Transcription
Your Google reviews start disappearing. Here is something that almost every business deals with, but most people don't understand.
Now, what inevitably happens in that scenario is they dial it up a little bit too much and a lot of collateral damage is caught in that net. You make a simple edit to your business profile and it either gets ignored, reverted, or even worse, your profile gets suspended. And when you reach out to Google support, you hit a brick wall.
And that's not because Google is trying to do you dirty. It's a big company with lots of people doing different things.
It all can feel very random, but what if it's not?
Not surprising that the amount of workflows that are being managed by humans they're getting fewer and fewer.
I think that where I've had the most amount of success is
Today, I am joined by two former Googlers who are going to break down how this all works behind the scenes. I'm joined here today by two of my favorite people, Brad Weatherall and Joel Headley, both who worked at Google for I want to make sure I have this right. Was it like a decade? Brad, you were there for like around 10 years?
Yep, I was there for a little for just over 11. Yeah.
Okay. Joel, how long were you at Google?
Yeah, about about just over 11, I think is same. All right. So something like that. Yeah. I I think Brad may have me beat.
How's that?
It's It's not a competition. Um one of the things I think that's really cool with talking to you guys when we work in the world of SEO, we always see things through a lens and the stuff that you guys share with me, it's it's very different than I think things you would know if you didn't work at Google. Um so I want to start us off by talking about one of the hottest topics at the moment, which is all around reviews. There have been a few things that Google has been doing lately with reviews that I would love to get your take on. Um so just to let's talk about missing reviews to kind of start us off here. So Brad, I'm going to start with you. What causes Google to like trust or not trust a review? Cuz everyone is very upset at the moment about Google removing all of their real reviews. So, what do you want to say about that?
Yeah, it's it's it's cyclical, right? And and I think that what I mean by that is this time of year, right? The the the March-April time frame is when Google really starts looking at the algorithm that determines what is spammy behavior or not or what is a real review or not. And it's it's kind of funny because I think that if we were having this conversation in a different part of the year the lens might be a little bit different. It might be like, "Why is there so many fake reviews?" Right? Like it's like Google is opening the floodgates to to all of these reviews and then it's running rampant. And so, you know, it's it's it's a constant balancing act for the trust and safety team there that they're the team within the walls of Google that are responsible for maintaining that balance. And it's it's a tough job, right? Because, you know, on one hand, you know, if the if the algorithm that they're using to detecting fake or policy-violating reviews is is too relaxed, then you see the lots of fake reviews showing up. You see a lot of these companies that sell reviews really kind of like ramping up their efforts. And and what you see is right now in you know, in that cyclical time of that you know, like the the the March-April time frame, you know, the the trust and safety team and no doubt experimenting with new algorithms, right? And they're trying to combat that behavior. Now, what inevitably happens in that scenario is they dial it up a little bit too much and a lot of collateral damage is caught in that net, right? So, there's a lot of legitimate reviews that are being um flagged as being policy violating or or or fake. Um and you know, so that the symptom the other side of the pendulum is we get a lot of con support contacts coming in saying, "You've deleted a lot of my reviews. Where did they go? We want them back." Right? And so, um it's it's a constant balancing act and you know, I remember many conversations that I've had with the with I had with the Trust and Safety team where I'm reporting to them that there has been a a a massive spike in support contacts related to the increase of deleted reviews. Like reviews have gone missing and they they want them back, obviously, because they're um there's a lot of investment from businesses to acquire those reviews. So, like losing them is kind of a big deal. Um and so, you know, that then I think the the the Trust and Safety team will tweak that algorithm down a little bit um so that it's not as aggressive. Um they run the reviews through the um through the modified version of the algorithm and and a many of them come back. But, what we're seeing recently is that, you know, not all of them are coming back and I think that's somewhat to be expected. You know, I think that as the algorithm gets adjusted, you know, if if you were to lose 10 reviews, like with the adjusted algorithm, you may be able to get six to eight back, but like you know, those two probably going to be gone. So, so it's um it's an unrealistic expectation, I think, to to set to get 100% of those reviews back. It's not impossible, but um I just want to temper people's expectation that it's usually related to investments that the Trust and Safety team are are making to try and combat this fake behavior, right? And so, it's um it's that it's that seasonality that we're experiencing where um now's the time where they've got a little bit more opportunity to uh to experiment with those algorithms. Conversely, they're not going to do this in Q4.
Why is that?
Well, the holidays, right? It's like they don't want they don't want to mess with businesses during during Thanksgiving, during the Christmas break. That and the engineers like to take vacation at that time. So like you know, like Google will go into a code freeze and and that means that you know, the the system is somewhat stable while engineers are able to go spend time with their families for the holidays. So like it's there is some interesting reasons why there's some seasonality involved in this, but uh but yeah, I think that this is this is kind of where if I was to predict the future, we may have a similar conversation this time next year because it's it's just the time where you know, coming back through Q1, the beginning of Q2 is really seems to be a good time for the for the trust and safety team to really start trying to fine-tune their algorithm on detection of of spammy and fake behavior.
Well, I got to tell you, somebody's got to tell that to the search team cuz they always like to release core algorithm updates right before the holidays, which makes me kind of grumpy. I wish they followed that cadence. What do you guys think for trends? So right now you're saying Google's like adjusting things. Are there any patterns that you're seeing like what types of reviews or what patterns are becoming problematic that are leading to reviews not publishing? Anything you're noticing?
Uh Joel, do you want to lead off on this one?
Yeah, well, I I think if I take you back to some of the things I used to say a decade ago. My goodness, that's a long ago than that. This joy.
Long drift.
Uh I I used to say it used to be focused only on flagged reviews, right? You could flag a review, a team would look at that review, and they would say, "If you can't If a team can't prove in the content of that review that it's violating policy, they will leave it up." Right? And uh the way the trust and safety team And And that's because that's the only process we had initially. Trust and safety team came in later, right? Um we formed a little map spam team. Uh it was actually two people working on it. It's It's obviously grown since then. But the way they would do this is they would look at the reviews that they figure out are spam, right? They say, "Here's a review. It's spam. What are the some of the markers on this review that make it spamming?" Right? Well, it's coming from an IP address outside of the local area. It uh The content looks a certain way. And so, when we talk about the algorithm, there's not actually a algorithm. What they do is they have a collection of scripts and they write one script and they say, "I'm going to try to target some of these features I see in this example and see how many I can detect." And so, as a result in Brad mentioned that, they obviously don't aren't aren't looking at every review they remove. Right? So, so they're looking at the quality of an example review that they understand to be spam and saying, "How can I write a script that is going to take down that example?" And when they s- uh sample this, they go, "I've taken down 80% of these examples. I've also taken down 20% of things that aren't in this example. Am I winning or losing in this case?" And 80 to 20, we're doing a pretty good job. We're removing 80% of what I think I'm targeting here. So, the things that the team is going to focus on are things that they hear and they see themselves, right? So, if people are complaining in the space about, "Hey, do you see these spammers? The tactics that they're doing?" They're going to look at that. And then they're going to try to create rules around the tactics of spammers. So, the the things that Google is focusing on to pull down. And where it becomes problematic is that some of the things a spammer does may be some things a legitimate business owner might do. Um might And And so, when they're going out and doing campaigns, for example, and especially for a business that hadn't done campaigns in the past, all of a sudden doing campaigns, they're getting a whole bunch of reviews where they didn't get reviews previously, right? Um all of a sudden the reviews are showing up and they're showing up from accounts that have never reviewed before, right? Um those were kind of the older things. Now, they're going to be looking at accounts that are reviewing industry-specific things. So, they're going to be looking at accounts that are doing doing things a little bit fishy that actually, if you're a consumer or marketer in the space, you may be focusing on those kinds of reviews and getting them taken down. So, I I think it's the patterns that you see in that someone like Jason might be saying, "Hey, here's a new pattern of spam that we found." Those patterns are going to have a waterfall effect into this process where they're going to look at those patterns and start taking down. And I don't know it's hard to say that there's something a pattern in the legitimate takedown because those are casualties of war, not um not uh targeted. When we I think when we get a little bit too much into hey, for that casualty of war, we need to stop doing this or change so it doesn't become a casualty, it may not be exactly what you think in content or behavior that that this script particular script is targeting. So I I think it like we love to do this is it, this is what's going on, but often that's not always the right thing that we're targeting on.
I I if I may add on to that, Joy. I I think you're exactly right, Joel. I think that a lot of people in the industry have a tendency to kind of like deal in absolutes, which is I'll give you an example. So um so one of the trends that we're seeing recently is this this concept of like a review party, right? Like where you get a whole bunch of people to your location because, you know, the the signal of like oh, I'm at the office of the business and leaving my review when I'm physically there is a strong signal of truth because like, you know, Google will know that oh, you're you're geographically located at the business, you must it must be a legitimate review because you're standing right there, right? Like and and so we've started to see that that's actually working against businesses and that like, you know, our reviews are being removed when one of the signals is that oh, they're leaving reviews at the location. And so the danger of dealing in absolutes is people are like, don't get them to leave reviews at your business anymore, right? Because like Google is looking at that as a signal as as a spam signal and it's going to it's going to flag you and it's going to remove that's not the case, right? I think that the the phenomenon I think that Google is trying to identify is that a the technique that spam is able to use is this concept of a review party where you get a whole bunch of people at a location, they all leave the review at the same time at the same place. So, the proximity of the time that the review was left and the location that it was left at is part of that algorithm that's getting flagged and detected. But, that's not to say that if you're using that strategy appropriately where you're kind of like getting that drip of reviews in like, you know, every day, every couple days, every couple hours depending on the industry that you're in, it's still a very valuable uh signal and and valuable strategy for for review acquisition. So, so so it's the algorithm tweaks that the Google is making, they don't deal in absolutes. They don't deal in like, oh, if it fits this criteria, we're going to take it down. It's always weighted and the um and and the skill that the trust and safety team are really trying to figure out is what is the right balance to be able to capture the spammy behavior and leave the legitimate behavior, right? Like and minimize that collateral damage. And and Joel is 100% right where a lot of these engineers will be like, "Oh, we got 80% of it. So, like 20% collateral damage, that sounds good, right?" Um in the conversations I was having with the team, it's it's more like 90/10, right? 90% 10% collateral damage. But like when and and I used to throw this argument out there, right? When as I was dealing with um listing suspensions, I'm like, "When you're dealing with 100 million businesses, like 10% is still 10 million businesses, right? Like that's still a lot." Um and and so the scale that you're operating at is is painful. So, I I was constantly pushing to try and get that um the the comfort of the collateral damage way down, you know? But, that's um that I think Joel is exactly right in the in the categorization of it. And, you know, to to your question, Joy, around like what are the trends that we're seeing, this phenomenon of kind of like, you know, in-person reviews is is something that I've seen lately that has been a little surprising to me, but no doubt Google is responding to spammy behaviors that they're seeing elsewhere.
Yeah, there was two tips that Claudia Tuminas shared with me when I interviewed her a couple months ago and I want to know what your guys thoughts are on why these work. So, one of her tips is that if you have a filter review, you should basically get the person to edit it and just change a few words and then it sticks. And I've tested this recently and it works. Like we had a client who left us one didn't publish, literally emailed them to say just edit it slightly and now publishes. The second tip she gave was to not link directly to your business profile, which is hilarious cuz that's what Google tells you to do and they even like give you a QR code and say like use this. So, she says don't do that. Instead link to the search results that show your business profile so that they have to click around first before leaving the review, which then basically like gives the review more trust because the person's clicked on your profile. What are your guys thoughts on those two tactics and why they work or if they work?
I can tell you why they would think they work. One, edit spammers don't are are like like create and go. They They don't come back to things, right? They're focused They're They're the most iterative bunch you've ever seen, right? They're going to try something, try try try, then they're going to try something else. They're not going to worry about is that review did it still appear? They don't care. They know they're going to do 10,000 things and only 10% of what they're doing is going to show up. So, they're already suspecting a lot of what they do isn't going to work. So, they're not going to try to muddle with that 90% that would that turned out to be garbage, right? If a user comes in and says, "I'm going to edit this review." It makes sense that Google would say this is probably a real person cuz spammers that's not the behavior of spammers. So, that's one. That makes sense to me. That makes intuitive sense to me. And I I think it would make sense to the teams writing these things. The other piece is in terms of um uh what was the other one?
Not not linking directly to the video.
Right. So So you you already know how something like reCAPTCHA works, right? When it's just a checkbox, you're like I'm not looking at any pictures. I'm not filling any obscure listening to anything. You're filling in obscure cuz all I'm doing is selecting a checkbox. Well, the reason reCAPTCHA works is it it's actually looking at the behavior of your mouse and clicks. And recognizing that mouse and clicks don't move around on the screen. So even the simple thing not not so much because you've interacted with the profile, but because you've moved your mouse on the screen. They know in reCAPTCHA that that is a human and can validate and has a trust has a higher score of being legitimate, right? The less clicks that happen or the less movement that happens on a screen to a text box where you start writing the review, no mouse clicks had to happen. They're measuring less and don't have more confidence that that is a real user. So think think of reCAPTCHA and how that works today, and I think that help when it's just a button click, right? And it can be worked into uh contact forms. It's just the contact form send button. Why does that work? It's because it's measuring how someone's navigating on the page.
Yeah. I I have a slightly different take. Si- similar themes, but like slightly different. So I think that um so the so so the one which is about um editing your uh your review, you know, I think that um Google is really um looking for fresh content, right? Like fresh data freshness is is is really important and I think that So, there's there's two parts to that strategy that that might be contributing to the review sticking, right? Be creating a bit more stickiness. I think Joel is right. I think that it's not really the behavior of of spammers to kind of like go back and edit and tweak it. Um but I also think that like fresh content um and um the um the the actual editing capability itself, right? Like being able to go in and go in and show that engagement, creates a a new set of signals, right? Like where where did they edit that? Like what was the IP address of that edit signal? Because usually when they do a review, it's an algorithm that'll go through and scan the review. And and it kind of comes back to this concept of trust score. Right? Like where there is like what what is what trust does Google have in the person that's leaving the review, you know? It's a Is it a real person? Um and and this is where you know, I think the um the the other elements right? Like around, you know, like so so so the freshness and the editing themselves, is is really around like is it triggering a fresh review of the algorithm itself? So, if the algorithm originally was more aggressive and it got flagged, and you go in and you make an edit, then it gets reprocessed. Right? Like it runs through that algorithm and and maybe the trust algorithm has changed since uh since that first process went through. So, um so you're kind of like triggering the Google system to do another analysis of fresh analysis of that of that data. And the more that you engage with the profile, right? Like kind of like not not looking at the link directly, right? Like, you know, like not using the the QR code, but kind of like navigating. I I think that it's working, but I don't think it's that exact recipe. I think it's the working for the reason you think it's working. It's not like, "Oh, navigate around the the QR code to actually um leave the review." I think it works because um there's more engagement and there's more history and the more engagement and the more history increases the trust that that is a real real activity. So, I I think that you're you're describing the symptoms and it's like, "Oh, that if you do that, it'll work." And And I think that the root cause is or or the real intelligence behind this is that the more you behave like a real person, the more and and the more you engage with that profile and you you kind of like simulate that interest, the more likely Google's going to believe that it's a legitimate review and you're giving it more information for the algorithm to make a decision.
Yeah, I'll I'll throw a a new one at you guys that everyone's really upset about. Um so, Google just recently put in their guidelines that you should not be leaving like specific employee names in reviews. Like employees basically should not be telling you to mention them by name. And this is a problem because so many businesses have strategies around this. Like I've talked to home service businesses and this is how they incentivize their staff. Their staff literally get bonuses based on how many reviews they get. And this is a great way for businesses to like actually get their staff to care. And then same with restaurants. Can't tell you how many restaurants I go to where the server gives me a card and says, "Hey, my name's Patty. Leave me a review and my boss will give me a bonus." And like so many businesses are doing this and they're really really upset with this new guideline. Um do you guys think this is even going to be enforced and like why is it even a guideline? I don't I This one is like perplexing to me.
Yeah, I'll I'll kick off on this one, Joel. Um so, like this is this is interesting to me as well, right? So, like I've I've I've been out of Google for for 2 years now. So, like clearly I don't have the um the the insights for this particular scenario, but my um but my speculation is that um there's there's likely a reason behind this, you know? And And it could be um like uh protecting minors. Like you use the restaurant example, right? So, it's like, you know, what if the the the waiters and waitresses are 16, 17 years old, right? Like then it's like protecting identity. And I know Google took personally identifiable information incredibly seriously.
Yeah.
And so, when when you're looking at identifying individuals, you know, I think the the general rule was that like if this is a public figure, right? Like if you if it's like, you know, someone who is promoting themselves as the business, you know, like then it's it's kind of fair game, right? Because like that that you know, you can you can review that person because the business is that person. Um but if it's a restaurant and and even dare I say a chain restaurant, right? Like and you're starting to now make reference to individual employees that could hypothetically be under the age of 18, then it gets a lot more sticky, right? And and and it's more more risk involved for for Google and and and being defamatory and and all that type of stuff. So, you know, my my guess is that there's probably some some story behind that as to like why they're kind of like leaning in. But I also think that my my general experience with the guidelines is that they're they're also intentionally kept vague. Um you know, because they Google wants the flexibility to pick and choose how they apply judgment to that, right? And and so, you know, it's it's it's I think an overstatement to kind of like look at the policies and say um you know, like this is this is gospel. Right? Like because um you know, a lot of the times when I was having internal debate with with the policy organization because internally the way that it worked at Google is there is like teams that will set policy for Google overall, right? It's like these are the things that Google really wants to enforce. And then there's like product specific policies and guidelines, right? And so a lot of the times when we got into a debate, often it was around a suspension or a review that was either left up or taken down, and we were debating the merit of a particular review. A lot of the time, the policy team would just say, "This is a business decision." I I used to I used to hate that phrase, right? Because basically what that meant is, you know, the the product manager of that organization gets to make the call. Like meaning um it's not necessarily a violation of the Google-level policy, it's a business decision. So the product can make an independent decision on how to apply those guidelines. So it is just a fascinating way when we start talking about, you know, like, "Well, what does the guideline say?" is one thing. How is it going to be enforced, and is there going to be any disciplinary action, right? I mean, we talk a lot about, like, we all at least we have historically about name stuffing, right? Like and it's like, "Well, what is the disciplinary action?" Like, what will Google allow you to get away with, and then what are they going to um penalize, right? And and it's a it's a moving target, right? Like Google will will move the goalposts on us a lot of the time, right? Which is which is why they keep those policies and guidelines intentionally vague, right? Because they want the ability to monitor the usage of the platform and adapt to um to human behavior and to the usage. I don't know your opinion on this.
Would you guys like advise people right now to do anything different or just kind of wait and see what happens?
I mean, I think that that being being cautious is always a good thing. Like, you know, I don't know that um I I don't know that I would wipe it out completely, you know, like are you talking about the practice of putting technicians' names. But personally, I think it's going to be industry-specific, right? I think that for for restaurants where you're dealing with waiters and waitresses, I would probably advise don't don't recommend putting the the names in the in the reviews. But if you're in a service business, right? Like a plumber, a locksmith, you know, HVAC, then I would. Because I'm like that the quality of the technician actually matters to the business. So So I would my answer would change depending on the circumstances.
Okay. I actually have a different question for Joel because it was something you told me many years ago that was very eye-opening and this is not something that people mostly know. Which is this concept that Google doesn't actually delete reviews. So they're always there in a database. Like they're not gone. So I wanted to ask you about that. Do they ever get removed? Like I talked to a business that for example had a listing that 3 years ago he deleted and he was planning on moving and he deleted his old listing, which is obviously stupid. But I was like these reviews from 3 years ago, is there any hope of him getting them back? Like or after 3 years is there like a purging point or anything like that?
Yeah. Um I Brett said he's 2 years removed. I think I'm about 10 years removed. I I can't remember. So So just put that in context a little bit. So did the In general, Google doesn't delete stuff. Users delete stuff. So if if I have an account and I go to delete my account and that has reviews attached to it, those reviews will eventually get really good deleted, right? If I'm a user I choose to delete my reviews, they will actually get purged and deleted from the systems. They will get deleted from backups and they will no longer exist, right? Google does have this process and it was something that was put in years ago. A process of kind of I think it came before the right to be forgotten, but it it's it was around that time period where we needed to make sure that if a user the content creator has decided they're going to delete things, they will be be deleted. Now, reviews um follow that process, right? If a content creator has decided they're deleting their content, they will be deleted from Google systems. If uh Google has decided that the content needs to be removed, it gets hidden from being served to the public. And we know that the spammers actually see all their reviews. This is my favorite thing about spammers, right? Leading, misdirecting them, and leading them down wrong path. They always see all their reviews and any given profile. So, hopefully they think that review is still published, even though Google has stopped it from being served elsewhere. Additionally, um if if we deleted them, then they would get instant feedback saying that review got caught, right? And we didn't display it. So, we do need to keep them. We Oh my goodness. It That long and I still say that. So, Google still still wants to keep them. I I and to be able to serve them where they need to serve it to serve the purpose of helping people understand that they just were not deleting their content. But at the same time, that stuff is hidden. Uh Brad is right that that content gets reprocessed, and that's the only way reviews can come back. Now, if he I deleted a listing 3 years ago, if he figures out a way to resurrect that same listing, he may get the same reviews. If a new listing is created and he gets really creative and figures out how to tell Google can move those reviews to the new listing, that's going to be harder, but that I think that's that was possible back in the day. You could You could have reviews moved to a different listing. So, everything's there. That content continues to be content that goes through these scripts or algorithms to get things removed, and that's why you still see things pop back up. Because all it is really is a flag to say, this is flag that's spam. The rule is when it's flagged as spam, we don't serve it to general users unless you created it. And one other thing that I wanted to mention, Joy, you say Google's own rule says use the QR code.
Mhm.
We have to remember Google isn't a thing, right? The team that wrote the policy, the team that created the QR code, and the team that's removing reviews are all different teams, right?
Mhm.
And they're all working to serve the same purpose, however they attack it differently, right? So, um when Brad says the policies are intentionally vague, they have never been so detailed, right? And I think you can appreciate that most, Joy. When I when I first wrote the policy, back in my day, when I first wrote the policy, it was three paragraphs long of what you should do, right? It was very simple, and people were like, "Wow, thank you for the transparency." I don't know. It actually was something new, brand new transparency. And now, I look at that page, it is ridiculous how much information is in there. But the policy is written on that It starts with the business purpose, not how are we fighting spam. It starts with the product manager saying, "I want This is the kind of things I want people to do on the platform. And I want them to describe their business so that it matches the any signage that they have. So, so when someone drives up to see that business, they see the signage. It doesn't start with we're trying to get people not to spam categories. Usually. Right? It starts with how do I want my end user to like perceive things? And then you have maybe the trust and safety team going, since that's the policy, maybe we can write a script to uh do some behavior. Or again, it's a different team than the team that's editing. And that's not because Google is trying to do you dirty, it's because they just don't communication isn't 100% all the time everywhere, right? It's a big company with lots of people doing different things. And so when we say Google recommends this and then they're telling and then it doesn't work very well, well, they're different teams that are doing that.
Yeah, I've a lot of those stories. So, let's say you're a business, you've done nothing nefarious, but yet you're still, you know, running to this issue. You have a bunch of reviews aren't publishing and yada yada. And so you reach out to support. Is support even doing anything these days? Like I am genuinely wondering if there's any point in reaching out to support when you have missing reviews. My uh success rate on the forum lately has been 0% of Google reinstating them. Um so I'm not even sure if they're even looking or what their current workflow looks like, but um it seems like they are just like, nope, the algorithm's doing the job. Is there any manual review that you guys are aware of anymore?
I mean, there there was when I when I left. Again, I'm like 2 years out. I mean, it could have changed and uh it does certainly feel that way. I I I agree with that uh sentiment, Joy. But like, you know, there there was I I think that when when scenarios like what's happening now occurs, right? Like where there is like, you know, an an influx of uh of of requests to the support organization saying, my reviews have disappeared, I want them back. It's right, and usually when that happens, Google will create a process for dealing with those, right? Like dealing with those cases. And so, um in in that scenario, yeah, there will be kind of like a um usually a Googler will kind of like look at that flow, and they'll say, "All right, so the you know, we've got a backlog of all of these kind of like, you know, deleted review user journeys, and we need to we need a way to process them or to to resolve them." And, you know, what they don't want to do is say, "Hey, Mr. support agent that's probably operating in India, you make the call." They're not going to do that, right? They're going to They're going to [clears throat] to have a a defined process on like, "Here is how we deal with that." Most likely, it's going to be rerun it through the algorithm, and if the algorithm holds it holds firm on its, you know, that it's policy violating, then the support agent will say, "Sorry, we're going to maintain our position." Um or if it comes back and says, "Yeah, you're right, there are a few false positives in there." Um we'll let it go, but I think that it's kind of a mixture of the two, Joy, to to your point, because what's not happening is the agent is not looking at each review one by one and making a decision and saying, "Yep, that's good. No, that's not." That's not the process that they're following. They are using um a a an authoritative source that's not them. It's not human judgment. They're actually going through and they're saying, "All right, I'm going to run it through the machine, and if the machine says yes, then yes, and if the machine says no, then no." Um you know, I think what's what's what's super interesting too about the deleted review case is that it's not uncommon for for agents to come back and say, "I can't find them." Right? Like that's And I And And to me, that's a support agent just being lazy. Because they they can find them. Um like they th- there is a there is a partial truth to that statement because when a like the the review in the in the data architecture is actually not owned by the profile. It's owned by the person who left the review. It's owned by the user. And so, when a review gets removed and and Joel was talking about this a second ago where like they can get removed from the profile, but they don't get removed from the user. This is kind of like, you know, he was talking before about how spammers look at their reviews that they left and they still see them, but they're just not associated to the business anymore. That's because that connection between um the review that is associated and the review that's published on the listing, that connection has been severed and it's no longer there. So, so when a support agent has the listing information and it's like, I can't find the the reviews that you're you're claiming or been removed, there is a partial truth to that. Um and and the burden is really on the on the business to say to provide copies of like these are the reviews that went that went missing, you know, and so like I've I've had a lot of clients that will you know, go back through their email box and it's like, I'm sure I got an email notification when I got this one. Let me try and track that down and and provide that as evidence to the support agent who then can go in and find the association with the user and ideally use that as use the representative example to collect all the other ones that are missing and and recover it. But it is actually a bit more work for the support agent. So, when I said that it's the support agent being lazy, they're just like, that's a lot of work. I don't know if I want to deal with that. I'm just going to tell them that I can't find and you know, like and and and they you know, a percentage of the audience will probably just go away. Right? But but but for those who understand the process like I hate that it's not like, oh no, you're not going to pull that one with me. Right? Like it I'm like, I know you I know they're there. I know you can get them. Right? I think that where I've had the most amount of success is when you have a backup of the reviews. And so, you know, like there there are a number software providers out there that that kind of like do, you know, GBP listing management. One of the services that is offered is is is it backing up a copy of the of the reviews so that when they do actually get deleted, you can run a report and and and see which reviews have been deleted. Um and then provide if you can extract that to a to a Excel document or something or Google Sheet, provide that as evidence to the to the support agent. They they can't pull the I can't find them card, right? Like because you just provided the evidence. You're bringing receipts to the table. It's like, "Hey, here are the reviews that have been removed. Can you check those out?" And so, it it does actually eliminate that um that that rebuttal from the support agent, you know, to to to to kind of eliminate that that blocker that will allow them to help you more easily.
So, something you you told me, Brad, um in one of your presentations at Local U was how there's like different levels of support. So, you have like level one, two, and three. Um I'd love to know some use cases when you want to ask for a different level because I've definitely succeeded in getting a different level of support. I don't believe that the second level gave me the result I wanted though. So, it didn't actually result in um which actually was a review case. It was a review case where this business lost like 200 reviews because they were working with a sketchy SEO company who set up several fake listings for them. And I believe the reviews were on one of the many suspended profiles that they had in the system. The agent that I talked to actually told me there were six. And but she would not move them to the right listing and I was like, "Listen, like I get it. Don't don't like reinstate the listing. They're not supposed to have six. It's a plumber. They're supposed to have one." But they they would not move the reviews and I was trying to argue that the reviews in and of themselves were not spam. They were real customers. It was just true. Like, erm, believe that is true. I got nowhere. So, I would love to know like what use cases, is it worth your time to fight with support and ask for another level? I'd love to hear some like Yeah, I agree with you.
That great great question, you know, and yes, there are multiple levels. And the way that we had it structured was it's the level is location specific. And what what I mean by that is I think I probably had about 11 different call center locations strategically placed around the world that that that makes up GBP support. Right? And so, you know, like there there was, you know, a couple of different locations in India. We had a really big call center that was very effective in Athens, Greece. You know, and then we had a lot we we had a small one in the US and then one in São Paulo, Brazil and one in Argentina, in Buenos Aires. Many others as well, but just to give you an idea of strategically where they were. So, so so the relevance here is that um level one and level two agents were all hosted together. So, like we had English level one and level two out of India. A lot of the the European languages were out of Athens. We we had a location in Krakow, Poland. I think that's now gone. But like the Spanish speaking was was operating out of Argentina. Portuguese was operating out of out of Brazil. And um and and all of those all of those support levels had both level one and level two agents on the floor in those locations. So, like, you know, and and the the communication method between level one and level two was usually real-time chat. They would chat consult the the case over. So, they could get information real-time. So, the support agent would would send that over to the agent to the level two, to the supervisor if you will, and they would like um you know, provide feedback on the case and and direction. Um and then it would um and then the level one agent would obviously um trans translate that and and and and give you updates. Um if it ever went to level three, that's usually um uh an indication of a defect, right? Like a a bug of some description that would go up to um a small set of agents that were in the US. Um and they were really responsible for logging bugs and defects with the engineering team and then following up with them to try and actually get them resolved. Now, if you ever had that experience, they take forever, or at least what it feels like, right? Because like um if it goes up to an engineer, um there's a whole prioritization exercise and it just takes a long time. So, like there's there's kind of like that that three-level structure where you got level one and level two um um strategically located together. And so, to answer your question, yeah, I do think that it it is valuable sometimes asking for a level two agent, right? Because um the level one agents are usually the least experienced, right? Like and um there there is at least there was um historically a decent amount of turnover in that staff. So, like constantly learning and learning and learning and then if they do good, then they get promoted to a level two. And if they don't, then they get rid of them and then they get new ones. Um so um so so asking for
A level two agent is, is, is, is not a terrible strategy. Um, at least it, you know, you may not change the answer that you're getting, but at least it will guarantee that you're getting, uh, a higher, more experienced, in, in that person that's making the decision.
The other pro tip I'll give you though, which, um, uh, which is kind of an interesting fun fact for you, like an exclusive for your, for your listeners, um, is that, um, when I was always looking at the performance of each of these call centers, they're not all equal, right? Like, um, you know, the, the, the India-based English-speaking agents, um, typically had the worst quality scores. Um, and so, like, if you wanted better quality, um, some people would strategically call up in Spanish and get the, um, and get the Argentina support who had a much higher quality score. Um, so, if you happen to speak Spanish, hit up the Spanish-speaking, uh, like, go, go through the support queues in Spanish, you're going to get redirected to the Buenos Aires location, which has a higher quality of agent, and you got a higher probability that they're going to help you. So, uh, so, that's a, a little pro tip on, on tweaking kind of, like, you know, the, the language modifier. I >> I kind of wish I spoke Spanish, but I don't. Um, Joel, I like to >> Google Translate it, you know? >> How do you get someone competent with Google support, Joel? >> Um >> [laughter] >> I, I think I, I think one, what Brad said is that's new for what it's worth since my time, because we had when I started it, we had one call center in the US, and they were all very good. I don't know if it, we actually had like 98% like CSAT. I mean, it was like we were super ecstatic when we first launched that, cuz they were all trained really well. And part of the reason why I resisted for a long time, and it was hard, I, I saw the team going to India to lower costs, is that there is context that they lack. Um, right? There's just something natural about understanding number, street, city, state, zip, right? That's something that typically we don't train, because you go someplace else, it is, uh, different, right? You sometimes start with a zip code or start with a num, start with the name of the street or that. So, those contexts are hard to, are hard to train because it's, it's cultural. It's, it's not just information. I think what, um, How do you, how do you get someone that's trained? Uh, that I don't know anymore, but I, I like the idea of speaking Spanish. I mean, I love that. I, I think what's good to remember is that things that reach, that if I hear it correctly, things that reach that level three are things that are bugs. And having a missing review, having a suspended profile, having something wrong in data is not a bug. It's just how, it's just an error in data. Right? And so, so we, we tend to overemphasize the idea, "Oh, Google must have a bug." Bug is really like someone can't access driving directions because they're using a certain kind of phone. That's the level of bugs that those team members want to hear and want to prioritize. It, if it is a data issue, you're up the creek without a paddle, right? If you're trying to get that extra support, I think you need to speak a language that helps someone understand that one, you understand what you're talking about. Two, I remember saying, "I know you can look on Business Builder and see this information. Can you please look there and tell me what you see because this is going to help me understand, right?" So, uh, understanding what you're doing and understanding how things work is a key way to help people understand, "Oh, this person not only do they understand, but they also maybe understand a little bit more than I do." And I, I think I, I think you've seen examples where, uh, customer service agents have said, "For better SEO, do these things, right?" Which >> Oh my god. >> they should never do because that, they, they're trained not to. So, that's it. That is an example of, you know, that agent is new and green and doesn't quite know all the process, right? And they're, they're actually trying to be helpful, right? These are humans. They're not, they're not trying to steer you astray. They're trying to be helpful, but and at the same time, this is another example of Google is not one Google. It's a collection of teams doing things. So, when you have like an agent that is trying to help people get verified and knows kind of that process, they may not understand what's going on in reviews as well as you understand what's going on in reviews. They may not know how to troubleshoot that very well. And because most of the review stuff is trust and safety. In fact, I think I was, I was still at Google, but I was a couple years removed from the team and I had to work with trust and safety. And these, and the person was relatively new to Google, maybe at as new as I had been outside the team. And they're like, "Oh, you, we don't give access to those tools." I, and I think it was to people like me. And it was a little bit >> surprising cuz I was like, "I can't build those tools." They, you know, it was one of those things where there was like this moment in, in my history at Google where I recognized that these groups are growing and, and they do need their processes. It's, it's fair to say like we access control this stuff, but it, it was, it had grown to organizational purposes where the organizational purpose of trust and safety is very different than the organizational purpose, uh, customer and consumer support. And so, those, those groups have separate processes. I think Brad was kind of explaining that. And, and to understand that you may need to get that person to trigger a process that goes to a different team, right? And, and that's hard, but, but recognize like support is isn't purpose-built to undo spam. It is purpose-built to to get people verified and help them understand how to use the system as is. And you can correct me if it's changed, Brad, but I, I, I think understanding that is, is important.
>> Yeah, I think I think you're right. In fact, and, and you can, you can blame me for this one. I'm sorry. But like there, there was a period of time where, um, the agents, the support agents, were incentivized on CSAT, right? They, so, so every at the end of every support interaction, a customer satisfaction survey would be sent out to, to the person and they'll be like, "Hey, can you rate me on my performance supporting you in this situation, right?" And they would be incentivized on CSAT. The higher the customer satisfaction score that that agent got, they would be rewarded, you know, often monetarily, right? Like they'd get bonuses and things if they got that. Now, that was a terrible idea, it, what turns out, because they just said yes to everything, right? Cuz it's like there was no like their, their incentive wasn't to make the right decision. It was the, the decision was anything that's going to get me a high CSAT is the score that I want, right? And so, you know, we ended up like spammers really went to town on that, you know, it's like, "Oh, all I have to do is is call up, be grumpy, and, and then the, the agent's going to say, 'Oh, no worries. I'll, I'll reinstate this, this fake profile for you. No problem, right?'" Because then they get a positive CSAT on the other side and they get incentivized by it. So, I had to change that. I had to change the incentive model. And so, the incentive was about, um, the quality of the like, the quality of the support interaction was, did you make the right decision? Like, it, in, in many cases in the case of a suspension or, or, or a review, like sometimes the right decision is no. Tell them no. Right? Like, and we would then incentivize them on making the correct decision. Now, that obviously is complicated in and of itself because now we needed a quality control process to, to audit everyone's decision-making and make sure that, that they are making the same decisions that, that we would make in that scenario. So, that increased the complexity dramatically. But, it's, um, it, it really is a good reflection, I think, of, you know, like, the, the goal is to make the right decision in the moment. You know, I think that no, no reasonable person will disagree with me when I say we want to keep bad guys out and we want to keep the good guys in, right? It's just there's so many layers of gray in that statement, right? Like, and it's like, and, and I think this is where I don't envy the trust and safety organization because they're constantly working on that pendulum. And, uh, and, and, you know, sometimes as I said earlier in the, in the session, sometimes the complaint is that you can't trust reviews cuz they're all fake. And then on the other side of the pendulum, it's all the reviews have been removed and, you know, so, so, you know, it's, it's a very difficult, difficult balancing act to, uh, uh, to, to balance.
>> So, I think the number one thing that people want from Google support is manual, right? Like you kind of nailed it when you said like if they're looking at your missing reviews and you expect them to be manually reading the reviews, that's not likely happening. So, are there workflows currently that are manual? Like I believe there is a manual verification, for example, where in that case somebody is literally looking at your thing, your very patient. Um, yeah, what, what is still manual that you can request?
>> Yeah, I think that, um, so, so you mentioned manual verification, right? I think that's a, that's a, an interesting one. Also, um, we, we called it ODVV internally. It's on-demand video verification. And so that is, um, um, you've probably experienced this. Like when you, when you attempt to verify via video, it doesn't work. Um, you contact support and it's like, "Hey, like I don't really know." And they'll be like, "You know what we'll do? We'll get you to contact an agent and they'll walk you through it, right?" They'll, you'll be a real human on the other end that does a, a video call and walks you through what they want to see, right? So that is still, um, human-based. Um, you know, but, and I think that, um, the manual verification process is another one which is, um, not, not very many people know that it exists, that process actually. And, and, you know, I can, I can give you a guide on how to access that actually, right? Cuz it's, yeah. The, the, the, the logic behind it is, um, um, Google doesn't want you to default to manual-based verification, right? Like they're, they're really trying to work through this, um, scaled approach, this algorithmic approach, right? Like, well, when I left 2 years ago, we processed 3 million contacts a year, right? Like, so just to, to kind of give you an idea of the scale that you're operating in, right? So it's like Google's not trying to solve the problem one at a time, right? They're trying to solve it in, in bulk. And so that's why it, the motivation is to try to push more to this algorithmic approach, right? Like, in, in, in logic. And so the, the manual verification process is really reserved for, I used to call it an escape hatch, right? Like, it's really reserved for the exception. Um, and so the way to access it, right? Um, unfortunately, the first thing you have, you have to attempt, at least attempt video verification, right? So, like assuming real-time verification channels aren't available to you, um, when I say real-time verification channels, I mean, um, verifying via phone or email, right? The, the options that sometimes are provided, um, as verification options. In the scenario that Google already knows something about your business, right? So, they no longer verify, have to verify the existence of the business, they really only have to verify your affiliation with that business, and that's where the, the real-time verification comes in. Now, if that isn't available to you, and, you know, it'll default down now to video-based verification because we got rid of postcard verification, um, you know, a few years back, and the video verification was ultimately the replacement of postcard-based verification. Um, and so, you attempt your video verification, and it fails for whatever reason. Now, if you've done that, what actually happens is in the back-end, a little bit is flipped in the database, and it's like, all right, they gave it a go, right? Like, they attempted video verification. And so, now, if you go through the support workflow, right? Like, and you go to what we refer to as the authorable workflow, which is, you know, if you go to the support channels, type in verification, you get to the bottom of the support flow, and then it kicks you out to another workflow, right? That second workflow is what we refer to as an authorable workflow. If you click through that, and, um, you'll get to a page where if you have never attempted video verification, um, you'll get to a point in that workflow where there's a little link that says, um, you know, like, send feedback or something like that. That is the indicator that you're not eligible for manual-based verification. But, if you've attempted video verification, and you go through that flow again, you'll go to the same page, but this time, instead of saying send feedback, it'll say contact us, right? There's a little big. And, and the reason why it says that, and it's, it's just, it's a little blue text, it's not a big button or anything like that, it's a little blue text. And the reason why it's said is because when you attempt video verification and fail it, then, then a, a bit is set in the database, right? Like, and it's triggered. It's like, all right, they tried, they gave it a go, didn't work out for them. Let's now enable the option for them to do manual-based verification. So, if you get that contact us button, you click on that, it'll take you to a dedicated form. Um, unfortunately, we can't just copy the URL and email the link around because you have to, it, it does all of the, the back check and it makes sure that it's, uh, it is actually eligible. But inside that form, you now have the ability to upload evidence, right? Offline. All right, so, so you can collect your evidence, you know, your documentation, your photos, whatever it is, upload that into the system, and then you can submit that. Now, that does go to a human. Going back to your original question, at least it did 2 years ago, right? So, and I, and I think it does because Google is safeguarding that. You can tell me describing how to access that is complicated, right? It's, it's hard to figure out how to get it. Um, and so, they're safeguarding it because it is one of the few workflows that I think is still manned by humans. Um, and so, that human will then get the evidence, review the evidence, and make a decision, right? Like, based on the, on, on the evidence. Very similar to that of the, um, on-demand video verification where, where the, the human will walk you through that process. So, there, there is a finite amount of manual processes that, that are left, Joey, and I think that, you know, a lot of, um, a lot of my old employees, like they, the, their job was to try and eliminate manual process, right? Like, and, and try and figure out ways to kind of like maintain the quality was always a requirement, but to automate it and scale it. So, so that is the motivation of a lot of employees at Google, my former employees would make careers on the fact that they could eliminate the problem from occurring or automate the solution. Um, so, that's, that's a big part of what they're focused on, and, you know, knowing that I've been out of the, out of that, um, that, that role for 2 years, I doubt the philosophy has changed. I think that everyone's still trying to push all that automation philosophy. So, not surprising that the amount of workflows that are being human managed by humans, they're getting fewer and fewer and they're just going to diminish, you know, potentially completely. That would be their goal, at least.
>> So, I hope you've enjoyed our conversation up to this point. Now that you've seen how the Google system works, in the [music] next episode, we're going to talk about what you should do when it turns against you. We're talking negative reviews, suspensions, how you should deal with them and fix them fast.