Unfor­tu­nate­ly for dig­i­tal mar­keters, there is no yel­low brick road that leads direct­ly to good search rank­ings. How­ev­er, at SMX Advanced recent­ly, Rand Fishkin, the so-called Wiz­ard of Moz, looked into his crys­tal ball to pro­vide some insight into what may mat­ter to search engines and impact rank­ings in the future.

Smart mar­keters who want to stay ahead of search algo­rithms and be found by con­sumers at crit­i­cal moments know they need to watch where Google is going.

“It’s very pos­si­ble these sig­nals are pure­ly spec­u­la­tive, but I think there is inter­est­ing evi­dence and/or patent appli­ca­tions from search engines sug­gest­ing it might hap­pen now or in future,” Fishkin said.

Here’s a look at Fishkin’s take on some more spec­u­la­tive rank­ing sig­nals, as well as his insight into what Google and Bing might use to rank results going for­ward.

1. Usage Data Of Pages & Sites

Fishkin uses the exam­ple of a query for “cloud com­put­ing” and points to Google Chrome and Android device usage as per­haps ele­ments that are already inform­ing Google of what con­sumers are search­ing for. In this par­tic­u­lar exam­ple, a site with few­er vis­its per day ranks high­er because Google is send­ing searchers to the page with greater vis­i­tor loy­al­ty and engage­ment, he said.

And in anoth­er search for “silk flow­ers”, he notes this type of rank­ing input could be behind the strong per­for­mance of pop­u­lar brand sites like Michaels and Ama­zon in queries where clas­sic SEO ele­ments are lack­ing, like poor key­word tar­get­ing and rel­e­vance, as well as few links, because the sites prob­a­bly have stronger traf­fic and engage­ment than their com­pe­ti­tion.

2. Accuracy vs. Popularity Of Information

Fishkin point­ed to a query for “do vac­cines cause autism” and not­ed the web­site www.howdovaccinescauseautism.com ranks #2, which he said was ini­tial­ly trou­bling until he clicked on the link and dis­cov­ered the site says, “They fuck­ing don’t.”

“It makes you proud of the Inter­net,” Fishkin said. “By look­ing at mul­ti­ple data sets, you real­ize an algo­rithm could deter­mine the con­sis­ten­cy of accu­ra­cy shown by a giv­en web­site and poten­tial­ly boost rank­ings based on that.”

3. Query Structure As An Anchor-Text-Like Signal

Not­ing this par­tic­u­lar sig­nal is “high­ly spec­u­la­tive,” Fishkin said how we con­struct search queries may be a fac­tor that deter­mines rank­ings in the future.

In oth­er words, many searchers are using query struc­tures in a par­tic­u­lar fash­ion that could con­nect brands and mod­i­fiers to key­words, which, he said, opens an oppor­tu­ni­ty for mar­keters.

For exam­ple, a search for “sun­glass” could bring up sug­gest­ed results for the brand “Sun­glass Hut.”

What’s more, pop­u­lar search­es around a brand could indi­cate asso­ci­a­tions that man­i­fest in rank­ings input, he added.

4. Brands As Entities, Entities As Answers

Brands are increas­ing­ly becom­ing enti­ties in Google’s Knowl­edge Graph, Fishkin said. And in many com­pet­i­tive SERPs, there seems to be a cor­re­la­tion between brand drop­downs and high­er rank­ings.

“I see folks say­ing brands aren’t some­thing Google talks about, but enti­ties are,” Fishkin said. “Brands are becom­ing what Google clas­si­fies as enti­ties.”

In oth­er words, Google can see peo­ple also search for not just a giv­en query, but a spe­cif­ic enti­ty, which is “at least sub­tly show­ing some brand bias,” Fishkin said.

He also not­ed there is a strong cor­re­la­tion between brand drop­downs and high­er rank­ings.

Using the exam­ple of a search for “Seat­tle real estate mar­ket,” he points out that real estate web­site Zil­low is so tight­ly con­nect­ed to the key­words and asso­ci­at­ed with the query that it appears in the sug­ges­tions drop­down.

“You just want to be what peo­ple asso­ciate with the query,” Fishkin said. “Now you own your traf­fic,” which, he added, is a poten­tial SEO tac­tic for the future.

Google Sug­gest also shows brand queries that earn strong con­nec­tions to URLs. Even some gener­ic queries – like for “maps” or “toys” – bring up brand­ed domain sug­ges­tions for Google Maps and ToysRUs.com.

Fishkin uses the exam­ple of ask­ing a room­ful of peo­ple – like, say, the SMX crowd – about what results they expect for a giv­en query, like, “tech news,” or, “New York best restau­rants,” and how it is fair­ly easy to pre­dict which brands will rank high­ly and then say­ing, “Yup,” upon see­ing expect­ed results.

In fact, Fishkin said the best way for brands to rank in 2018 is to “find a way to be the first, ‘yup,’ that every­one yells out in the room.”

5. Tracing The Visit Path To An Answer

Google wants to dis­am­biguate the query path to get con­sumers to com­ple­tion faster, Fishkin said.

In oth­er words, if Google sees that many peo­ple who per­form these types of queries – like for “best ramen noo­dles,” “instant noo­dle brands,” “tasti­est pack­aged noo­dle,” etc. – even­tu­al­ly end their queries on the top­ic after vis­it­ing instant noo­dle review site The Ramen Rater, Google might use the click­stream data to help rank that site high­er, even if it doesn’t have tra­di­tion­al rank­ing sig­nals.

6. Weighting Elements Of User Experience

Since launch­ing its Pan­da algo­rithm update, Google has tried to sur­face not just qual­i­ty con­tent, but high-qual­i­ty web­sites, which is why Pan­da can hurt a site with good con­tent by hav­ing lots of bad con­tent on it as well, Fishkin said.

And if Google isn’t already doing it, it is at least think­ing about how to mea­sure UX and rank sites that do it bet­ter high­er, he added.

7. Replacing Flawed Humans With Deep Learning Machines

Fishkin also points to a shift from human-cre­at­ed algo­rithms to machine learn­ing, which he said “changes the equa­tion for how we are think­ing about SEO.”

In machine learn­ing, a com­put­er is giv­en images from across the Inter­net and trained to rec­og­nize and qual­i­fy said images and com­pa­ra­ble images as pic­tures of, say, cats.

Deep learn­ing takes it one more step, Fishkin said.

“We give the machine a bunch of pic­tures and we don’t have to give it cat­e­go­riza­tion, but it fig­ures out these things should be cats,” Fishkin said. “That’s super pow­er­ful and sug­gests if Google can do that for YouTube videos and images, why not for rank­ing ele­ments? Replace YouTube with the Web and cats with any giv­en search query, and it’s not hard to imag­ine Google cre­at­ing a deep learn­ing rank­ing algo­rithm.”

For exam­ple, Fishkin said that when he search­es for “Paul Gra­ham,” Google knows that based on his his­to­ry, he is prob­a­bly look­ing for Paul Gra­ham the pro­gram­mer and not the pho­tog­ra­ph­er.

And, in the future, even Google’s search qual­i­ty engi­neers may have no idea why some­thing ranks or whether they’re using a par­tic­u­lar fac­tor in the rank­ing algo­rithm, Fishkin said. The machine will sim­ply ask, “What algo­rithm pro­duces results that searchers engage with best?” and then make it, he said.

For Google, It’s All About User Experience

In addi­tion, accord­ing to Fishkin, Google seems to be going down what he calls “a strange path” in which total searchers, num­ber of searchers, and search­es per searcher are going up, so he ques­tions whether Google is sac­ri­fic­ing ad impres­sions to make searchers hap­py and whether Google is will­ing to take away queries that pro­vide rev­enue.

In the end, Fishkin said he thinks Google is think­ing long term.

“They want addict­ed searchers pro­vid­ing data about them­selves so they can charge more per ad unit,” Fishkin said. “Face­book has shown Google that more data about users yields more dol­lars per impres­sion and click. I think Google will chase bet­ter UX to almost any extent in order to keep searchers and get data, even at the cost of their exist­ing mod­el.”

In oth­er words, Fishkin said Google cares about user expe­ri­ence almost more than mon­ey and could be sac­ri­fic­ing rev­enue direct­ly to pro­vide a bet­ter expe­ri­ence, which he calls “pret­ty inno­v­a­tive and brave.”

“Face­book has shown that by hav­ing all that data, it can charge more for impres­sions and clicks than Google can, which is scary to Google. That’s why there are so many algo­rithms around users, user expe­ri­ence, click­stream mod­els,” Fishkin said. “It’s pre­dict­ing people’s hap­pi­ness. If you keep chang­ing to bet­ter the user expe­ri­ence in order to keep searchers, you can keep col­lect­ing data. Google will chase bet­ter UX to almost any extent.”

What’s your take on Fishk­in’s poten­tial future search sig­nals?