I still don’t get the point of bing’s latest Thelma and Louise commercial. If you haven’t seen it yet, check it out below:
What exactly are they trying to say? It just came on with the TV on MUTE and here’s the story it told me:
Women are being chased by cops, they look at google, don’t like what it says so they try Bing. Bing gives them directions, and they follow them right off the cliff.
Ok, that can’t be right. Let’s try another interpretation.
Ladies try switching from Google to Bing, then become so frustrated with the quality of Bing’s results that they drive off of a cliff.
Hmm… that doesn’t seem like a good marketing message either.
I really don’t get this commercial. To me the marketing message is just off. It tells me that a.) Bing caused them to drive off a cliff or b.) they’d rather drive off a cliff than use Bing. I don’t think either of those are the message they were going for.
Pro Marketing Tip: It’s usually never* good to show customers killing themselves or dying when using your product.
*unless of course your product is a funeral casket or assisted suicide machine.
SEO Isn’t science. It’s not rocket surgery either, but it’s definitely not science.
I’m not sure if having a computer science degree makes me a scientist or not, but I do know how to spot science. See, science has a definition. According to Wikipedia, that definition is: a systematic enterprise of gathering knowledge about the world and organizing and condensing that knowledge into testable laws and theories.
It’s that last bit that causes all sorts of problems for traditional SEO. We SEOs are pretty good at gathering and organizing knowledge. We’re even better when it comes to sharing it, but when it comes to testable laws and theories we kind of suck.
Ever since Branko’s presentation at SMX (which I liveblogged by the way ) I’ve been thinking about testing and the scientific method – and how little of it we actually do in the SEO community. It prompted me to start testing META descriptions with Alan Bleiweiss. Even though we both “knew” what the result was going to be, we did a scientific test anyway – and that’s what the SEO community needs more of.
The problem though, is that in addition to being technologists, SEOs also have to be marketers – and marketing and science don’t really have the best relationship. Marketing tends to be about making claims whereas science is more about collecting and publishing data. As SEO/Marketers we not only have to do the research, but we have to then go in front of a client and convince them why they should spend money to make changes. Often, data alone isn’t enough to sway a client.
A successful SEO is one who is able to wear a science hat and a marketing hat, but a great SEO is one who knows when to wear each one of those hats.
It’s not that SEO can’t be science, it’s just that we’re not doing a good enough job of making it a science – and that’s bad for the industry. Unverifiable claims are one of the quickest ways to get the snake oil reputation. Science is how that nasty reputation can be avoided.
Unfortunately, many SEOs (myself included) get too caught up in the lifestyle to worry about actually doing SEO. Talking about SEO, doing the conference circuit, and living the A-list twitter life are all things that come to those who make SEO claims – but sometimes we get so caught up in that life and making those claims that we forget to actually test our claims.
Sometimes, it’s our own egos that prevent what we do from being called science. “I’ve been doing this for X years, it works, I don’t need to test that, everybody knows it.” How many times have you said something like that? Read through the comments on Alan’s META description test and count the passionate opinions there that are soley based on ego without any data to support them. There’s quite a few.
The true SEO scientist doesn’t just make a claim. He gathers data, then posts that data for others to examine. The problem though, is that posts full of data don’t get retweeted and they don’t get onto the front page of Sphinn. Posts full of claims however do get lots of retweets.
It’s that community aspect of SEO that’s holding us back from achieving our true potential. I can hear a few people muttering under their breath “Oh, he’s just jealous that he’s not an A-lister” and well, that’s true, but it’s not the motivation for this post. I firmly believe that all the A-list SEO people have earned their status. They did awesome work, wrote great blog posts, and did everything else to earn the success they’re enjoying today.
Science and fandom don’t mix.
The problem with reputation though, is that people stop questioning A-listers. They’re no longer required to produce data or backup their claims. While they’ve deserved that right, it’s not good for science. A-listers and SEO rockstars can be wrong too. Nobody’s perfect. Just look at how many advocated pagerank sculpting several months after Google quietly changed how nofollow worked. (then, look at how many refused to admit they were wrong.)
Pay attention next time Danny Sullivan or Lisa Barone write a blog post. Almost instantly they’ll get 10-12 retweets. (and usually, they deserve them too as they write awesome stuff.) The problem here isn’t the retweets, it’s the amount of people who retweet before they actually had time to read the blog post. If retweets are the social equivalent to links on web pages, then so many people are turning their twitter feeds into free-for-all directories. You wouldn’t recommend a doctor that you’ve never visited, why would you recommend a blog post that you haven’t read? What if Danny’s blog had been hacked to include a Viagra post? I bet he’d still get several blind retweets. Stuff like that doesn’t help SEO get taken seriously as a science.
We’ve all heard that the squeaky wheel gets the grease, but in SEO it’s the loudest shouter who gets the attention. Usually, those shouting loud know their shit, but not always. It’s important for SEO that we still continue to scrutinize and think critically about what people are saying – no matter who’s saying it. Everybody makes mistakes.
When we’re dealing with over 200 inputs like in SEO, it’s very easy to mix signals, miss relationships, or confuse correlation and causation. Rather than simply making claims, we as an industry need to focus on becoming more scientific. Don’t be afraid to run tests, share your data, and discuss methodologies. You may be right, you may be wrong, but either way you’ll be starting a scientific discussion in which everyone is bound to learn something – and that will make us all better SEOs.
SEO in its current form is not science. It’s starting to be though. Several great people and companies are attempting to put the science back in SEO, and I’m excited to see where our industry can go once more people get on board.
Update: Google’s Maile Ohye has Confirmed my hypothesis that META descriptions do NOT affect relevancy. But please, still read this post to learn about testing SEO.
Last week after SMX, Alan Bleiweiss started running a META Description Test to see whether or not the META description has any affect on search engine rankings in Google.
The article generated lots of comments, and several SEOs (including Jill Whalen) argued that the META description does affect rankings. Somebody even claimed that Google employee Maile Ohye even said so at SES Toronto, but I was unable to verify this claim. For years, my theory has been otherwise.
From my experience, the best way to come up with a hypothesis for an SEO test is to do just that – think of things from Google’s point of view. This is where knowing how to program can greatly benefit an SEO. Ask yourself “if I were coding a search engine, would this make sense? Would it make my results more useful? How could it be spammed? How would I do it? Is it robust? Does it Scale?” You’ll find that answering those questions from an engine’s point of view will almost never steer you wrong with regard to SEO theory.
I didn’t believe that META descriptions affected rank so I decided to join in with a test of my own. We had just finished listening to a panel talk about testing SEO so testing was still fresh in my mind. I also wanted to keep the test as scientific as possible, so here’s what I did:
I created a new page and set the META description to a Unique Phrase that currently had no results for an exact match.
I then changed the META description of an already established site to also include a different unique phrase. The reason I did this is because Jill argued that new pages and old pages may be treated differently – so I didn’t want any bias in my data.
I then planted several links to both of the sites on various blogs, including the test post you see below this one (which will be removed in a few days.)
On the existing page, Google re-visited the page rather quickly and updated its cache. On the new page, it took just about 24 hours. That was several days ago and the spiders have been back several times since then.
After several days of waiting, neither the new page nor the old page rank for the unique phrases contained solely within their META descriptions. You can verify that by clicking the two “unique phrase” links above.
For the sake of completeness, they don’t rank in Bing either.
No matter what I tried, I was not able to make a page rank solely based upon its meta description.
Do some sites appear to rank for the text in their descriptions? Most likely, but we can’t confuse that correlation with causation.
If your site is properly optimized, the terms and phrases contained within your META description will also be the same terms and phrases used both on your site and in links to your site. Several tests have shown that when people see their search terms in titles and descriptions they tend to click through at higher frequencies (just ask your paid search people about this one!) Therefore, it’s a very common strategy to use the same wording in META descriptions as on the site – which then causes people to use that same wording in links.
Also, when it comes to sites who haven’t changed their META descriptions in a while, a funny thing happens. people scrape that content and use it on their spammy sites. Those people even sometimes link that scraped description back to the original site. In that case, it’s very possible that a site would appear to rank for terms contained within the meta description, but it may not actually be because of the meta description.
I’m glad that my test shows what I thought it would show – not simply because I enjoy being right, but because META descriptions are very easy to spam.
In most cases, Google doesn’t actually use the META description you provided. They prefer instead to use a snippet of text taken off of the page. It all goes back to the same “increase clickthrough when your terms appear” idea mentioned above. This means that users very rarely see the meta description you provide. Google has long said that they don’t like to rank sites based on factors that don’t appear to users (see META keywords) as they often invite spam.
So there’s the results – please leave your insights, comments, or observations in the comments. Also, please run your own tests if you feel fit. I’d love to see your results and compare your findings with my own.
If you missed SMX this year, you missed a great conference. I’m really thankful that I had the opportunity to attend, and I’m looking forward to coming back in future years.
I did a lot of live-blogging during the sessions, and lots of tweeting – but I wanted to collect some of the important points here on the page. This is stuff I learned not only from sessions, but from talking to other industry veterans at the bar as well. There’s a few unique insights here that many may not be aware of.
Here’s my summary of important takeaways from SMX:
1.) Page speed is a minor factor in rankings. Unless you have noticeable problems, it’s not worth worrying about for SEO. Worry about your user experience and your SEO will be fine.
2.) If you’re worrying about bounce rates affecting SEO, stop! Worry about lowering that bounce rate because it’s costing you customers. Matt also said that bounce rate is NOT used directly in search rankings.
3.) If you change your screen resolution, Google shows less images in results to accommodate for the smaller screen size.
4.) When it comes to vertical search and trying to rank for videos, ask yourself this: is a video really the point of entry I want my customers to choose? Will it have an effect on my conversions? Ideally, you don’t want a video to show up and push down your web page that already ranks well – unless you’re a video specific niche.
5.) If you can’t get a straight answer out of Matt Cutts, judge by his reactions or whether or not he’s looking up and to the left.
6.) Pagerank DOES decay over 301 redirects. Especially if you have multiple 301 redirects chained together. I asked Matt about this in the bar with an example of a chain of 5 redirects and he said “google says that’s too many redirects.” He also hinted that most of the pagerank from links to the original URLs would be gone.
7.) If you have a 301 redirect or rel=canonical on a page, it will slow the crawl rate at which googlebot fetches your pages. This came directly from Maile. If your site is down for maintenance, use a 500 response to avoid reduced crawl frequency.
8.) When measuring page speed, Google uses the toolbar time and not the googlebot time – so javascript and things like that actually matter. The estimate given in webmaster tools is very close to what the algorithms use.
9.) Marketers talk about claims. Scientists talk about data. Are you an SEO marketer or an SEO scientist? Too many SEOs are focused on making claims rather than letting the data speak for itself. Tip: Test your shit, publish your data, open it up to peer review.
10.) According to Maile Ohye, the goal of Mayday was to make long tail results more useful to users. Remember, searches don’t think of their query as “long tail.” The goal was to remove pages that were technically not spam, but not very useful. We called these pages “weeds.” Many people speculated that this was a response to Mahalo, eHow, and Demand media, but we got an “up and to the left” look out of Matt on this one…
11.) When it comes to paid links and link farms, Google’s first attempt is to take a “scalpel” approach where they just cut those links out of the graph. There’s no use worrying about a competitor damaging your site with links.
12.) Pretty soon Google rich snippets will happen automatically – just make sure you’ve got microformats on your site.
13.) Google has the ability to determine sentiment, but it is NOT used in rankings. Google analytics data is also NOT used in rankings.
14.) Real time search rankings are calculated based on: Author Quality, Site Authority, Trust & Relevancy. In other words, it’s not about how many followers you have, it’s about how many quality followers you have.
15.) To get into bing news, email bns@microsoft.com with a request and an rss feed.
16.) Google supports video sitemaps, Bing doesn’t. Both Bing and Google support local listings.
17.) You can disable Bing’s document preview with a meta tag: meta name=msnbot content=nopreview
18.) Bing & Yahoo rankings will be exactly the same, the display and user experience will differ. Sadly, no decision was made on the future of Yahoo Site Explorer.
So there you go. I learned a lot at SMX, not just about search engines and SEO but about the SEO community as a whole. All in all, it was a very interesting time and I hope you’ve learned something from my summary. I’ve learned a few things, and I’ve even come back with a few things that I need to test a bit more. I’ll close this article by mentioning two of my favorite quotes overheard at the event:
“The SEO community is like high school” – Matt Cutts talking to me at the bar.
“Insecure High School loner seeking vampire who sparkles.” – John Lebaron in his Craigslist ad.
The SEO Vets take all comers with Danny, Rae, Bruce Clay, Alex Bennert, Vanessa Fox, Greg Boser, Todd Friesen, and Stephan Spencer is starting soon. I’ve got my seat in the front row and I’ll be liveblogging soon.
screw it ustream: http://ustre.am/iWi6
Phone died during ustream. I’m going to pick up here liveblogging.
Rae has lost her voice. One of the panelists said it’s an early xmas gift for the rest of us.
Bruce and Vanessa are talking about news headlines. Don’t use “giant wave” in the headline if you’re talking about a Tsunami. Bruce says they can’t correct after the fact on CNN – so they have to do it during creation.
Vanessa is talking about crawling again. Whoever asked this question didn’t go to the last architecture session. See my notes below for the answer here, I’m not retyping it.
add &start=990 to google to see the end of the results. &filter=0 will show you omitted results
Don’t worry about bounce rate affecting SEO – worry about bounce rate affecting conversions.
Since Cutts won’t give a straight answer, a good strategy is to ask him questions and judge his physical reaction.
Matt’s taking notes about changing the “omitted results” list’s name to “a list of crap” I like that Idea. “would you like to see a list of crap from this website?”
Many SEOs here don’t believe official Google answers about crawlability. that’s shocking – why would Google lie about how they crawl the web? There’s no motive. Google wants everything crawlable and findable – it gives them a better search engine.
Should you ever stop link building? The answer: if you owned a brick and mortar store would you ever stop trying to get customers?
Nobody admits to buying links, yet lots of people do it. If you get penalized, you have to clean it up and file a re-inclusion request.
Not sure why so much paranoia over linked networks of sites or paid links. You have to go very obsessive to get banned. IN most cases, I don’t think many SEOs have to worry about links in bad places.
Google recently penalized Google Japan for buying links. If Google is willing to penalize themselves, they may penalize you. Google’s really good at finding links.
So what happens if your competitor buys links for your site…. Vanessa says it’s unlikely that you could get penalized in this case as bans usually result for multiple factors.
Don’t buy links, just give out free phones in return for links.
Is facebook and the open graph going to kill Google? No – but should you diversify into social media? Yes.
Rae says she’ll take any links, she doesn’t care where they come from. I’m in the same boat. One of the biggest traffic drivers to NoSlang.com is a nofollowed link on somebody’s site. Sometimes we lose track of why we needed links in the first place.
Don’t go crazy with universal search. Do you really want a video to rank higher than the link to your site where people can buy something? Most videos don’t have a link to go complete a conversion after watching it…those links that ranked do.
Alex Bennert says to email her if WSJ mentions your site but doesn’t link to it. Good to know, they’ve done that to noslang.com in the past.
The live blog of “Build it Better: site architecture for the Advanced SEO” will start momentarily. Vanessa Fox, Adam Audette, Maile Ohye, Lori Ulloa and Brian Ussery will be speaking. Stay tuned here and refresh for my thoughts, insights, & recap.
Vanessa and Maile must have a lot of clout here, as there’s a facebook session going on across the hall and the site architecture room is standing room only rightnow.
Lots of the room has complicated problems – not sure how many are related to search though.
Maile is up first. She’s using Google store. It has 158 products but 380,000 URLs indexed. How does that happen?
First point: protocol and domain case insnsitivity. http://www.example.com and HTTP://WWW.EXAMPLE.COM can be different.
Advocating a consistent url structure to reduce duplication and facilitate accurate indexing. suggestion: keep everything lowercase.
301s and rel=canonical can cause slower crawling. Google crawls 300 and 400 status less than 200s. If your site is down for maintenance use a 500 response code to not reduce crawl times.
Don’t be like the Google store, use standard encodings and &key=value stuff. no crazy stuff in place of key value pairs.
Google crawl prioritization.
Indexing priorities: URLs with updated content, new urls with probability of unique/important content.
Sitemap information (xml here) is used.
Ability to load the site (uptime, load, etc) also comes into play.
To increase Googlebot visits:
strengthen indexing signals above. (links, uniqueness, freshness)
Use the proper response codes.
Keep pages closer to the homepage. Further clicks away = less frequent indexing.
Use standard encodings
Prevent the crawling of unnecessary content.
Improve “long tail content” Be wary that we as webmasters call it long tail, but to users it’s the content they want.
Seek out and destroy duplicate content and use the canonical and 301. Google can find these for you in webmaster tools.
Include Microformats to enhance results with rich snippets. Gives the ability to include reviews, recipes, people, events, etc. Her example is the hrecipes format.
Create video sitemaps/mRSS feed. (only Google supports these currently. BingHOO says they’re working on)
Adam Audette is up now and he’s shilling Vanessa’s book – marketing in the age of google.
First, make the best user experience, then leverage that for SEO. That’s good advice.
He’s using amazon as his example. Talking about the top and left nav and how they make search prominent.
Sweet, they’re giving out jane and robot stickers. Gotta get me one of those.
Google shows less images in search results depending upon the size of your screen resolution.
adding dimensions to your images in addition to alt attributes can help browsers and search engines, and increase speed.
Use .png files if possible. Muchsmaller than.gif or .jpg
use EXIF, Tags, Geo, and what not.
Looks like we’ve killed twitter too.
Things you can test:
1. pages indexed. everybody knows how to do a site: query
2. Canonical (www and non www)
3. in-links using yahoo site explorer. (there was a plea to binghoo to keep this tool live)
4. sitemaps – blah blah blah
5. site speed – I’m so tired of hearing about this. it’s not a big factor at all.
Maile says that Google still doesn’t want search results in their search results. Disallow your search pages. Category pages however, are still very welcome.
Maile also said that Google uses toolbar page load times not googlebot crawl times. Good to know, but still don’t like people obsessing so much over speed.
text -indent -999px is not a safe technique to use instead of alt text.
pubsubhubub is an open protocol that lets you push content to search engines rather than let them try to index you. It’s not yet incorporated into google’s pipeline.
I’ll be live blogging the “so you want to test seo” panel at 10:30 pacific time. Check back here for live updates. This should be a good session.
Actionable, testing, etc. These aren’t words you normally hear when people talk about SEO. So glad to hear them.
Several people in the room admit to testing SEO. One guy admits that he’s perfect and doesn’t need to test his SEO.
Conrad Saam – director Avvo is speaking now.
He’s talking about statistical sampling and the term “statistically relevant” – I feel this is something that many SEOs fail at.
The average person in this room has 1 breast and 1 testicle. A good example of how averages can be misleading.
He’s now talking about sampling, sample size, variability, and confidence intervals. Also the difference between continuous and binary tests. This is very similar to my college statistics class.
His example of “bad analysis” looks awfully similar to some of the stuff I’ve seen on many SEO blogs. It would have been real easy for him to use a real example from somebody’s blog.
Bad analysis: showing average rank change in google based on control.
Good analysis: Do a type 2 T Test. Excel can do that.
It’s all about the sample size when doing continuous testing
Common mistakes:
Seasonality.
Non representative sample
non bell curve distribution
not isolating variables. This one is huge in SEO as there are over 200 variables considered.
Eww… he’s talking about the google sandbox. Just lost some cred with me, as I don’t believe in a Google sandbox – but it does make his point when testing SEO – that we can’t be 100% sure some changes actually caused the results.
Next up, John Andrews
An seo wants:
to rank better
robustness
avoid penalties and protect competition.
As an agency one wants to actionable data to help make the case why we want that.
Claims: (that need to be tested)
PR sculpting does/doesn’t work.
Title tags should be 165 characters
only the first link on a page counts.
there is no -30 penalty
John says that authors of studies and blogs place more value on the claims and not so much value on the claims. the difference between marketing is that marketers tell stories and make claims – scientists deal with data.
Problems with SEO studies
Remarkable claims get the most attention.
Studies are funded by sponsors who have something to gain.
There’s virtually no peer review.
Success is based on attention not validity.
“citations” are just links – and not as valid as real citations.
Note to self, but a copy of “the manga guide to statistics”
So how can we contribute?
Science is slow boring and not easy.
Most experiments don’t produce significant results
scientists learn by making mistakes
As SEO’s we’re stat checkers. We’re too busying seeing how much we just made and how many visitors we just got to deal with experiments. That’s so true.
Tips: Publish your data without making claims. Be complete and transparent. Say “this is what I did and this is what I saw” and people will email you, cite you, or repeat your experiment. Invite discussion about your test.
A good example of this was Rand’s .org vs .com test where he didn’t account for wikipedia bias and also didn’t alot that most .com domains were brand names (which he excluded)
When it comes to SEO testing, just say what you saw. Let the data tell the story and let others come up with the same analysis that you did. That’s science. Publishing claims is often just a push for attention. Man, that’s so true.
Next up, Jordan LeBaron.
Don’t trust Matt Cutts, test your own shit. Different things work in different situations.
Branko Rihtman – a molecular biologist who runs seo-scientist.com
Define question, gather info, form hypothesis, experiment, analyze and interpret data, publish results, retest. That’s the scientific method.
Choose your testing grounds. don’t use real or made up keywords, use nonsensical keywords made up of real words (like translational remedy or bacon polenta)
How to interpret data:
Does the conclusion agree with expectations? Does it have an alternative explanation? Does it agree with other existing data? Bounce the findings off of somebody. Don’t have definite conclusions.
Statistical analysis is hard. get help from somebody who knows statistics. Understand correlation and caustation, understand significance. Don’t rely on average.
Avoid personal bias. Don’t report what you want to see or what you thought you saw, report what you actually saw.
You can learn a lot from buying branko a becks. It’s a known fact that scientists can’t hold their alcohol.
