is about cherry picking and personalization?

As Tom Hanks said in BIG! - I don't get it... and neither, apparently, does anyone at Microsoft!?!

Labels: branding, microsoft

¶ 2/12/2007

While others have been interpreting the tea-leaves about Google's recent "you can't handle the truth" statement about video, Robert Cringely, ubër-pundit, tells you what it really means:

And the effort began this week with Google stating what I've been hollering for weeks, that the Internet is going to be buried in video. "The Web infrastructure, and even Google's (infrastructure) doesn't scale. It's not going to offer the quality of service that consumers expect," Vincent Dureau, Google's head of TV technology, said at the Cable Europe Congress.

Cringely thinks Google has a grand plan on video, and attendant advertising. I'm usually skeptical of anything conspiratorial that would require a big company (like Google) to depend on future innovation and make grand plans farther than 1 year out; however, I must admit, I'd put my money on Cringely at this point:

The future lies in watching precisely the video you want to watch precisely when YOU [...] want to watch it. This requires a client-server distribution system for more than just ads and it will come as bandwidth and server capacity increase. I've spent the last three weeks explaining exactly how that will happen. But first Google has to make itself indispensable to European cable companies, to get a toehold in their market. Only when they are totally dependent on Google ad services will the search giant reveal its true video ambitions in Europe.... and the world.

Bandwidth, P2P and video distribution predictions are things Cringley does best.

¶ 2/12/2007

Aaron Wall of Seobook.com fame is giving away free consulting - normally worth over $500 / hr! He'll give you an idea that will make good money within 3 months.

The best part, it's a really simple plan - anyone can do it. I can't believe he's doing this for free!

Here's a sample from an IM chat with him:

AW: you market THAT angle. no i am saying right now, put another way, i could work 1 hour a day developing the blog idea i gave you and i would bet inside of three months it would make far more than your idea, so listen to the idea and run with it...

The only problem is that you have to read a long blog entry to qualify for this free advice. However, I guarantee it's worth your time.

¶ 2/09/2007

Bad news / good news: Yahoo apparently decided to use the Yahoo account database for telemarketing. That's bad.

A friendly telemarketing agent working for Yahoo Publisher Network from Fargo, North Dakota called my home number to try to get me to put YPN ads on one of my websites. She claimed she got the info from the "Yahoo Database", but she didn't have my email!?

I asked her what gave Yahoo! permission to call my home number, and she said it was a mistake for them to do that.

What's the good news? At least Yahoo is making an effort to market YPN! Of course, I've already been signed up for YPN for a while, but apparently they didn't cross-check that database.

Labels: telemarketing, yahoo, ypn

¶ 2/07/2007

My contest to rename Panama came up with some good entries. I choose the best name from about 40 submissions:

Jon Roig submitted that name, and it did pretty well in the user poll. I think it's a good brand name - it's descriptive, and not too narrow. It's sounds cleaner and broader than AdCenter (from Microsoft), and it will easily be identified as distinct, yet competitive to the leading brand, Google's AdWords.

Next step: I'm going to email Terry Semel, and my contacts within Yahoo with a persuasive argument that they really need to create a brand for Panama, and offer Yahoo! AdMarket as a suggestion. No I haven't checked the trademark database, but the point is to get Yahoo! to do something positive.

PS. Jon Ruig, please leave a comment on this post, pointing to a place where I can find your email to send you your winnings.

Labels: contest, panama, yahoo

¶ 2/06/2007

For many years, the Overture Keyword research tool was the main source of data people had about search terms. It ostensibly showed what key phrases had significant volume on Yahoo! Search.

Now Yahoo has gone, and canceled the free access to the Overture tool, which many SEO keyword tools (like Aaron's) had relied on.

Here's a good substitute: Wordze. I've used wordze for 3 months. It's not free, but it is very useful for doing keyword research.

My favorite features are:

• The "dig" keyword tool - you type a keyword, and it builds a list of thousands of related keywords. Typically, you will get between 1000 - 8000 KWs for a single word term like "humidifier".
• The WordRank tool, which analyzes the competition level for a keyword.
• The Wordze API - which can be queried to return XML keyword lists automatically for programmatic research. This is pretty valuable in itself if you have the right application (i.e. automated SEO)

For example, a dig query for "projector" took about 12mins to complete. It returned 4848 words, all very relevant phrases, mostly 2-5 words in length. It provided a frequency count, estimated volume, and a KEI for the top 200 or so. Here's the top 10:

Keywords	Count	Est.	KEI
projector	392	9179	0.18
projector screen	240	5620	3.45
lcd projector	234	5479	1.43
overhead projector	189	4425	1.09
video projector	179	4191	1.84
projector screens	166	3887	2.97
projector lamp	132	3091	3.92
3m overhead projector	119	2786	55.37
projector headlights	115	2693	9.74
dlp projector	92	2154	0.33

A wordrank query for "humidifier" returned a lists of domains, counts of links, types of links (Edu links & Gov links), domain and URL pageranks, along with search demographics by country, and city

Wordze also has filters and options for finding keywords in profitable niches such as Adult, Drugs, Gambling, Warez & Hacks. This means you can get lists of those types of keywords if you are into those areas.

Most of the Wordze tools work on a batch basis - you put a query in, and it works on it in the background, so you can quickly enter 10-20 queries, and wait about 20mins for them all to be completed. Then you go download the results from the download manager.

Support on the tool is good. I've gotten several replies in under a few hours from Levi - the author, whenever I had a question.

The downsides to Wordze? Sometimes you have to use multiple steps to really expand a keyword list. For example, it might return basically the same set of keywords for related terms like "refigerator" and "frigidaire". Then you may need to use the refining tools to get a more expanded list for the terms. Doing this is a pretty good way to find long-tail keywords that competitors may not be using, however.

Wordze doesn't really provide any bid / cost information about keywords, but it does indicate competitiveness. I never found the Overture bid estimates to be accurate at all. The only way to tell how much stuff will cost is to go bid on it.

I use Wordze mainly for SEM - quickly building lists of keywords to buy. I think most people that use it probably do SEO research with it.

Wordze costs about $35/month and payment works with PayPal. To me, it's a no-brainer, because if you do any serious keyword research, it can easily pay for itself in 1 day.

¶ 2/02/2007

Yahoo's going to build their own automated splog system. (Splogs are sites that are basically spam blogs - containing little original content - it's mostly scraped.)

Except Yahoo calls it "Brand Universe". GigaOM has some opinions on it.

Their first splog is wii.yahoo.com. Overall, it's embarassingly lame. Yahoo is announced they are going to build 100 more. Why only 100? That ain't a universe?

I wonder why they don't already have 100 sites up? How hard is this? The content is easily grabbed from the Yahoo APIs - including the Flickr pix, News and Blog results, Yahoo Answers topics, etc.

In fact, when I get a chance later today and tomorrow, I'm gonna build 100 brand universe clones, using the Yahoo APIs. Then I'll make sure that the news items / scraped content I put up gets linked from the real yahoo versions.

And I think I can get my ads more relevant and targeted as well... Should be fun.

Labels: splog, yahoo

¶ 2/01/2007