Google's Seemingly "Snap Reactions" to Changes on a Website

[JD_Toims]

I've seen a number of posts lately about a change being made on a site or even only advertising [Adwords] one day and a site's visits dropping significantly the next.

The frequency of posts somehow related to the preceding makes me wonder if there's be some type of update we've missed? I know, yes, really, I do *know* correlation!=causation *but* the only "given" we have is Google's algo is in a constant state of change, so the question is:

Is it time for us to quit simply discounting any statement about a change made yesterday possibly having an effect on rankings today, or not?

I know sometimes when I read here it seems like we're "stuck in 2005", so I don't think it's a stretch to wonder if there's something we've missed/discounted based on "yesterday's knowledge", without even discussing to try and determine if a change in the immediacy of drops in rankings that are being reported has been made on Google's side.

[micklearn]

Maile Ohye (creator of google webmaster tools)

On a sidenote, I think she's on the GWT team but I think Vanessa Fox was the creator of GWT.

[brotherhood of LAN]

JD_Toims, indeed, how easy is it to go from an advanced discussion to establishing some very basic truths :)

Here's a proposal for the cause. Consider what might be relevant if someone were to conduct an experiment and report back here, and have that report in a structured format. i.e. basic facts about the site, age, size, traffic, niche... mixed in with specifics about changes they wanted to measure- alongside a tracking of ranks.

Surely after some volume of input we could start making some good public observation. It'd be crucial for someone with experience to knock together the structured format though.

It's also potentially easier to experiment with other peoples sites, particularly if you can get a relatively static SERP or two.

Patterns would inevitably emerge over the randomness. Just a thought.

[lucy24]

What I have seen are more frequent posts where people say they changed [something] and their rankings/traffic were "nearly immediately"

But that's your absolutely textbook case of selective causality. It's got a formal term which I can never remember, but you know what I mean. You only remember changes if there's an immediately following search-engine action-- which is then perceived as a reaction. If you changed three minor things within a week, you can't possibly know that it's the most recent change that triggered the perceived (re)action.

For the correlation to be meaningful, you have to keep a record of all your changes. Not just the ones you remember after the fact.

[webcentric]

For the correlation to be meaningful, you have to keep a record of all your changes. Not just the ones you remember after the fact.

Precisely! You need a clearly defined state of affairs to use as a baseline. Each metric that you want to eventually measure or track needs to be defined and documented.

Then if you're being scientific, you change 12 things and wait to see what happens -- NOT -- ;) (as long as I'm stating the obvious).

[JD_Toims]

Precisely! You need a clearly defined state of affairs to use as a baseline.

True, but I don't think there's any reason to not "brainstorm" about perceived changes/impact before we get "all scientific" about things, because there could well be patterns we miss if we don't evaluate the perceived-changes/perceived-impact on rankings.

EG

If we only allow "carefully tracked" changes and what seem to be "snap reactions" by Google's algo, we could miss: Webcentric made URL changes that seemed to have a negative reaction algorithmically, while I made changes to URLs, titles, H1s, page content, etc. that seemed to cause a temporary boost in rankings.

There was another fairly recent thread about someone adding H1s to pages and dropping in the rankings, but I changed H1s and got a boost, so imo, it's possible there are "thresholds" for changes -- Maybe, possibly, if you only change one thing, algorithmically it looks like a test/manipulation but if the template, source-code, titles, H1, content, etc. are all changed at the same time it looks like an upgrade algorithmically?

---

Bottom line for me is: I think we need more feedback on what seems [appears] to have impacted rankings before we "get scientific" about things, because like netmeg said, there are infinite possibilities, and more feedback/examples of possible cause/effect situations give us a better idea of what to test.

[webcentric]

OK, then I'd start by compiling a list of most commonly reported types of changes that people think might have impacted rankings...which could include just about everything you could possibly do to a site but generally might include...

1. Adding more ad slots
2. Adding a significant amount of content/pages
3. Restructuring/reorganizing content and/or related navigation
4. HTML rewrites/adjustments - everything from migrating to a new flavor of HTML, to simply changing the headings on a page.
5. Theme and plugin changes

And the list goes on and on and on but that's all I've got at 5 am...well, that and a case of insomnia.

[JD_Toims]

well, that and a case of insomnia.

LOL -- Been there, for sure!

Your list is pretty much what I remember too, and thanks for sharing it, but I don't remember the timing of the change onsite v. ranking/traffic changes off the top of my head and really don't feel like digging up the threads I've seen, because I think trying to have this discussion is a bit pointless.

We used to approach things with an "open-minded, exploratory, figure-it-out" attitude here, but that's been replaced by "close-mindedness, polarization and know-it-all-already" type responses, even though the algo was reportedly updated 650+ times last year.

Those type of responses really detract from "exploratory discussions" to the point of making those who may want to contribute to exploratory threads unlikely to post in an effort to "stay out of harms way" [or so I've heard], which means, bottom line: We're probably not going to get anywhere or figure anything out by continuing this one, unfortunately.