Autocomplete Changes Affecting Brands - Might Explain Recent Updates

[anallawalla]

I look after eight brands in the same industry, all with separate websites, so have the luxury of comparing them in Google Search Console (Search Traffic > Search Analytics > Clicks/Queries).

The symptoms of the algo update stretched over December and January. They affected our brand name searches very selectively. ("Example" is a brand name)

e.g. "Example green widgets" was affected but "Example widgets" and "Example reviews" was not.

Total click volume did not change much, as these are branded searches and most of the first page in the SERP are our pages. Hence we didn't notice this problem earlier. However, checking each query in this GSC report showed a chilling change.

Rankings never changed, but both Impressions and Clicks dropped at the same rate after 15 January and have stayed to about a third of their 2015 levels.

To cut a long story short, I looked at the Google Instant autocomplete suggestion and checked the click history for the first recommended keyword. Bingo!

Looking for "Example green widgets quote" showed what happened.

On 16 December, there was a brief but large spike in clicks and it died in a couple of days. This was perhaps a trial.

Then a flat line of low clicks for a month and suddenly a rise in clicks around 12 January, and now the clicks are at that level.

I have seen this in three of our eight sites that sell a similar product. You might like to check out your branded terms. Chances are that another of your branded terms has picked up the lost traffic, even though the total volume of clicks is roughly the same. In our case, our top money terms lost out and a lesser value term picked up the traffic. :(

Anyone else seeing this?

[aristotle]

Is it due to a change in the threshold for how often people search for that particular term, when they enter the full term manually ? In other words, if people don't search for that particular term often enough (above an algorithmically-set threshold), then it doesn't get an auto-complete suggestion.

Edit; To clarify what I meant, Could google have changed the algorithm to set a higher threshold?

[anallawalla]

No, "Example green widgets" is the best value term for us because it is the "Example" brand name plus the best head term. I don't think that 200-300 of them each day would stop being interested in that three-word term, but continue using a broader "Example widgets" when the broader term covers other, unrelated "colours". Someone needing "green" cannot use "red".

Google has possibly changed the algo, but this is so selective that it is odd, particularly across only three brands. (The company acquired similar companies over the years and they mostly serve a single state, hence so many brands.

[aristotle]

Well I meant that google probably did change the algorithm -- the part of it that determines what to show in the auto-complete. This is generally thought to be based on how often people search for a particular term, but maybe there are other factors as well. It's even conceivable that some kind of perceived over-optimization for a particular term could be involved.

[JS_Harris]

Chasing keywords at such a granular level isn't going to serve you well in 2016 now that Google is allowing machine learning (RankBrain) to play a big role in all results, in my opinion. I compare analyzing the suggestion box, which is always changing as you point out, to chasing a rabbit down a hole. It's a great suggestion tool but there is no guarantee Google would accept your domain for any of the suggestions if you targeted them. My advice: perform a "related:example.com" to find your immediate competitors and focus more on what they cover that you do not if you need ideas. Otherwise just keep building great content and tuck it away in the back of your mind that those suggestions will keep on changing over time and you didn't get stuck down that rabbit hole.

[anallawalla]

Thanks for your comment, but this is an observation within a single brand, To illustrate in a different industry, it's the equivalent of "Samsung Galaxy phone" no longer showing in the suggestion box (and your clicks drop) but "Samsung phone" is now the top suggestion and has picked up the lost clicks. In branded searches one usually dominates page 1 and sometimes a couple more SERPs.

We have 14 brands in total in this region and sometimes dominate the top 10 in unbranded searches. Fun when your competitors are yourself. :)

[Robert Charlton]

They affected our brand name searches very selectively. ("Example" is a brand name) ...e.g. "Example green widgets" was affected but "Example widgets" and "Example reviews" was not.

I have experienced autocomplete drops that followed this pattern... ie, from specific three-word phrases to broader two-word phrases... though not with branded terms.

I'm thinking that generically, at least, the pattern is understandable and not unexpected when you consider the nature of autocomplete. I don't know whether it makes sense, though, within your branded offerings.

Assuming that Google is following its longtime basic pattern of displaying suggested queries in order of search volume, your brand name "example" in the lead off spot could drive a more rewarding three-word phrase down in the list. If there are enough two-word phrases triggered just by "example" that have significantly greater search volume than your three-word phrase, your three-word phrase drops off the list. On Google desktop, there are only five suggestions, so it's not a huge list.

Note in the above that factors like QDF, personalization, and geo-location would probably override search volume in many cases.


When autocomplete was introduced, I saw it bury "big red widgets", eg, a great niche term, replacing it with highly searched two-word terms like "big widgets", "big gizmos" etc in the drop down list. I'm using "big" as an example, but I saw autocomplete substantially change search patterns for this vocabulary. Three-word terms beginning with "big", already below the two word queries in volume, simply lost even more searches. There were no added suggestions of modifiers within the two-word phrases, and to most searchers it seemed like the general two-word searches would cover it.

I also remember that when Instant and autocomplete were launched, Google claimed that searcher behavior was not changed by the suggestions. Google said that searchers overwhelmingly disregarded those suggestions and simply typed what they had intended. I didn't believe it then; I didn't see it then; and I certainly don't believe it now, particularly with mobile interfaces.

I know you're aware of all this. So, I think you may essentially be asking... why are you seeing your particular change with your terms all of a sudden now?

It's possible, eg, though I don't know whether it's likely, that RankBrain is simplifying anomalous short queries to shorter phrases. Seems to me I've been seeing this happen randomly, as autocomplete appears to be dropping superfluous vocabulary. Suggestions for phrases like "how can I tell" are becoming "how to tell". It may also be, though, that this is the way searchers are typing things in now... that it's not RankBrain at all, or at least not RankBrain by itself.

It also appears that while autocomplete is shortening some queries to essentials or less, it's also encouraging searchers to use longer queries. At the same time, it's caused some 3-word niche phrases, as I noted above, to get buried.

...this is an observation within a single brand, To illustrate in a different industry, it's the equivalent of "Samsung Galaxy phone" no longer showing in the suggestion box (and your clicks drop) but "Samsung phone" is now the top suggestion and has picked up the lost clicks.

I doubt that "example" is a brand name matters, except that it gives you more insight into what's going on, but that's just a guess on my part. If you can provide a better sense or characterization of the "green widget" phrase without getting too specific, that might help. Is "green widget" a compound word? Right now, I can only assume that it's likely that a three-word phrase would have less volume than a similar two-word phrase.

I would also look at what's replacing your bottom terms in autocomplete with regard to likely search volumes, and how competitive the phrases are. Again, with only five suggestions, it might not take many items to obscure "example green widgets". How many phrases changed? If the terms don't make sense from your traditional feeling for the phrases, then maybe it's RankBrain somehow shifting searches to the top term, but it might be some kind of Google recalibration at the bottom, moving something else into your list.

Another possibility... complete speculation... is that Google is using autocomplete to test user preferences to see where less or more disambiguation is desired. I think you'd have to track this over time to see if there's a regular cyclic behavior of some sort.

[JS_Harris]

This will be a long read but will make sense as you get towards the end I hope.

I understand what you are saying but I have to stand by what I said above, here's why. The suggestion box offers you the most common search terms but the highest ranking pages for each of those typically receives traffic for many, many more search terms. In fact a page designed to rank for only the suggested phrase will now fail miserably at getting traffic for its one target at a very high rate. In 2016 you don't even need to have the keywords on the page anymore to rank for them.

"did you see that light show last night? Man, the sky lit up so bright, I saw spots for hours. It made me feel like I was dreaming, real twilight stuff, it looked like a fireball"

The only term that sentence might return for is "comet" or "comet + location", even though the word Comet is not there. Intent has trumped keyword use thanks to RankBrain.

When you consider that Google is heading in that direction you've got to step back and look at the page as a whole, not as a sum of its keywords. Could this affect brands? Absolutely, as you point out the phone model was removed in your example. That is what you'd expect machine learning to do when it spots a trend in which the model changes but the intent stays with the rest of the phrase, or Samsung + phone in your example. If you keep that in mind you want a page to embody "Samsung Phone" and focus much less on the model of phone because that has become an interchangeable aspect of the page. ie: When the next model comes out you can re-tool the same page to be about that model, while still being about Samsung phones.

If the page always stays about the same company and product but the model swaps out to whichever is most popular that's a clear signal to Google that your page reflects what people are looking for. Instead of creating a new page for each model look at it more in terms of pages being keystones for subjects where things like model can be changed. You'll end up with a much stronger page for "samsung phone" and can always link it to previous model pages that contain only basic info(since the product is likely discontinued). ie: 1 main page about "Samsung Phones" and 1 archive page containing a growing list of past models and basic info about each. 2 pages in total, both heavy hitters for a wide range of searches... that's where google is heading with RankBrain. The 1 page optimized for one phrase days are behind us, ask an affiliate site creator how that's working out for them.

Prediction: When RankBrain evolves to become the #1 factor in determining results(it's #3 now?) and it gains full trust from engineers I'd expect it to go to work in feeding Google's coming voice results. ie: if you want info you ask a mobile device and it will answer you as if you're having a conversation. That's the next step, imo, and it takes breaking away from a keyword driven system which they are doing nicely now.

If I'm right it will HUGELY affect brands, especially when voice kicks in. The device will know who it's talking to and will learn about them. Lets say they don't live near a WalMart, you can expect WalMart never to get mentioned if the device knows the user doesn't like to order online. Perhaps they have a phone plan with a company that does not offer a Samsung phone or Samsung support(unlikely but hey), then you'd expect the mobile device to suggest another brand. Right now is the time to get thinking about that, stores would do well to ensure that their information is readily available online in case a user says "show me". That applies to restaurants, entertainment, anything the person is involved with.

"How do I sharpen my lawnmower blade?"
response: "Well, if you want to do it yourself you'll need a file. Would you like me to find one at that place you like to shop at and see if it's in stock? I don't think you have one now"(because he took a picture of his workbench and the phone didn't see one)

The future... will be fun for webmasters.
- If you regurgitate widely known specs or details, that's not value anymore
- If you give opinion, yours is a support page for another site and not a source
- If you price compare your pages will return to having little value if a system checks actual store prices directly
- If you offer guides, write them well because those will be fact checked by a machine and used

Basically get your mind around how you can help a machine help others, how websites are now will change to such a degree people will feel like "the web" is an antique. Online will be for social, facts, news, info and the rest will be spoken most likely. If a user wants to SEE a page, it will probably get auto-generated like Google's "feedback" system is now minus the link to your page(s). I'm getting ahead of myself but, well, Google is heading there so I need to lay some groundwork to that end or I'll get left behind too.

Thought: Google would thus be designing a system that ends their cash making advertising machine (adwords/adsense) as it moves towards a true conversation on a mobile device. Is Google into mobile? Are they building the platform? Are they building the AI? Do they plan on having you free to talk to a mobile device while you get self driven somewhere? Are all vehicles getting wifi to accommodate this? The only thing that surprises me right now is how much companies are investing in their online presence in ways that just won't be compatible with an "ask your device" world. Likewise the privacy advocates will have a hard time knowing if what is not said by such a system is due to censorship or not.

Instant contact, heck even contact like this in a forum between me and you, looks like it might become a dinosaur of a system in the now not too distant future. Chase that rabbit, not the suggestion tool :)

Globalization: Right now I can't talk to Dimitry in Russia(example) because I don't know Russian and he doesn't know English but such a system will become a bridge as it speaks as a middle person for us. The world will become smaller, so to speak. If two people are into the same subject and want to share ideas then language won't even be a barrier anymore... fun. We ARE heading there, quite quickly. 15 years? Every click on that suggestion box gets us closer, Rankbrain learns from it.

Bonus idea: Why do you think Google, after becoming "caffeinated", suddenly reversed course and slowed how quickly pages could reach #1 and instead has only rolled out updates that are aimed at spam without really changing any top results? My guess, it needs stability in those top spots for a while while RankBrain learns.

Google compare is toast, but it wouldn't be needed if your mobile device can answer comparison questions for you now would it. Virtually every piece of Google seems geared towards such a system in the not too distant future. They are betting the bank on people interacting with their devices by actually talking to it and asking it questions. Illiterate people who can't otherwise even SEE online much less interact with it will suddenly be able to speak to someone of a different language about anything, and find all of the info they could want and then some. It has benefits.

I'm stretching, but as a webmaster who has been stretched recently... not much choice but to see where I can get into the game!

[anallawalla]

Robert_Charlton and JS_Harris,

Thanks for your extremely detailed posts. I find this an odd situation because I am not seeing the same issue in 5 of the 8 consumer brands I manage. We have 14 but I have not looked at the other brands as they are B2B brands and some are in NZ. The Samsung analogy is perfectly apt although that is a product and we sell services.

Let me invent an example in the service industry (and assume this is some household brand you hear on TV and radio every day) -- this Acme brand sells massages - Thai, Swedish, Shiatsu, etc - so Acme Swedish massage drops off but Acme massage picks up the traffic. I cannot imagine a RankBrain designer thinking that Acme massage is a better suggestion. In our industry, someone who needs a Thai massage simply cannot get any value from the Swedish massage, despite the superficial fact that both are massages. :)

If you can provide a better sense or characterization of the "green widget" phrase without getting too specific, that might help. Is "green widget" a compound word? Right now, I can only assume that it's likely that a three-word phrase would have less volume than a similar two-word phrase.

Yes, the two-word query gets about a third of the search volume as the three-word query. "Green widget" are two words, where "Red widget" is completely useless to someone who needs the Green widget. Alternatively, think "Acme boat servicing" being useless to someone needing "Acme car servicing". "Acme servicing" has no commercial relevance to a customer as a Google Instant suggestion. It only has relevance in an academic sense that both involve servicing mechanical devices.

[jpalmer]

Has anyone considered spoken word searches on mobile devices as a factor in these results? Just m2c. ;-) cheers.

[tangor]

If Acme offers massage of various kinds, is there a reason those kinds do not all appear in one result for Acme and massage?

Single purpose pages, particularly product and service, will start falling by the way side. The two word search is more powerful than the three word (or four), but nothing prevents the user from extending the number of terms per search.... other than laziness as the autocomplete makes that first suggestion.

BTW, works the same way in Bing.

[anallawalla]

If Acme offers massage of various kinds, is there a reason those kinds do not all appear in one result for Acme and massage?

Someone selecting "acme massage" sees the home page as the first result, followed by six sitelinks. The service that lost the clicks is the only one there, so it is still prominent (and so also counts as position 1). The other sitelinks are the usual spread of Contact Us, Get a Quote, Policies etc.

The remaining kinds of "massages" do not show at all in the SERP for 3-4 pages, if at all. Therefore a search of the broad branded service does not result in all the services coming up in the SERP.

You could check other service industries to see if a similar pattern emerges -- lawyers comes to mind - "XYZ family lawyer" is a person who only needs a family lawyer, not a contract lawyer or a commercial litigation lawyer.

If anything, Google is placing more emphasis on specifics.

[JS_Harris]

I cannot imagine a RankBrain designer thinking that Acme massage is a better suggestion

That's what you conclude when you think along the lines of keywords. I'm suggesting that Google, via RankBrain, has moved beyond keywords which makes the text in the suggestion box somewhat irrelevant to a webmaster seeking to draw data from it. The text displayed in it is designed to separate different data sets but those lines can be drawn differently by a thinking computer not focused solely on keywords. they can also be re-drawn repeatedly while RankBrain refines its data sets. In fact I see my own pages ranking for keywords that are not even mentioned once on the page. That box is now a rabbit hole, I'm telling ya!

[tangor]

I'm suggesting that Google, via RankBrain, has moved beyond keywords which makes the text in the suggestion box somewhat irrelevant

True, and like all close relationships there will come a time when one says to the other: "Honey, please shut up and listen to what I AM SAYING and not WHAT YOU THINK I WILL SAY". When "honey" doesn't listen because they know better the inevitable divorce will soon follow. Or a separation (ignoring autocomplete).

I like science fiction as much as the next person, but in the meantime I will code clean, neat, CONTENT PACKED pages and continue the vigil (and speculation) of where things are going. What I do see is some changes that might be good and many changes that SHOULD be made and have not, and some changes that will continue the "wrong turn" policies of recent years (last 6 for sure!).

[Robert Charlton]

If you search for [example green widgets], is your desired page ranking?

[anallawalla]

That's what you conclude when you think along the lines of keywords. I'm suggesting that Google, via RankBrain, has moved beyond keywords which makes the text in the suggestion box somewhat irrelevant to a webmaster seeking to draw data from it. The text displayed in it is designed to separate different data sets but those lines can be drawn differently by a thinking computer not focused solely on keywords

If RankBrain truly works like that, I have no further comments on it.

If you search for [example green widgets], is your desired page ranking?

Yes, to the exclusion of example red widgets, blue, pink etc. Despite that continued #1 ranking (actually a home page with six sitelinks including the desired one). However, the majority of searchers no longer continue to type that search in full. They seem to accept an autocomplete, which is for every other colour but green.

[Robert Charlton]

If you search for [example green widgets], is your desired page ranking?

Yes, to the exclusion of example red widgets, blue, pink etc. Despite that continued #1 ranking (actually a home page with six sitelinks including the desired one).

Just to be clear, this sounds like the home page is ranking, with six site links, including green widgets, which is not the same as the green widgets page ranking.

To illustrate what the pattern I'm trying to describe... if I search for [webmasterworld]...
- "webmasterworld google" is one of the autocomplete suggestions
- and when I do enter [webmasterworld], I get the home page above six sitelinks, one of which is 'Google SEO News.....'.

If I search for [webmasterworld google]...
- 'Google SEO News and Discussion: - WebmasterWorld' ranks as the main page, above six other sitelinks.

My point is that it is 'Google SEO News' that's the ranking page for [webmasterworld google]
...and the sitelinks are in fact chosen by another algo, but I wouldn't say that the sitelinks are ranking for the query I entered.
- they're actually six top results from site:webmasterworld.com

But, again, it is 'Google SEO News' that's ranking for the query I entered, which was
[webmasterworld google]

As you can see, this is analogous to the [example green widgets] search... but that query, if I'm understanding you, doesn't cause your Green Widgets page to rank.

What I'm getting at is that autocomplete results, in addition to being chosen by frequency of search, must have a page in the serps that ranks for the query. That's one of the conditions of inclusion in autocomplete.

From what you've described, I thinking that in this case there isn't such a page that ranks. I'm not sure this is so... but that's what it seems to be.

If it is so, your problem appears to relate to why that page doesn't have enough juice and relevant text to rank for what should be a navigational search. More to be said about that, but apparently the brand name is so strong that either "green widgets" is being overpowered, or there's a problem with the green widgets page on that site.

As you point out, the reduction in the number of searches is going to make it hard to recover an autocomplete position for that query.

You would have to run through other colors and examples to see what might be going on with the rest of the pages. Hoping this makes sense.

Autocomplete isn't intended, btw, to provide results we're happy with. If it were, those "scam" results, etc, that plagued so many sites would never have been included.

Whether the results within a domain are useful depends on all sorts of factors within the domain. One of the few times I've ever "noindexed" a page to keep it out of the serps was when a Terms and Conditions page, with some fairly strong buyer-beware type warnings, turned up in sitelinks along with our main product pages... and I decided we simply didn't want the page ranking in search instead of a marketing page.

Anyway, at this point, I'm not sure that RankBrain has anything to do with it. I think it's more about whether you have a green widget page that's ranking for a sufficiently large number of queries.

Re the Samsung phones... if you enter a query for "samsung phone", you will get a page in search that links to the next level down of a hierarchy that's clear to me. I don't know whether the page displays all their models, but it's what they've chosen for now. In distinction to what we're discussing... Google isn't returning, though, the "samsung" home page. This even though the individual model pages are more popular than the Phone page (though for different queries).

More thoughts on service searches vs phone searches later if they're of interest.

[aristotle]

Sometimes you have to type part of the second word before you see what you want in the auto-complete. Most likely this is what some searchers for your product are already doing.

So as a test what does it show if you type the brand name as the first word and then type part of the second word?

[anallawalla]

aristotle,

So as a test what does it show if you type the brand name as the first word and then type part of the second word?

Pretty much unrelated terms. If I keep typing the full term, the suggestions disappear to zero. Not even the last letter coaxes out the suggestion. OTOH, if I search for ACME pink widgetting, it is normal behaviour. Since I have eight (and more) brands to check and most offer the same service to different states or market segments, I noticed that the other colours have actually gained a boost in their other branded colour traffic on the same dates onwards.

If I were a doubting type, I'd venture that a specific niche is being experimented with. To use a travel industry example, it's as major a term being deprecated as "ACME NYC harbour tour", while "ACME Boston harbour tour" and "ACME Honolulu harbour tour" have been given a boost. The latter two are far lower searched terms. The user is made to type the full term if they are determined, else they could be swayed. Someone needing a harbour tour in a specific city gets no joy when presented with irrelevant hints. Making them navigate from ACME's home page isn't helpful.

[anallawalla]

Robert_Charlton,

Just to be clear, this sounds like the home page is ranking, with six site links, including green widgets, which is not the same as the green widgets page ranking.

Correct.

This most valuable term's page only shows as one of the sitelinks below the home page. It also brings up third party sites such as review sites, comparison sites etc.

Any other colour's SERP does not bring up the home page + sitelinks, but a less interesting SERP that has the 2-3 pages directly related to this colour widget, followed by the home page, contact page, PDF results from the domain, and finally the third-party domains. Well, still six discrete listings from our domain rather than six sitelinks.

...and the sitelinks are in fact chosen by another algo, but I wouldn't say that the sitelinks are ranking for the query I entered.
- they're actually six top results from site:webmasterworld.com

Agreed. Our remaining five sitelinks are "important" pages for someone dealing with us, e.g. Contact Us, Login, etc and not much of a connection to the desired product/service.

What I'm getting at is that autocomplete results, in addition to being chosen by frequency of search, must have a page in the serps that ranks for the query. That's one of the conditions of inclusion in autocomplete.

No. The most valuable page does have a page in the SERP, except it is in a sitelink. But it is no longer a suggestion. It has twice the search volume as the broader term that omits the colour.

Moreover, this pattern of dropped clicks and impressions has only shown up in three out of my eight brands.

We are the absolute market leader in this country in this niche (all brands combined) and most of the coloured widget pages have fairly similar content. Although content keeps changing in small doses, nothing significant happened to these three sites.

As you point out, the reduction in the number of searches is going to make it hard to recover an autocomplete position for that query.

That's part of the puzzle. I don't know how often the Keyword Planner data is updated, but since our most popular product has stayed #1 in branded ranking, we are in a Catch 22 here, if people are being distracted by an irrelevant suggested autocomplete. I am confident that they land on the "wrong" page but can work their way to the desired product. (I can check path flows to see how that journey has changed) since 15 January.

Autocomplete isn't intended, btw, to provide results we're happy with. If it were, those "scam" results, etc, that plagued so many sites would never have been included.

I agree to some extent when it comes to amusing examples that result in magazine articles being written on that premise. If "McDonald's Big Mac" dropped out of the autocomplete but "McDonald's Chicken Nuggets" remained, it's not a question of being happy with the change. I am guessing that Big Macs are their biggest sellers. Never mind the McD execs being annoyed, a customer might get annoyed and Google surely wants to keep users happy, right? ;)

Interestingly, the suggestions that appear for typing just "Samsung" (google.com.au) are curious (I am an iPhone user) - I'd have expected the new s7 to be the top suggestion, but not so. They are:

• samsung galaxy s5 release date
  
• samsung galaxy s5
  
• samsung galaxy s6
  
• samsung australia
  
• samsung galaxy s7
  

and three more.

I cannot believe that the algo thinks that the 2013 release date for the s5 is the most useful query for someone interested in Samsung? :)