Episode 50b: Google guest post crackdown and dealing with fake PPC traffic

Play this episode:

Or get it on:

• Apple iTunes
• Anchor
• Breaker
• Google Podcasts
• Overcast
• Pocket Casts
• RadioPublic
• Spotify
• Sticher

What's in this episode?

In this episode, Mark Williams-Cook and Rob Lewis will be covering the Google guest post crackdown and how multiple publishers have reported receiving penalties related to guest posts. They will also be discussing fake PPC traffic and they will give a run through how to detect fake PPC traffic and a discussion of Rob's tools he uses to combat these clicks.

Transcription

MC: Welcome to episode 50b of the Search with Candour podcast! Recorded on Friday the 28th of February 2020, my name is Mark Williams-Cook and I'm again joined by PPC expert Mr Rob Lewis.

RL: Hello.

MC: And as you can probably tell it's the second time we have had a B episode; which means that we recorded 30 minutes of brilliant podcast a moment ago and our beautiful open-source audacity software failed to save and so crashed. So here we are again Rob, episode 50, obviously our lucky episode and today we're going to be talking about Google cracking down on guest posts and how Rob is dealing with fake, dubious and crap traffic with PPC.

So I was saying in the first recording, the first time around, that 50 does feel like a bit of a milestone for us. I remember, I think it was episode 5 that I was really happy when we did that recording because most podcasts,like 70%, 80% of podcasts do four or five episodes and then they kind of trail off and it's been pretty much a year, but it doesn't feel like it.

RL: No it doesn't.

MC: Yes it's gone really quick. I think episode hundred will be the big milestone for me at least so we're at least a year away from that in the future, but I hope things keep going as they are. I released some stats actually the other day for the podcast. We were really happy, we surpassed I think it was 5,000 listens on the episodes which is really cool, so it's averaging over a hundred listens per episode which I'm really really happy about.

So the first thing we want to talk about is Google guest posting crackdown. So this is an article that I actually haven't seen talked about that much in the SEO community, I spotted it on search engine journal, they did some nice coverage of this and it was essentially around that there's been a like a swathe of publishers reporting they've been getting manual actions from Google to do with guest posting. So this particular publisher on search engine journal reported - it's possibly paraphrased - the manual action email that said: “we have detected that some of your articles are guest posts, we have disabled your authority for your outbound links. please set your outbound links to nofollow and submit a review request” and apparently this has now been a whole host of publishers that have experienced this.

I think it's worth going back and checking just specifically what Google says in their webmaster guidelines about guest posting. I'll put a link to this in the show notes at search.withcandour.co.uk but if you just Google “Google Webmaster guidelines”, you'll get up their page and one of the topics they have on there is link schemes. So if you click on link schemes, this will go through the whole list of things Google doesn't want you to do, in terms of links, so it's got all the obvious stuff on there like, don't buy and sell links, don't do excessive link exchanges. But what they specifically say around guest posting is quite interesting; so they say, they don't want you to participate in large-scale article marketing or guest posting campaigns with keyword-rich anchor text links. So the keyword-rich anchor text links, that's something that really, I think, got that final nail in the coffin with the original penguin update from Google. So the penguin update was really targeting what Google was deeming as link spam and what SEOs were finding prior to this is we know that search engines use anchor text, which is the text within a link to determine what the following page of content is about.

Therefore when people used to link build, they used to focus on getting what I call the “money keywords” and what Google's referring to is this keyword rich anchor text. Whereas an internet user naturally might link to a website just by pasting the URL or they might just use the brand name, these links had a high frequency such as “buy car insurance” or whatever the search term was they were targeting and it really did work. And it was actually how, what we call the original Google bombing, was working. We had a quick discussion about this in our version A of this episode.

So the first example; I think I saw their main stream was when a bunch of SEOs managed to get the George Bush biography page to rank number one in Google for “miserable failure”. I don't think you'd seen this happen when it came out.

RL: No I didn’t.

MC: Yeah so I think this was achieved through lots of websites linking to this page with the anchor text “miserable failure” and we saw this actually happen a few times when, even outside the SEO community, people got wind of ‘oh this is how this works’ they would do

that because it would override, in some way, some of the other weaker on-page signals and just force that page to rank. So that's a really common thing that used to happen. Obviously this is really easy for Google to spot and it was one of the things that we saw the penguin algorithm update really hammer down on. Which is, if you've got these ridiculous link profiles where even 30, 40, 50, or even more ridiculous like 70, 80, 90 per cent of the incoming links you've got these key phrases and they're going to deep internal pages, that's just not how people are naturally linking on the web. so it really sticks out for Google.

The search engine journal article; going back to that, it talks about whether Google is issuing these manual actions for guest post articles or specifically for guest post articles that are paid or guest post articles you know there are over optimized. The warning email received from Google states it was a manual action specifically about paid guest articles, however the publisher and this is where it got really interesting for me the publisher reported that their site doesn't mention they accept guest posts. So the first thing we can draw from that is Google has figured that out right. What Google has apparently also done is of the various guest posts on the site - so the publishers had only done 15 guest posts in the last quarter, so it's like the last four months, there are not many guest posts, but Google managed to work out which sites were the ones benefiting from the guest posts, despite there being multiple links within those guest posts. So what the tactic some SEOs use is when they're supplying these guest posts they will link to the website, the client, whoever that they want to improve the ranking for but they'll also include a bunch of links to other websites in an attempt to camouflage who is the beneficiary of this link. But it does appear from the feedback Google has given that Google have worked that out and I think that's really interesting. So to me, without doing like a massive investigation on this, what it would suggest is that Google is using this cross-sectional data from other sites to work that out. So obviously there has been some kind of pattern that this particular site is appearing, maybe on several dozens of lower quality sites mixed in with seemingly random other links and Google has actually worked that out. I think this is interesting; if you are an individual freelancer, an agency, whoever, relying or using guest posting as something in your arsenal.

So for those maybe haven't done guest posting before, in a nutshell it is basically - you've probably had it if you own a website, where you've received an email and some very kind person is offering some highly relevant content, that is gonna be very interesting to your visitors about some vague connection to the topic your site's about and the deal is normally they will provide you with content which ideally your site can rank for, but their benefit is they're getting some links for that. This is what Google doesn't want you doing on an industrial-scale. Obviously occasionally guest posting on other good sites is fine, but this warning I think does start to draw the focus on maybe one of the few things I know lots of other agencies are doing, as a like day-to-day, month-to-month SEO tactic that hasn't been really stamped down on hard before. Search Engine Journal is reporting that they've had quite a few people say this has happened to them. I certainly know personally of some other agencies that have had these kind of notices as well, so it does look like there is a shift there, so definitely something to keep your eye on if you are doing guest posting,maybe if you don't have one it's time to start working on your plan B, Plan C and plan D.

Okay we're going to talk about some PPC now which is why you're here with us Rob, our resident PPC expert and you're going to talk to us about display traffic, traffic quality, fake traffic, dubious traffic, suspicious and just crap traffic.

RL: Yeah basically. Well this is all about display advertising this particular part and I'm a big fan of display advertising. I think it's a great channel. It's really fun to work on. It's a good way of reaching a lot of people in a short space of time and for raising your brand's awareness, it's a really good channel to work on for that sort of thing. But one of the things I've found over the last year is that the cost per click has just skyrocketed, to the point where sometimes it's just not even worth doing and I think this is mostly in part because Google is automatically opting advertisers, specifically newer advertisers, in to display advertising and a lot of them don't even know that they're running display adverts and in fact when we audit pay-per-click accounts from people who manage their accounts internally, one of the first things we check is whether they've been opted in to display and whether their search campaigns are running on the Display Network.

MC: I mean that's something we're covering in a lot of the PPC like fundamental courses we do isn't it? So we're trying to make people aware, maybe as the first step until they master it, to start on the search Network because I mean, I know we've discussed this before but there's a lot of those default options when you set up Google Ads accounts that it's sometimes better to dance around rather than just leaving all the boxes checked, right?

RL: Absolutely and the thing with display advertising is that it is upper funnel advertising, so there's always that expectation that conversion rates are going to be lower. And in the past we used to run on display advertising with the expectation that the conversion rate would be lower than the search, far lower than search, but with cost per clicks that used to average between 1 and 3p or a few cents per clear, it really didn't matter - it was swings and roundabouts, it was a low conversion but with a low cost per click, you could typically get a good CPA, a good return on ad spend from it.

But the other month I was looking at a display account and noticed that there was a placement on the Display Network that generated a click and it was a cost per click of 30 pounds and it was a really crap website, you know just to be blunt about it.

MC: That’s wild. That is expensive.

RL: Yeah, it is and I mean the average cost per click on that campaign wasn't 30 pounds or 10 pounds, but it was around the £1.50 mark and a lot of the clicks were from sites like that and that's something that I see a lot, not necessarily the 30-pound cost per click, but just dubious traffic from dubious sources and one of the types of campaigns that I see this most frequently in are lead generation campaigns that have running display advertising and I see a lot of fake leads being generated and fake goals and I think this tends to happen because a lead is being generated, but when you examine the lead, when you look at the actual content of the inquiry, it's either gobbledygook - just complete random arrangement of characters in the actual message or it's just spam. and so I see that a lot with lead generation campaigns and in fact if you're running a lead generation or your pay-per-click managers reporting to you that your display campaigns are generating the highest percentage of leads, then I would just make sure that you or your pay-per-click manager reconciles the actual lead itself that you received with the source of that inquiry and if you're finding that you're getting a lot of spam, then the biggest source is likely going to be from the Display Network.

MC: I mean that's really interesting so obviously the Display Network clicks are coming from Google's inventory of advertisers, so generally this is going to be made up from websites who have signed up for and opted into the Google Adsense program and that they're running ads for Google and I know there used to be a lot of ad spam so we called them MFA's, which were made for Adsense websites where people would spin out low-quality content, the idea being just to pick up some search traffic, jam them full of ads and actually make the site unhelpful almost, so people would click on the ads. Google did combat this, they had smart pricing which was trying to look at the value of the clicks that web sites were generating so I think the theory was, if they worked out that your site generates junk traffic for advertisers, what we're gonna do is actually pay you less per click and reduce that cost per advertiser, so it's interesting when you say that you see this fake or dubious traffic generating form fills, goals that kind of thing, because it does make me think maybe this is part of this operation some people are running - where they're generating fake clicks and they know if they send some traffic from their site through Google Adsense and it then converts, Google may think well actually this is good traffic I can then, therefore, charge the advertiser more and therefore that site gets more money per click.

RL: It's a really interesting point actually and I mean it's not just the lead volume that is a good indicator of a high-quality web site, it's also engagement metrics as well which I'll touch on in a little bit because it's relevant. But one of the things I've noticed is that these fake leads that do happen unfortunately on the displayed network, I've seen it happen more frequently since Google introduced the pay for conversions model, where you only pay if a conversion is being generated and I think the first time I was able to run that type of campaign I did, in fact, run it on a website that didn't have a capture form in place and the next morning, there were hundreds of conversions in this campaign and thankfully I managed to get a refund from Google on that case because the cost was high.

MC: But it obviously shows that there is some intentional activity, this isn't like some stray bots accidentally doing it, this is intentional activity perhaps

RL: Absolutely.

MC: And alongside this, I think you mentioned it's well in a previous episode, a while back Google removed the ability, and correct me if it is wrong, to exclude mobile apps from advertising, is that right?

RL: Yes. You can exclude mobile users but you can't excrete mobile apps, so in those cases you want to target mobile, people on their mobile, they account for the vast majority of traffic but you don't necessarily want to run your adverts on a mobile app because well the the chance of crap traffic is a lot higher for a mobile app - dubious traffic, accidental clicks.

MC: Yeah, accidental clicks must be high because you know, I’ve accidentally clicked on ads because you know, desktop and mouse - you've got precision there, my bumbly thumbs on an app you know I've clicked on ads before and I'm not exactly what you would call a technophobe or a new user to these things, so if I'm doing it, I assume there must be lots of other people doing it.

RL: Absolutely and you know Google does have an invalid clicks system, where it looks at clicks that it thinks are accidental but it's not bulletproof and I've proven that it's not by looking at analytics accounts and by looking at clicks that matched with obviously fake traffic, so they don't always know when it's a flaw.

MC: Have you seen any trends in types of apps or anything like that caused that sort of traffic?

RL: Yes absolutely and I'm sure people listening here will smile when they hear me mention these two words, if you manage display accounts that is. Torch flashlight apps probably one of the biggest accidental click type apps, which makes sense. Who uses an app in the pitch-black dark?

MC: Unless it’s like a buy bulbs app.

RL: There’s that, yeah.

MC: Emergency bulb.

RL: If you do sell bulbs, get in touch, we can help.

The other type of placement that I always see is VPN's, so virtual private networking placements, they're the first things I always have to exclude when I'm taking over someone's account.

MC: Any idea why? I mean I can see like torch apps, ad placement where the on/off is. Is this to do with things like what free VPNs, that maybe have ads close to the connect button or?

RL: I think another way of looking at it is the actual person knows what they're doing and they're influencing clicks perhaps and using VPNs maybe, I don't know - that’s one possible reason and the other, as you say…

MC: The placement.

RL: Yeah.

MC: That's interesting.

RL: So basically this is really rife on the Display Network and it's a problem I face every day, on every display account that I'm running I'm sure other advertisers know what it's like when you need to run a mobile campaign and you're constantly excluding poor quality mobile apps or just poor quality placements, that you don't want your ads to run on, that's been going on for years to be fair.

But one of the things I've noticed aside from that is that there are three main types of fake traffic that I consistently see and which I've managed to narrow down through segmentation in Google Analytics. So I can easily just run a report and see the prevalence of fake traffic essentially and this is just part of my day-to-day optimisation, when I'm managing a display campaign.

The first type of fake traffic that I see are users who have only ever visited through display campaigns, but yet have visited the website multiple times through display. So they may visit your website ten, twenty, sometimes in excess of fifty times, just through display advertising and through no other channel.

MC: I mean, we've said this doesn't even make sense in terms of if you try and think of how a user might do that. So if they've seen a display out on a website, the odds of them thinking ‘oh what was that website? Well I remember I saw the display out on this site’, if they go back to that site - because that site is just showing normally the rotation of Google inventory the odds of you seeing that ad again are quite low. So again that points me towards this is intentional activity, possibly by the site owner to generate this traffic.

RL: Absolutely. it's not even normal human shopping or browsing behavior either because normally what you see from analytics data is if someone has high intent to purchase or generate a lead and they've discovered you from display advertising, sure they may not necessarily just visit you at one time, they may visit you multiple times but they will visit you through multiple channels. the initial channel will likely be display if that's your acquisition channel and then they will revisit you perhaps later on in the day or a few days later to actually generate that lead but they may do that directly, organically or through a pay-per-click search advert, so it will very very rarely will it ever just be through display. I mean there are some situations but the data that I have shows it's very very low.

So that's the first type of dubious traffic that I segment into and the second type comprises of what I call ‘fake high engaged users’ - so these are users who arrive on your website through display advertising and they only ever visit one page on your website, so they don't browse to any other page, yet they exhibit a high average session duration and low bounce rate or 0% bounce rate, which is obviously impossible. And the only way that that can happen is if someone were to visit your website, sit on the page for about two or three minutes and then refresh or somehow reload the page again.

MC: Or I guess, it's unlikely, but if you had some event, you could trigger an analytics but they're not going to know that through bot traffic so like refresh is getting the extra hit in that session.

RL: Absolutely, yeah. So when you're running a new display advertising campaign or when you're just optimising for any reason, for engagement, the pay-per-click manager is going to divert traffic to the placements that have the highest engagement, so that is why you get that type of fake traffic because it will stop you from excluding it, regardless of whether it's generated a lead or not.

The third type of common fake traffic, I've already mentioned, is traffic through display that generates fake leads and you can identify these quite easily because you'll have, say a placement generate one user visit the site and they've generated five or six leads or they visited six times and generated six leads and it's the same user, which again it could happen, it’s unlikely it could happen but it doesn't happen very often and it's not necessarily a fake traffic instantly, but it's definitely a red flag and it needs to be looked into and you need to have a look and see what types of websites are driving traffic to sites that are doing that because the likelihood is, it's crap traffic.

MC: So as I understand you've put together your own tool that collates this information and highlights it for you.

RL: I have yes and it's been really useful and I've been running it on all of my active clients and it's it really saved me a lot of time and that's why I put it together, because I spent so much time filtering out poor quality traffic, identifying where the biggest offenders are in terms of low quality traffic and which ones are potentially the highest quality, that I thought I need to streamline this. I need to speed up time and spend more time actually optimising the account, rather than analysing what’s crap. So that's why I came out the tool and essentially in a nutshell, it works by using a weighted scoring system, it analyses all of the traffic that's been directed to your website through display activity, it looks at all of the placements and it assesses the prevalence of the three different fake traffic types I've mentioned. Obviously the higher percentage of fake traffic coming from one website - there’s something wrong with that placement; you need to look at it or exclude it. It looks at the volume of traffic that the placements received as well and the algorithm factors that in, which is really important because a placement may have generated one or two visits, generated one fake traffic that's a really high percentage of fake traffic but actually not enough data to go on.

MC: Yeah.

RL: So it factors that and the more traffic a placement generates and the more percentage of high fake traffic it generates, the more I guess “umph” the algorithm gives to highlighting that placement as a potential fake driver of traffic. also if it's a website that generates both leads and online sales, it factors in the fact of whether a sale has been generated as well, because you can't really fake an online transaction, can you?

MC: No, if you’ve got the money, you've got the money.

RL: Exactly, yeah and you know if someone wants to spend £5 with their own cash trying to add more fakes than go ahead, keep spending £5, you know it’s money in the pocket.

MC: I guess it's an interesting arbitrage thing there if you did fake some transactions, does that raise your CPC as a MFA advertiser enough. But that’s all good for you if you’re generating sales.

RL: Yeah, I've not worked on any e-commerce sites where that's a problem, but I mean obviously if you're running a search tool and you know, you need to factor that in then obviously factor it in. But essentially if there's a lot of fakes leads, a lot of fake traffic being generated by a specific placement, yet an online sale has been generated at a good return on investment, then the algorithm will factor that in and say, so what? you know I don't care if I'm getting odd fake leads or fake traffic, if an actual valuable sale has been generated. Although to be clear, that doesn't happen very often in my experience, fake traffic is generally fake traffic.

MC: Fake traffic is symptomatic.

RL: Yes, of course. But you know we need to fact this in, just in case. It then looks at the overall percentage of fake traffic and obviously there’s more to just fake traffic there's other things as well like, low engagement that it factors in and lots of other things that I could just prattle on about for ages, but I don't want to bore everyone. So to cut a long story short, among other things, it tells you which place is driving the highest quality traffic, so it scores them really highly if they're generating actual leads and not fake leads, it looks at whether it's a potential fake lead or a genuine need and whether it’s generating online sales and if it's not doing that, if it's generating a high percentage of fake traffic then it gives you an urgency metric to say you need to look at this or exclude it because it's likely just wasting money. And I've recently added a new add-on to the tool that factors in conversion delay; so if someone clicked on the display ad for 3 months ago and then converted three months later into a sale, it does it looks at that as well and factors that into the quality metric, if you will, just to make sure that we're not just looking at a last click basis.

So it's becoming really useful to me and it's helped, I guess just to clarify a lot of ideas of have had about fake traffic on Display Network because a lot of the theories I've had seem to be proven through the tool, when you see the constant patterns and so to me it's all about saving my clients money, I don't want them investing money in crap traffic and essentially this is what the tool does.

MC: That must be one really biggest “if” statement you’ve got there.

RL: It’s the biggest “if” statements I've ever written; ifs within ifs within ifs, it's great. But one of the other things as well as with all tools, there's a nice basic dashboard to it as well. So in real time, I can just log in in the morning, have a sip of my coffee and say, oh look 80% of all fake traffic comes from mobile apps, take another sip of my coffee, and then decide what to do about that. So it's just a really nice time-saving tool for me, it's been great.

MC: You've also been working on like a location-based tool.

RL: I have yes, and the reason I've been working on this one is because a lot of my b2b lead generation clients, obviously service different locations and it's really useful to know which locations perform best and you can actually look at that data anyway in Google Ads, they have a location report, actually have two location reports. The first is the user location report, which reports on where the user actually is geographically when they carried out the search and the second one they have is more of an interest location report. Now the problem with that report is, it doesn't really tell me what I want to know; which user entered a location into their search query, was it town? Was it a city? They don't really tell you that, they just say whether there was some intent on that location at some point, regardless of whether it was for a search or they'd previously visited the location or not.

So this led me to assess one of my clients, I was working on their account and I spent a lot of time assessing it one and I thought there must be an easier way of doing this, so I just essentially set up a database full of every single town and city in the UK and ran a tool that compares whether or not a town or a city was mentioned compares the cost per conversion against towns and cities, versus whether or not there was a search for a location mentioned at all in the search query and then just gives me the results straight away. And it's really handy, especially if you have a client or if you're running an advertising campaign, where you've got a limited budget and you want to get more leads but you don't have any more budget and you can quickly see whether people who carry out a search for a location convert better or worse. So if they convert better with a location, you know you can invest more of your budget into getting more of those searches and excluding those who without a location.

It also helps me to build out campaigns based on location a lot quicker as well, so it's been really useful one and again I do like a dashboard, so I could just run it on a client if I'm curious and it will just tell me the cost per conversion or the return on investment of people in towns versus cities versus whether they typed in a search at all for a location and then I can just make some really quick wins and reduce cost per sale or increase return on investment or make recommendations to the client for their own internal reasons. it's yeah it's a really handy tool.

MC: So that makes a lot of sense, especially for local clients, because as you say Google has moved to this slightly more opaque intention around location-based.

RL: That's putting it lightly.

MC: Yes, that's what I do. So I mean, all of these tools you've built, they're pretty new right?

RL: Yes, yep. I actually was inspired to create my own tools actually after listening to Hannah Rampton do a talk.

MC: Yeah, if you haven't listened to that, do. So Hannah came down to Search Norwich in January and did a brilliant talk about using the query function in Google sheets and showed us some of the stuff she's done and I think if you work in SEO you've probably seen Hannah's Google Search Console like insights Explorer tool she made, it’s really really good. As you saw today as well, Google Search Console has actually improved some of the exports you can get from GSC, which is brilliant because I think they've seen lots of people are now making their own front-end and analysis tools in data studio and sheets for that data.

Brilliant, thank you Rob, that was really really interesting. I hope you've enjoyed this podcast, this was our unlucky episode 50, so episode 50b - I hope episode 100 treats us slightly better and we don't have to have a 100b . We will be back in one week's time, next Monday the 9th of March. As usual, if you are listening online you can subscribe to us on pretty much any podcast platform you can find and you can find the notes from this episode at search.withcandour.co.uk. I hope you have a great week.