How I Accidentally Conducted My First Panda Experiment

Taking a look at an old search experiment with some interesting observations in the light of recent algorithm changes.

Note: I welcome all input, feedback and constructive criticism. However, if you’re of the opinion that SEO experiments are pointless, then please leave now before you get deeply upset by the ‘speculative’ nature of this post.

Test #1

In early February 2011, two domains were registered:


Both were linked from the same page in exactly the same order as above. For results and the technical set-up of the experiment you can read more here.

Search Query:

Why did hyphenated domain outrank the non-hyphenated domain?

The main differences between the two sites:

  1. Hyphenated domain was linked first on the source page
  2. Hyphenated domain was indexed first
  3. Hyphenation (claimed to be a negative factor)

Clearly, the number of hyphens in the domain did not prevent it to outrank the second domain in our experiment.

Google Results Pattern

Test #2

At around the same time I noticed a pattern of pages displayed in search results for the site: command and wanted to see what’s going on in a controlled environment. So we used the same domain and set up a  few pages on it. Each page was of a different URL length and they were all linked from the same source and in the correct order.

Search Query:


Date: 10th August 2011

As visible from the search query ( the results are sorted in the following order:

  1. INDEX
  2. 1234567
  3. 123
  4. 12
  5. 12345678
  6. 12345
  7. 123456
  8. 123456789
  9. 1234
  10. 1

Expected outcome against actual order:

  • 7:1234567
  • 3:123
  • 2:12
  • 8:12345678
  • 5:12345
  • 6:123456
  • 9:123456789
  • 4:1234
  • 1:1

Observations: There seems to be an underlying pattern with numbers funneling down and re-setting at certain points.

Pattern: 7,3,2,8,5,6,9,4,1

We followed up to this search query and found the pattern unchanged. Nobody has yet been able to figure out why this particular order. Perhaps an independent experiment may give an answer whether this same pattern would repeat or not.

So what’s Panda got to do with this?

Well, Panda 1.0 was released in late February 2011, just after my tests had finished. One domain has a single page in index, the other had a whole bunch of pages – all with very thin content.

I am starting to think that I may have actually tested Panda filter in action for the first time back in February last year. Without realising what I was doing.


Dan Petrovic, the managing director of DEJAN, is Australia’s best-known name in the field of search engine optimisation. Dan is a web author, innovator and a highly regarded search industry event speaker.

More Posts - Website