A History of Google's Panda Algorithm
"Why?" I hear the people in the cheap seats cry! Doesn't Panda make the job of an SEO copywriter or an SEO company that much harder, and cause you to wail and gnash your teeth every time Google announce a new Panda update?
Err...well, no actually. No it doesn't. Because for a specialist SEO guy / SEO copywriter like me - and I hasten to add: FOR SMALL BUSINESS AND SME OWNERS LIKE YOU - Panda is actually a Godsend.
Google's Panda Update Explained – The Layman’s Version
analysing and categorizing the value of on-page website content.''
Of course while the top-level view of how Panda works is deceptively simple, how Google algorithmically decides ‘good’ quality content from ‘bad’ quality content (which let’s face it is often subjective) is an infinitely more complicated matrix of mathematical madness, and is as closely guarded a secret as the Colonel's 12 secret herbs and spices, so nobody outside of the Googleplex, really knows. But before we try to get our head around the technobabble, let’s find out a little bit more about the tech who came up with it.
Navneet Panda – Education, Publications
- Efficient Top-k Hyperplane Query Processing for Multimedia Information Retrieval.
- Formulating Context-dependent Similarity.
- Concept Boundary Detection for Speeding up SVMs.
- Improving Accuracy of SVMs by Allowing Support Vector Control.
- KDX: An Indexer for Support Vector Machines.
- Exploiting Geometry for Support Vector Machine Indexing.
- Hypersphere Indexing.
- Active Learning in Very Large Databases.
- Speeding up Approximate SVM Classification for Data Streams.
- Formulating Distance Functions via the Kernel Trick.
Google's Panda Update Explained – The Technical Version
Up until this point, you have to remember that ‘bad’ pages were largely thought of by Google as ‘random gibberish’, which Google’s kick-ass Spam Team handled quite nicely, thank you very much. But the problem they now realized they were facing was one of qualitative subjectivity. Not relevancy, you understand, but SUBJECTIVITY. Google had been great at working out relevancy for years, thanks to on page factors such as Title Tags, H1s, H2s, key word density, et al. But how, they now wondered, could they algorithmically work out what written content on the web was actually ‘good', versus what what was inherently ‘bad’? A website such as ‘CheapViagraOnlineBuyNowCheckOutMyWoody!’ was easy for the Google Spam Team to classify and downgrade. But how could they work out which of those hundred websites legitimately selling Internet Marketing Widgets, actually knew what they were talking about? None of them were spam, not in the traditional sense, so which websites should end up on page one of Google, and which should be banished to Matt Cutts’ nine circles of Google Hell (page 2 and beyond)?
That’s where our good friend, Navneet Panda, comes in. Because Navneet was heading up a team of Google software engineers tasked with solving just this problem. And, in late 2009, after one too many late nights hopped up on, ironically enough, too much caffeine, Navneet came up with the breakthrough that would cause SEO companies and affiliate marketers to scream blue murder and reach for the nearest available bottle of hard liquor. A breakthrough that involved ‘Machine Learning’, and in particular a sub-section of machine learning, known as Support Vector Machines (SVM) or Support Vector Networks.
Machine Learning - despite what you’re probably thinking - is not that point in Google’s evolution when Skynet becomes self-aware and launches the nukes (no doubt at Microsoft, Apple and Facebook!), but is actually a networked series of supervised models with associated learning algorithms that interpret and analyze large banks of data in an effort to recognize and classify patterns within the data so as to run intensive regression analysis on it.
A standard SVM takes a predefined set of input data and predicts and forms the output (prediction) points, turning them into probabilistic binary linear classifiers for computational analysis.
Another technique Google is (rumoured) to utilise in the Panda algorithm is 'Latent Semantic Indexing' (LSI), which is an algorithmic approach that utilizes a mathematical technique known as 'Singular Value Decomposition' (SVD) to perform a taxonomy on predefined banks of data, to identify patterns and relationships inherent within unstructured text.
Confirmed Panda Updates
- Panda Update 1, Feb. 24, 2011 (11.8% of queries; announced; English in US only)
- Panda Update 2, April 11, 2011 (2% of queries; announced; rolled out in English internationally)
- Panda Update 3, May 10, 2011 (no change given; confirmed, not announced)
- Panda Update 4, June 16, 2011 (no change given; confirmed, not announced)
- Panda Update 5, July 23, 2011 (no change given; confirmed, not announced)
- Panda Update 6, Aug. 12, 2011 (6-9% of queries in many non-English languages; announced)
- Panda Update 7, Sept. 28, 2011 (no change given; confirmed, not announced)
- Panda Update 8, Oct. 19, 2011 (about 2% of queries; belatedly confirmed)
- Panda Update 9, Nov. 18, 2011: (less than 1% of queries; announced)
- Panda Update 10, Jan. 18, 2012 (no change given; confirmed, not announced)
- Panda Update 11, Feb. 27, 2012 (no change given; announced)
- Panda Update 12, March 23, 2012 (about 1.6% of queries impacted; announced)
- Panda Update 13, April 19, 2012 (no change given; belatedly revealed)
- Panda Update 14, April 27, 2012: (no change given; confirmed; first update within days of another)
- Panda Update 15, June 9, 2012: (1% of queries; belatedly announced)
- Panda Update 16, June 25, 2012: (about 1% of queries; announced)
- Panda Update 17, July 24, 2012:(about 1% of queries; announced)
- Panda Update 18, Aug. 20, 2012: (about 1% of queries; belatedly announced)
- Panda Update 19, Sept. 18, 2012: (less than 0.7% of queries; announced)
- Panda Update 20 , Sept. 27, 2012 (2.4% English queries, impacted, belatedly announced
- Panda Update 21, Nov. 5, 2012 (1.1% of English-language queries in US; 0.4% worldwide; confirmed, not announced)
- Panda Update 22, Nov. 21, 2012 (0.8% of English queries were affected; confirmed, not announced)
- Panda Update 23, Dec. 21, 2012 (1.3% of English queries were affected; confirmed, announced)
- Panda Update 24, Jan. 22, 2013 (1.2% of English queries were affected; confirmed, announced)
- Panda Update 25, March 15, 2013 (confirmed as coming; not confirmed as having happened)
- Panda Update 26, July 18, 2013 (confirmed)
- Panda 4.0 Released, 20 May, 2014 (confirmed)
- Panda Update 28, September 23, 2014 (confirmed)
If You Panda it Google Will Come
For this example I’ll keep the discussion to a fictional article page you’ve come across via a Google search for ‘Internet Marketing Widgets’ (but of course the rules presented hold true for content on any subject).
Top 20 Ways to Get Panda to Love Your Content
- Ask yourself – objectively – if you’d just stumbled across the article on Internet Marketing Widgets (via Google) and read it, would you personally trust the information presented in it? If the answer is yes, then chances are Panda will feel the same way.
- Does the article have a bare minimum of 300-500 words of copy on it (but ideally a whole lot more)? SEO / copywriting research done by the always erudite Neil Patel suggests the new Google content sweet spot may actually be north of 2,000 words of copy per page. So if this Internet Marketing Widgets page has well written copy galore, then Panda will think it has something to say on the subject. If it doesn't, then Panda will think the content is ‘thin’ and will downgrade it accordingly. After all, thinks Panda, if they know so much about Internet Marketing Widgets…why on earth is the article so short?
- Does this article read like it was written by an expert on Internet Marketing Widgets, or does it read like the person who wrote it wouldn’t know a widget from a wet willie?
- Is the article 100% original content, fresh to the web, and steaming with factual goodness, or is it just re-heated slops served up on a clean plate? Because if the content is original and actually adds something to the web’s discussion on Internet Marketing Widgets, Panda will love it, and rank it accordingly. But if the article is full of duplicate or thinly veiled ‘spun’ content, Panda will bring out its claws and tear it to pieces (and rightly so).
- If, after reading the article, you were asked to input your credit card to purchase a box of Internet Marketing Widgets from this company, would you feel comfortable doing so, given what you’ve just read? If the answer is no, then Panda will pounce because the trust value engendered by the content was obviously low. And if you – a person actively looking for Internet Marketing Widgets – don’t feel comfortable buying them from a company that sells them, then something is obviously wrong with the content.
- Does the article have spelling, grammatical or factual errors? Because if it does, it deserves to be penalized by Panda, because this sort of ‘Writing 101’ is the simplest of all the on-page content issues to get right; and not getting it right is just plain lazy.
- Is the article of genuine interest to people who want to find out more about Internet Marketing Widgets? Or is it just so many words on a page trying to trick the search engines and con you out of your hard earned money?
- How does the article compare to other articles about Internet Marketing Widgets you’ve read on-line? Does it offer any new insights into the subject? Any new data? Anything that hasn’t been said before? If the answer is yes, then all is good. If no, then chances are you wasted your time reading it, and Panda will feel the same way.
- Does the article discuss the topic fairly, giving the pros and the cons of using Internet Marketing Widgets? Or is it all just a one sided, ‘Our Widgets are Great – Buy Buy Buy!’ sales blurb that makes your eyes glaze over with the banality of it all?
- Is the article unique to this page, this website? Or is it viewable on other places across the web? Because if it’s not a one-shot deal, Panda will know it and act accordingly.
- Is this website a recognized authority on Internet Marketing Widgets? Or is this page just a token bit of non sequitur content tacked on to the website to try to trick the search engines into giving the site a bit more traffic, and make the website owner a few more dollars?
- Does this article contain any insightful analysis or interesting tidbits of information beyond the obvious? Did you learn anything new when you read it?
- Is the article on this page good enough to like, tweet or bookmark? If not, why not?
- Is the page the article is on crammed with ads that distract from or interfere with, reading it? Because trust me when I tell you that Panda hates this!
- Is this article well enough written that you could rightfully expect to see it in a reputable printed magazine, book or encyclopedia?
- Would you visit this page again or recommend it to your friends? Because if you wouldn’t chances are the page won’t get many visitors. And how many visitors a page gets is taken into account by Panda, as is average time spent reading the page and whether or not people interact with it, etc (think the benefits of having a low ‘Bounce Rate’ and a high ‘Average Time on Page’ ranking in Google Analytics).
- Does the article have internal links to other, Internet Marketing Widget related pages? Because if it does, this suggests to Panda that this site is an authority on the subject, while also going a long way to improving visitor engagement (and deeper Google indexation).
- Does the article have pictures and / or videos as well as words? Panda understands that a page with visual content such as pictures, videos, audio AND words, offers a far more rewarding and enjoyable user experience than words alone (though the words are by far the most important).
- Do readers have the ability to comment on the page? Comments are a great way to promote interactivity. And Panda knows this.
- Does the author of the article appear in SERP snippets? If they do, then it will improve the page’s click through rate as well as help users find their content. AuthorRank may not be a huge ranking signal yet, but trust me when I tell you that IT WILL BE, because people like reading content from people they acknowledge as experts in the field being searched for. In this case, Internet Marketing Widgets!
So there you have it. Everything you ever wanted to know about Google Panda, with the bonus of 20 sure fire ways to get Panda to love your content, thrown in for good measure. So now, whether your content is about widgets, or waveriders, wallets or wet willies, if you follow the above rules you can rest assured that Panda will notice what you've written and Google will turbo-charge your website content up the SERPs quicker than you can say, "Matt Cutts made me do it!"
Oh, and if by chance you have trouble remembering all 20 of these rules, just remind yourself of this, and you'll be fine:
Write for Humans. Optimise for Google.
Want to master SEO? Then book in yourself or your staff for some SEO Training Sydney today! Sydney's #1 SEO Trainer Brian M Logan shows you everything you need to get your company website on the first page of Google. SEO Training can be in-house (we come to you) or classroom based (you come to us).
Or you can always call us on 0425 204 887 if you want to leave your SEO Sydney requirements to the experts. Your website's 1st Page of Google results are guaranteed in writing in the contract.
Enjoy this free article? Great! Don't forget to Like it, Tweet it and LINK to it from your blog or website. Your sharing of the page with your friends is greatly appreciated! :)