This is the tenth installment of our Deep Dive into Local series. For the week ending 10/30/15, David Mihm, Mary Bowling and Mike Blumenthal shared their thoughts about the previous week in local. The complete video is posted in the Local U forums (paywall). In the second half of that video, they take a deeper strategic and tactical dive into one interesting area that caught their attention during the week. In this discussion we look at Google's new RankBrain algo. You will note that Mike is present in voice only due to his very own RankBrainError in the recording.
Mike: So with that we'll transition to our Deep Dive. We have some uneducated musings about what is it called?
Mike: RankBrain, right. I keep thinking BrainRank. And it's wrong, it's not BrainRank, it's RankBrain. So I said earlier, we now have a great new bevy of potential phrases when things go wrong in local, like "RankBrainFart." I think that would now describe one of the...
David: RankBrainDead if I might put it out to you.
Mike: There you go. All right. So with that, let's kick off the Deep Dive.
Mary: So what started this discussion is that Pub Con, Gary Illyes, who is the new Matt Cutts that Google sends to conferences as their public face, was asked about all the crazy results that we've been seeing, the changing from day-to-day on local search. And his response was that these are tests, that there are ongoing tests, and that they're trying to get rid of features that searchers are not using.
So RankBrain is not just part of semantic search and that sort of thing, but it's also ongoing testing, and I think that that explains an awful lot of what we've been seeing for quite a while, although it seems to have accelerated quite a bit in local over the past I'd say three months.
Mike: Right. So my understanding of the algorithm, which apparently is being applied in somewhere on the order of 15% of all queries, is that it tries to match new query data with existing query strings that typically have related to the same entity. So for example, if you had a query string like “Knicks game" in “Madison Square Garden New York City," there's this Madison Square Garden and New York City are the entities, Knick’s Game is another entity that's frequently paired with that. But if somebody wrote a query that was different, that was perhaps, "The Cavs Madison Square Garden, New York City," they would know that the Cavs are a team, and they would then return the schedule for the Knicks against the Cavs, that kind of logic. I think as I read Bill Slawsky's article, that's what I understood that it's a way to deal with long tail queries that they never heard before, which would explain in local why you see a lot of variety, because you see a lot of really weird, long tail queries, right.
Mary: Yeah, you really do.
Mike: So back to David RankBrainDead, where does that take us?
David: Well, before we get there, I'd just like to point out that was the first public example of a sports analogy that Mike has probably ever said. So, with completion, a red-letter-day here in the world of local search.
Mike: I'm gonna head back to…..
Mary: Really, he knows two professional basketball teams.
Mike: Well, I struggled to the second one if you noticed the delay in my voice, but when I...
Mary: Where do they play?
Mike: I don't know that. But when I'm done today, I'm gonna head back to my man-cave.
Mary: Oh, okay.
David: Nice, there you go. So Mike, to your point, I just gonna pick up on it a little bit. We've seen some of this in the Google Maps “hacks” that have been reported this year. Earlier in the year there was the "N-house" and those related phrases that returned the White House. Just this week there was something reported on I think it was a link from a SearchEngineLand newsletter, somebody had reported that "death to Jews" returned a holy site in Jerusalem, sort of similarly to the N-house thing. We've noticed the “Cecil the lion” query returning the dentist office in Minnesota. A little closer to home here, if you search "Oregon shooting," that now returns “Umpqua Community College," as the single one box result in Google Maps.
So we've started to see some of this stuff, phrases that you wouldn't ordinarily associate with a specific entity, a location entity. We've seen that over the course of the year that those things seem to be increasing in frequency, and it's clear that from the RankBrainDead standpoint, that the brain does need a filter with some of this offensive…
Mike: That's my problem. I need a RankBrain filter.
David: That's right. That's right. But I think that I'm as clueless as anyone else out there as to how this thing is actually working. But it certainly seems to me to be a “how can we return more knowledge panel results for exact phrases that are somehow related to entities”. So I think that that's my sort of elevator pitch for what this thing is.
Mike: I think that's a really good example. You noted before we started that in the Upqua Shooting that it took almost a week for it to show up.
David: Well, at least a week. I certainly checked the day it happened, the day after the day after. So it was certainly there's a few days where it didn't return. It was returning Oregon shooting ranges all over the state, so...
Mike: Right. And I've seen times where it returns it same day. I'm wondering if that's some sort of batching that they're doing, or could because their index…. I mean, the Umpqua shooting was clearly national news, was in New York Times. They're indexing that stuff every day, but they must not be associating it with the entity necessarily every day.
David: Real time, yeah.
Mike: That would explain the differences between my observation and yours, which is interesting.
Mary: And it kind of makes me wonder how much of that is a result of what people are selecting in the results when somebody searches for Oregon shooting.
Mary: So to me that's the most interesting part is that of all these people that Google's kind of been poo-pooing that user behavior is affecting results. They most definitely are effecting results and now they're finally admitting it.
David: Exactly. Well, you just can't take it. I'm frankly...
Mike: Oh, David, that tin hat is coming out, I hear it.
David: It is. It’s a good day, I should have gone with the Tin Man for Halloween, I could have actually been part of this.
Mike: Yeah, I can go as a scarecrow.
David: Anybody who's take what Google has said at anything close to face value since about 2005, has I think been setting themselves back in terms of their optimization success. So I think that it's very clearly user behavior, especially click through rate and user engagement have been affecting search results for a long time. Mary, to your point, I think that this thing is clearly accelerating that affect, and the application of this thing is...the keyword setup across which this thing is applied seems to be increasing by the day.
Mike: It also portends something that we've been talking about in a tactical way, but that is the difficulty of playing your marketing strategy to the algorithm, right? If they're doing AI and machine learning, and this is adjusting to people and people adjusting to it, and there is very little input from direct engineers in terms of that algorithm, which is currently 15%, but goes to 30%, or 50%, or 70%, it just makes the idea all the more absurd of doing anything other than broad brand building, and semantic writing, and those sorts of things, rather than key words and link building, right? I think it will put the final nail in the coffin of that kind of SEO, not that there won't be a role for technical excellence in web-design and in marketing there will be, but it won't be what it used to be
Mary: Well, I think that Rand Fishkin at MOZ is spot on when he talks about there being two algorithms that we need to satisfy now: the technical SEO algorithm of having a quality website that's very crawlable, but also that we need to think of the algorithm for users. What is going to attract people to our pages, what's going to keep them on our pages, what's going to send them down the funnel that we'd like them to go down.
Mike: So with that we can rename Panda, RankCrap algorithm, and I think we can give you the last word on this Mary, and with that we'll have a wrap. How's that sound?
David: Sounds good. Just if I could put in a very quick plug, so a lot of this stuff with respect to what Google is doing algorithmically, and some of the cross-platform, cross-app searching that Apple's doing and Google's doing with "Now on Tap," This is all stuff we're gonna be talking about at Lovely Williamsburg next March 5th, 2016, and the link is on our website if you wanna sign up for that. We've got an early bird rate for the next week, pre-agenda I should say for the next week. So this is the time to do it.
Mike: We'll end it with that. Thank you very much. Have a great weekend you guys.
David: You too.
Mary: Thanks. Bye.