Odd GoTo suggestion tool results

Now I'm all confused

[georged]

I wanted some suggestions around 'machinery', so typed it in. In the bottom half of the list I found that in October there were 51 searches run using these search terms:
'metalworking industrial supply cutting tool power tool abrasives carbide tooling machinery shop supply precision measuring mro cnc'

Now it's conceivable that it's the same guy/company every time, or perhaps an optimiser, but 51 times the same terms? I am really, really puzzled. Anybody got any idea why this might happen?

[bigjohnt]

Perhaps an error in loading his WPG or other reporting software? I see this alot. Copying and pasting a list without line breaks would send it as a one string query, every time the report is run. Whomever probably kept trying it, confused by the results.

[NFFC]

I think they just sample searches and then estimate what the monthly total would be.

[Woz]

They also seem to be playing with the results using an AutoCorrection algo. This has affected misspellings as discussed earlier, but also the number of choices available.

One example is an industry of interest to me where the main product has two related names:- 1 = word1 word2 (with a space), and 2 = word1word2 (without a space). They used to be listed seperately, now they are combined.

It seems to me that whilst they are cleaning up their searches, they are also reducing the number of options available to bidders at a time when bidding is getting a little out of hand with the new alliances.

Interesting! I would actually be looking at providing more targeted variety in the searches.

Onya
Woz

[tedster]

>>I think they just sample searches and then estimate what the monthly total would be.<<

Yes, NFFC, that would explain what I keep seeing down in the lower numbers. Maybe maintaining an exact database of every search adds too much overhead.

Makes me wonder about the numbers I get for kw searches when buying banner ads from Yahoo and so on. My guess would be that they are also estimates from samples.

[georged]

So, if I'm getting this right, they might sample ten percent of searches, and if the unlikely string I referred happend to come up five times in their sample, they'd x10 it?
If they are sampling, which makes a hell of a lot of sense, would it be more likely that they might sample a couple of peak hours a day, or take every xth search?
Thanks for those suggestions, BTW :)
P.S. rhetorical:Why do things appear blindingly obvious only after people explain them? I remember the first time I heard about <noframes>, it was like being slapped about the head with a fish called 'obvious'.

[tedster]

My guess is that they would sample off-peak hours, since the extra operation would add some overhead. But all of this is guesswork on our part anyway, as far as I know.

>>Why do things appear blindingly obvious only after people explain them?<<

LOL, it's always been like that, all my life. It sometimes blows me away that all of this complexity is still a big pile of ones and zeros. That's pretty simple.

[Brett_Tabke]

You guys really think they are sampling? I don't. I think they are keeping track of every search done. They are in this for the money and the data figures = the money. I'd imagine that every search is logged. It really isn't that much to store even with the massive amounts of searches they get.

I think what you see some times in the weird results are machine generated queries. Some one will type into their se utility a string of keywords they are interested in, and then dig down through them. I see it all the time on jf [joefarmer.com]. There are so many people running automated se tools right now that corrupt results are bound to happen.

[tedster]

I concede to you on this Brett. I don't have enough experience on high traffic servers to know.

When I spoke about overhead, I wasn't thinking about disk space but response time under peak load conditions. However the point you make about corruption from automated queries is right on and would explain it all.