"Data Refresh" Defined - Do You Agree?

My Interpretation of Matt Cutts Video Explanation

[egomaniac]

I think I "got it" today about what a "data refresh" is.

Matt Cutts released a video last week defining the term data refresh [video.google.com] (thank you Matt).

I interpreted it this way: the algo elements are the dials that we refer metaphorically too, the index is all of the spidered information about our websites, and a data refresh is someone changing the settings on the dials to adjust the relative importance of various algo elements.

If my interpretation is right, then a data refresh is similar to an update of years past in that the relative importance of algo elements changes. What's different today from the past is that no new algo elements are introduced during a data refresh - those are introduced whenever they are proven ready as Matt explained in the video.

Note that Matt says in the video that the algo element that was affected by the June 27 data refresh has been an algo element for 18 months.

What do you think?

P.S. I don't care if I am right or wrong, I just hope to get to an understanding through discussion...

[Nimzovich]

I don't see it like you; if so, big changes on SERPs would have been reported. What has been reported is a lot of sites dropping and most remaining the same.

Actually I don't believe they are only running an unique algo constituted by some elements, but different algos constituted by some elements. For example, an algo for ranking new domains, an algo for ranking new pages on old domains, an algo for filtering spam... But who knows.

[gcc_llc]

" I don't see it like you; if so, big changes on SERPs would have been reported. What has been reported is a lot of sites dropping and most remaining the same."

Umm...That bascially what everyone is reporting. The 3,4, and 5 page drop in SERPS came directly from June 27th.

[Genie]

I believe that Matt Cutts was saying that a data refresh is a re-calculation of one component of the algorithm. Let's call this particular component OOF (for over-optimisation filter). Nowadays PR is re-calculated constantly, but it seems that OOF is recalculated once a month, though Matt did say that it might change to fortnightly or weekly IIRC.

Matt discussed actual changes in the algorithm separately. He said those were like puting a new engine in a car. Whereas the data refresh was like putting in fuel.

[Genie]

By recalculation I don't mean changing the formula for that component. I mean applying the formula to the index or part of it. In other words recalculating how that formula is actually applied to sites/pages in the index.

[jimbeetle]

I think Matt explained it fairly clearly.

Index Update: Done on a continual basis, so should not really notice one unless an index has to be rebuilt for some reason.

Algo Update: Dialing the knobs up or down. Matt used the example of making PR a more or less important part of the algo.

Data Refresh: An update of the information that the algo works on. Can be continual; Matt used the example of newly recalculated PR being continuously fed into the system. Can also be on a once every blue moon basis; Matt said the last two burps were instances of data used by parts of the algo being refreshed.

[Genie]

You put it much more clearly than me. That's exactly how I understood it.