Google treats same page differently

depending on search terms

[snowfox121]

If I do a search on "blue widgets," my page "widget.htm" comes up in the results with a fresh date of Dec. 26.

If I do a search on "widgets" the same page "widget.htm" comes up in the results but with no fresh date.

How can this be? Both results point to the same page, which google has obviously visited on the 26th. Why this difference in the results?

[rfgdxm1]

You probably hit a different box the second time that hadn't yet got the fresh page. Google isn't one computer. It's a bunch of Linux boxen in 7 datacenters around the world.

[snowfox121]

oy, thanks.

[soapystar]

may not be the case here..but ive noticed that google will serve the most relevant version of my homepage.....if the fresh tag is more relevant i get that..sometimes ive seen it serve a chached page more than 3 months old....

[Powdork]

You probably hit a different box the second time that hadn't yet got the fresh page. Google isn't one computer. It's a bunch of Linux boxen in 7 datacenters around the world.

I posted this same question once before and got the same answer. However, as I'm sure is probably the case with the lovely snowfox, no matter how many times you try the two searches, it keeps happening. This would suggest that different searches automatically are routed through different centers rather than randomly or according to traffic. I doubt this and am inclined to follow soapystar along the lines of relevancy. That is; for awhile, while the minty freshness is being updated, the index can consist of both pages (fresh and stale) and will return the most relevant of the two.

[jomaxx]

Similar experience here. The fresh page normally comes up, but for searches that don't match the fresh version, the regular index page can be served up instead. And quite a few times I've seen pages from previous indexing cycles as well.

[stevenha]

I thought of another possibility. I am presuming that Google pre-calculates and stores the SERPs for thousands-or-millions of the most common search requests. Doing so could save a huge amount of computations. ( But it would still do the full ranking computation for rarer complicated searches.)

In order to mix in minty fresh results, Google would probably create a cycle for "expiring" its pre-calculated SERPs, so that new ones can be re-calculated periodically. Depending on the randomness of that cycling process, the amount of minty fresh results available will vary. Does that sound plausible? (As usual around here, I'm totally guessing.)

[snowfox121]

That is very interesting. Now 24 hours later, the Google results are exactly the same. i looked at the cache and the one that is not fresh goes back several months. It also DOES relate to a common internet search phrase, so you might have something there.

[rfgdxm1]

Interesting theory, stevenha. It would make some sense to cache the most commonly searched phrases, rather than have to them over and over again. This cache could have a time to live value the same way ISPs do the same thing with DNS caching. Once the cache hit the time to live value, it would be deleted, and replaced with a new cache the next time someone did that search.