The Power of Search Suggestions

Or how 5 k uniques a day can vanish with no apparent reason

[vivalasvegas]

About a month ago [webmasterworld.com] I was asking for your opinions in regard to a drastic traffic loss one of my websites had suffered. Here's a reminder of what happened: 3 word phrase had been sending me around 5,000 unique visitors a day for over a year. I remember I was surprised to receive quite a few visits on this phrase starting about 2 years ago. At the time I was ranked 5th. The keyword tools I was using didn't suggest the phrase in question to be so popular or competitive. But we all know that these tools are not always accurate so I didn't pay much attention to the "little" detail. Instead I made a few onpage changes in an attempt to get the first place for the phrase, especially that my website offers exactly the information people were searching for. A few months later my goal was achieved and the bonus: 5 k uniques/day. Traffic was so steady for over a year with so little competition - I'd already started to take it for granted. Until July when suddenly it dropped to 200 visitors a day. No apparent reason - my website was and still is ranked first. I couldn't find an explanation and I was sure someone was somehow steeling my traffic. Today I used the Insights tool to search for my phrase. Guess what - dramatic drop in searches starting exactly the day I noticed it.

My conclusion after doing some more research: the apparently popular 3 word phrase was made popular by the Search suggestions feature. It seems that people were typing in the first 2 words (or even just one word and a half) and the first suggestion was my phrase. Combined with the fact that the first 2 words make a very popular phrase with several million searches reported per month - the result was some nice traffic spilled in my direction. Needless to say - the 3 word phrase is no longer a favorite suggestion.

[foolsgold]

We wonder if these are changed to reflect seasonal trends? Will be inetersting to see if one we watch, and was dropped at the 'right' time of the year, will come back.

[1script]

As a user, I'm in love-hate relationship with search suggestions myself. I find it useful only in times when what I'm looking for is the suggestion #1. Otherwise it distracts me and sometimes Firefox starts feeling playful and won't let me complete the sentence with my own words and insert one of the suggestions instead (must be a mouse problem,actually).

But whatever it is, it's a complete traffic killer if you were ranking for very long tail keywords: by the time a user would type everything in, they'd be bombarded with dozens of suggestions and chances are they'll give up and take one instead of what they originally wanted to type in. Or, as my experience tells me, their Web browser would do the honors and pick one for them.

Google (and Bing of course) are not only trying to guess the users intent, they are in very real terms shaping it.

[Robert Charlton]

While the Suggest feature is intended to motivate searchers to refine their queries, it generally works well for that purpose only if the additional modifiers are at the end of the query.

If you're targeting [great red widgets], eg, but [great widgets] is extremely popular, you may find that after searchers have typed in just g-r-e-a that [great widgets] is suggested... and many searchers go with that.

Suggest can paradoxically lead to more searches for the general phrase... not the disambiguated phrase that might have been searched without it. This can also drain off a lot of traffic.

[1script]

Suggest can paradoxically lead to more searches for the general phrase... not the disambiguated phrase that might have been searched without it. This can also drain off a lot of traffic.

And helps Google to ease the CPU load by letting them cache results of much fewer queries. Additionally, having fewer word in the query in the first place makes building the SERPs a less intensive CPU task - also a welcomed characteristics for an overloaded Web property.

I plan on experimenting with search suggestions on my own sites for exactly the CPU load, not necessarily user experience reasons. Site searches are very burdensome and I've seen many sites resort to putting captchas and other means of slowing down search requests. I would much prefer to gently nudge my visitors to search for terms that have already been cached instead.

[blend27]

And helps Google to ease the CPU load by letting them cache results of much fewer queries.

Or maybe it helps Google to show more expensive ads if the suggested search applies to some extent. The more generic the term user arrives to at Gorgs Serp the more expensive clicks will be for the Adwords for the companies to get in a top 3 of the Ads.

That is the way I'd do it.

[incrediBILL]

Likewise, suggestions sometimes appear that cause traffic to ramp up and also vanish, it's a fickle beast.

Or maybe it helps Google to show more expensive ads if the suggested search applies to some extent.

Suggested search probably does boost Google's revenues because it creates an artificial limit on the number of search terms being used by the average searcher. Fewer search terms mean that only the best sites will ever rank for those search terms. The net result in theory should be more site owners being forced to advertise and compete in AdWords on those suggested terms.

This is a situation where smart people that dominate those suggested search terms can make out like bandits given the right circumstances.

Assuming you have a site that ranks high for one of these terms, and selling competitive advertising isn't a conflict of interest with your site in general, you can offer direct advertising at a rate far below what Google charges yet more than you would earn off AdSense with their smart pricing.

Don't get too greedy and you'll make out like a bandit as my direct ad sales bloomed after suggested search was implemented.

People afraid to get their hands dirty selling direct ads are just leaving big bags of sticky sweaty cash laying on the table.

Don't say I never gave anyone a freebie :)

[blend27]

big bags of sticky sweaty cash laying on the table

and that made me smile... Dorothy, lock that door, would ya...

[vivalasvegas]

I almost never use these suggestions. I know what I'm searching for and don't pay attention to them. In fact I type pretty fast so I don't even see them. I guess this is why I was so surprised to find that so many people use them.

Suggest can paradoxically lead to more searches for the general phrase

I guess we're back in time. Advice such as "try to rank well for longtail phrases instead of the most competitive ones" may not sound so wise these days.

[kaz]

For every down there is an up. Someone else is getting that traffic. Part of keyword research should include reviewing search suggestions.

[LostOne]

Lot's of potential traffic for local with my big two word search phrase... After finishing the phrase seven out of ten suggestions are keyed to local results.

Example:

Green widgets...

seven of ten suggest terms are towns or suburbs around Atlanta

[paulguy]

I've been reading here for a while but holding back from the conversation. I've finally got to say though, that more and more I get the feeling to NEVER take Goog's traffic for granted, and that if your business is dependent on Google search for survival, you'd best be looking for another way to keep numbers up now... What shocks me is that some big newspaper has yet to put a solid investigative reporter on the search beat. Here one little change from Google can make or break a business, but no one appears to be watching...

[tedster]

Coinciding with the big change that removed the host crowding filter for some searches [webmasterworld.com], there seems to have been a significant positive shift in the Search Suggestions.

For the search terms I was tracking at least, no longer are there suggestions for "Company scam" or "Company ripoff" when there are no significant search results for those searches (thank you, Google!)

These negative suggestions were creating a reputation problem based on nothing more than search frequency. Google even lost a lawsuit in France about the practice last spring. To be clear, when there are published negative opinions the suggestion still may appear - and that's fine.

[Sgt_Kickaxe]

For every solid article about the primary keyword set it seems you need a couple of checkdown articles covering the lesser sets. This can happen naturally when you write about something more than once so just make sure to link to the primary page from each of the others within the content.

It's best to simply produce content to the point that when you look at such a keyword list you are ranked on page 1-3 for every combination. A solid internal link structure and good SEO practices can get you that far assuming you have a content rich site. When you've reach that point you can begin to sort it out and adjust the pr flow to move specific pages up. Trying to do it too soon results in a wasted effort when you later add content that changes the structure again.

Also, if you're feeling that there is an aspect of a subject that needs expanding, but you've already covered it 3 times for example, don't be afraid to modify/add to an existing article.

[commanderW]

This is quite a phenomenon. I hope you were able to monetize it phenomenally :)

It is also phenomenal that you were able to finally figure out how it occured ! Too bad it was after the fact. But now you have this new knowledge to go on. So do the rest of us. Thanx for posting that. This seems a pretty big SEO insight to me.

I can see where such a thing becomes ephemeral however. I would expect search Suggestions to have some relation to current search request statistics. Typing in the letter 'b' gives me British Petroleum in the sixth position. But on the other hand, the first three are static local business -largest bank, largest electronics discount store, largest public transportation service (The fourth is for Bing !)

Maybe next you will figure out what it was that changed for your suggested result. Why does the Google bar show something different now for that third word. For instance, did something change in the social sphere, in the search statistics that inform the result (presuming that some search statistics are informing the result at all) ? Or maybe Google changed the way it chooses this third word. You might be able to find the trend and/or the method that controls this suggested result. That will be even more amazing !

[cls_wired]

As soon as suggest cropped up, Black Hats immediately started attemts to use it to put their' keywords into suggests.
Of course, Google use some kind of antifraud, but not all of that "fake suggest" is possible to suppress.
I won't show examples here, I am just observing that field, not taking part.
But if you have found your site loosing traffic, it's useful to check suggest as well

[tedster]

Good observation, cls_wired

There's a nifty and free online tool for 1-click scraping of Google's Search Suggestion content - Übersuggest [suggest.thinkpragmatic.net]. I use it for keyword and even full-blown content ideas, but it's also a potential diagnostic for a traffic loss that can't be explained by ranking changes.

[phranque]

i recently learned about this which might also be useful for some:
http://www.google.com/complete/search?output=toolbar&q=keyword+phrase [google.com]

[aakk9999]

I am wondering what these numbers mean. For the few keyphrases I tried, the numbers are completely out of whack with the google keywords tool. Not just by volume reported - suggest search numbers are way higher (between 1000 to 50000 times higher than keywords tool) but there is also a different keywords order by search volume between the two(ordering suggested keywords by search volume order does not tie in with the order of volumes reported in google keywords tool).

One thought was that maybe this is the total search volume for the keyword from Google foundation to today (perhaps explaining a different keywords order by volume), but even when I divide suggest volume with 4745 (13 years in days), I get daily number of searches way too high for a few keywords tried.

[phranque]

i haven't been able to find an explanation for the significance of the "num_queries" value.