These days, Google algorithm updates seem to come in two main flavors. There’s very specific updates — like the Page Experience Update or Mobile-Friendly Update — which tend to be announced well in advance, provide very specific information on how the ranking factor will work, and finally arrive as a slight anti-climax. I’ve

Weather report: Penguin data refresh coming today. 0.3% of English queries noticeably affected. Details: http://t.co/Esbi2ilX

— Matt Cutts (@mattcutts) October 5, 2012

This was the idea that major ranking fluctuation was sometimes caused by algorithm updates, but sometimes simply by data being refreshed within the existing algorithm — particularly if this data was too costly or complex to update in real time. I would guess most SEOs today assume that all ranking data is updated in real time.

But, have a look at this quote from Google’s own guidance on Core Updates:

“Content that was impacted by one might not recover—assuming improvements have been made—until the next broad core update is released.”

Sounds a bit like a data refresh, doesn’t it? And this has some interesting implications for the ranking fluctuations we see around a Core Update.

If your search competitor makes a bunch of improvements to their site, then when a Core Update comes round, under this model, you will suddenly drop. This is no indictment of your own site, it’s just that SEO is often a zero sum game, and suddenly a bunch of improvements to other sites are being recognized at once. And if they go up, someone must come down.

This kind of explanation sits easily with the observed reality of tremendously authoritative sites suffering random fluctuation.

Test & learn

The other missing piece of this puzzle is that Google acknowledges its updates as tests:

This sounds, at face value, like it is incompatible with the refresh model implied by the quote in the previous section. But, not necessarily — the tests and updates referred to could in fact be happening between Core Updates. Then the update itself simply refreshes the data and takes in these algorithmic changes at the same time. Or, both kinds of update could happen at once. Either way, it adds to a picture where you shouldn’t expect your rankings to improve during a Core Update just because your website is authoritative, or more authoritative than it was before. It’s not you, it’s them.

What does this mean for you?

The biggest implication of thinking about Core Updates as refreshes is that you should, essentially, not care about immediate before/after analysis. There is a strong chance that you will revert to mean between updates. Indeed, many sites that lose in updates nonetheless grow overall.

The below chart is the one from earlier in this post, showing the impact of each Core Update on the visibility of www.reuters.com (again — only among MozCast corpus keywords, not representative of their total traffic). Except, this chart also has a line showing how the total visibility nonetheless grew despite these negative shocks. In other words, they more than recovered from each shock, between shocks.

Under a refresh model, this is somewhat to be expected. Whatever short term learning the algorithm does is rewarding this site, but the refreshes push it back to an underlying algorithm, which is less generous. (Some would say that that short term learning could be driven by user behavior data, but that’s another argument!)

The other notable implication is that you cannot necessarily judge the impact of an SEO change or tweak in the short term. Indeed, causal analysis in this world is incredibly difficult. If your traffic goes up before a Core Update, will you keep that gain after the update? If it goes up, or even just holds steady, through the update, which change caused that? Presumably you made many, and equally relevantly, so did your competitors.

Experience

Does this understanding of Core Updates resonate with your experience? It is, after all, only a theory. Hit us up on Twitter, we’d love to hear your thoughts!