Candour

Blog

Episode 11: Combating PPC click fraud and a new Google SERP

Mark Williams-Cook
Mark Williams-Cook , SEO Director
28.05.2019 • 28 min read • Podcast

Play this episode:

Or get it on:

What's in this episode?

What's in this episode? Mark Williams-Cook and Rob Lewis will be discussing:

A new Google SERP What changes are coming to how both organic and paid ads are shown in Google.

Click fraud Detecting and combating click fraud on the display network.

Podcast transcription:

Show note links:

New Google Search blog post: https://www.blog.google/products/search/new-design-google-search/

Google's blog post about enabling a safe digital advertising ecosystem: https://www.blog.google/products/ads/enabling-safe-digital-advertising-ecosystem/

Transcript

MC: Welcome to episode 11 of the Search with Candour podcast recorded on Friday the 24th of May. My name is Mark Williams-Cook and I'm joined by my favorite person in PPC, Mr. Rob Lewis!

RL: Hello!

MC: In this episode we're going to be talking about some changes in the actual Google SERPs, it's quite a big story this week in that Google is changing what their core search engine results page looks like in terms of ads and organic. A little bit of a longer, almost feature segment, we're going to be talking with Rob about Google Ads click frauds and examples and what you can do about it.

Before we get onto those subjects I just wanted to talk about something I actually saw a few hours before we recorded this podcast which was that the Google Webmasters Twitter account today this morning just tweeted saying:

So this tweet is actually referring to yesterday, Google confirmed as well yesterday that they had another issue with indexing pages that seem to be affecting a lot of sites (tweet seen below), and if you listened to our earlier episodes you'll know we covered issues previously with Google dropping a single percent figure of its entire index.

You know we covered some caching issues, you know we covered some issues with Rich Snippets that Google's been having as well as data loss in Search Console, so they're really having a bad month technically. I said at the time in those episodes and this is my personal feelings, I think some people agree thee I find it hard to believe that these issues aren't related especially as we've just seen the new deployment of their Googlebot service so this all seems to be happening around the time that they've switched their live web crawling Googlebot to this ‘evergreen’ configuration to keep it in line with Chrome. It may just be a coincidence but it seems like a big coincidence to me, a few other people have said that they've noticed in the last 30 days they seem to have had more problems than they've had in the last several years so I'll be interested to see if they can sort these out. That was just a kind of by-the-by so if you are having any indexing issues it may not be your fault - it may be Google's fault this time.

The first thing I wanted to talk about today with you is this new SERP design, there's a post on the Google blog, I'll put a link to it in the show notes on search.withcandour.co.uk, and it's titled ‘A new look for Google Search’, and I'll go through the post and just explain to you what these changes mean.

So they say:

So it was the podcast before last when we talked about the new Google feature where you're going to be able to view 3D models and augmented reality directly from search results, so that's what they're talking about there. They go on to say:

There's a screenshot below, but I'm just going to talk you through what these changes are.

Google says:

Then they give a helpful link where you can specify your preferred icon for the organic listings, so the way to summarise this is essentially: favicons are coming to Search so you can choose a tiny icon and it's going to go next to the name of your site in the top of the results card, which all sounds good and fine. What is interesting though, and they've put it as the secondary point in the blog post but it's the one I've seen the most conversation about. They're changing how ads appear in Search, so at the moment (and on their screenshots) they show when you run a Google Ads, you get the title at the top and then underneath you get the green ad label and your green URL under, that is the ads we are used to, and these are still really different. I'm sure Rob remembers too, many years ago Google Ads, apart from just having a tiny green box that said ‘ad’, before that it was kind of a full green box that said ‘ad’ and way back when the whole background was a different color.

So we've always seen this slow move towards blending in ads with organic and I think now from the mock-up they've given of the new search results, that the ads look even more blended in and there is essentially some just tiny black text at the top. It's not in a box, it's just two letters just as ‘ad’, instead of a favicon, instead of a small image (I guess they weren't technically be called favicons but I'm going to call them that for now) and then it gives the URL and then it's on to what looks like identical listing from then on to an organic listing. I can understand, certainly I mean why Google would want to do this because we all build up this kind of ‘ad blindness’, and this armour - we know what ads look like and a lot of people will skip them when they identify what they are, and we've seen many times on different platforms when you change what the adverts look like you at least temporarily get a spike in the click-through rate. So I would imagine this change will, at least in the short term, see extra revenue generation. I guess this is something probably they've done tests for, what do you think?

RL: It’s something they've probably tested, I'm sure they have I'm sure everything they roll out they've tested amongst a small segment of the population haven't they. If it had a lower click-through rate I can't imagine that they would have rolled it out.

MC: So good for Google. Arguably not as good for consumers in that I think it's helpful to be able to clearly determine which results are ads and which are not, and I guess that's it's good if you work in PPC maybe?

RL: Absolutely, anything that can help to drive more traffic and sales through that channel is great. It's interesting actually because I looked at the new proposed layout and I see actually that it is a lot easier to see the ad to determine that it's an ad because it doesn't have a favicon.

MC: So well the favicon is ‘Ad’, it says ‘Ad’.

RL: So in my opinion it's actually a lot clearer to see and differentiate between an ad and an organic listing than it was before, but that's just perhaps because I work on pay per click every day, maybe that's why. So in my mind it would have led to a reduction in click-through rates. But I'm guessing that's not going to be the case at all.

MC: It will be interesting to see when this does go live, because obviously we are running quite a few campaigns I guess we'll be able to see the historic results on the on the same ads if they're getting more or less clicks or higher or lower click-through rate. Google says “this redesign is coming first to mobile and will be rolling out over the next few days”, from that I can assume it means it's coming to desktop as well because it says it’s first coming to mobile.

Stay tuned for even more fresh ways Search can help you, the actionable things from this are that you'll be able to specify a brand image that goes next to your search results so that I think is going to be definitely something that's going to be worth doing. We assume that the click-through rate on paid ads will, likely at least in the short run, I’m confident in the short-term they’ll increase - in the long term it will be interesting to see how people adapt to that to that ad format.

Okay we're going to talk about fraud and we're going to talk about fraud specifically on the Google Display Network. So if you don't if know what that means it basically means that if you're running Google Ads. apart from the search results that you can run adverts in so that's when people do a search and they see you they see an ad. Google has a vast inventory of its own properties and third-party sites that you can show adverts on a display basis that they call the Google Display Network. We want to just do a short segment about combating fraud on the Display Network which Rob's got an example for.

So by ‘fraud on the Display Network’, this is a term we're using for non-genuine clicks, so this may be automated programs that are clicking on adverts to generate revenue for someone else. So they're not genuine people seeing your advert and clicking on it because they want to buy something, so potentially damaging if you're an advertiser and you get lots of these fake clicks. so something Rob deals with on a pretty much daily basis I guess and I think you've got a good example you want to run through us about click-fraud and what we can do about it.

RL: So recently I've been carrying out an investigation into a potential ad fraud case for one of my clients and during the process of the investigation I ended up uncovering some really interesting patterns regarding display traffic behaviour. I've always known that the Display Network is more susceptible to fraud but unfortunately recently I've come to the conclusion that far more spend is being generated by potentially fraudulent traffic than we may realise.

MC: When you say the Display Network is more susceptible for fraud I assume that's because we get, you know the internet’s full of bots clicking around on websites, and it's just more likely they're going to click on adverts that kind of exist statically on websites rather than bots aren't going out there as much so in doing searches on Google and triggering results.

RL: Yeah that’s right, because Search adverts appear when a relevant search query has been typed in and display adverts as you say are static, they're always there across a plethora of sites and properties.

MC: So bots basically aren't doing lots of searches everyday.

RL: Absolutely, first a bit of background I suppose to set the scene, last year back in November 2018. Google in conjunction with the FBI helped to investigate and eventually take down a multi-million dollar ad fraud scheme on the Google Display Network - codenamed 3ve. It's one of the biggest fraud schemes that have ever been uncovered so far, I'm guessing we can leave a link to that article in the podcast.

Millions and millions of fraudulent ad spend was generated by 3ve when it was running at its peak and I'm going to read a snippet of text just directly quoting Google when they came out publicly about the scheme, so this is what Google said:

Just to explain, website owners and publishers earn money for any traffic or impressions that they generate if they're part of the Google Display Network. So 3ve was a way of driving fraudulent revenue publishers which could be potentially a lot of money if you can get away with it.

MC: So when you said “produce more than ten thousand spoofed fraudulent domains generating three billion clicks” I'm assuming this means that through computers infected etc. these are fake sites they've set up as well. So they've set up, this isn't necessarily targeting websites that are running ads themselves, it's that the fraudsters set up fake websites to show your customers’ banners on.

RL: Absolutely. In fact I could give you an example of something that happens time and time again on every single display campaign I've ever managed. Which is, let's just say you're running a remarketing campaign so you don't really mind where the ads are showing as long as it's showing to the right people, you start seeing traffic appear on torch flashlight apps and you think those flashlight apps - I've looked at the data they've never converted because people have accidentally clicked on the ads so, you exclude the flashlight app and then the next day you have two more flashlight apps appear of a slightly different app name that's been submitted to the Google Play Store. So you're continually excluding these low-quality placements.

So I'd imagine that 3ve operated in a very similar way in which it continually automatically generated and submitted all sorts of dubious placements that were low-quality and ultimately not valuable at all for advertisers. I could talk and talk and talk about how good the Display Network is because it's such an excellent channel for raising brand awareness, it's brilliant for remarketing. I quite often see it used to a lot of success, sometimes it performs better than Search but any pay per click manager who's worked with a client who has invested many hundreds and thousands of pounds into Display activity will tell you that the potential for wasted spend is immense, particularly if it's not monitored closely - just as with any digital marketing activity I suppose. But I once had a client who was a lead generation business-to-business client, and they came to us to manage their Display Network spend, and when we bought them on they'd spent close to half a million pounds in one year just on Display Advertising traffic alone.

The reason they kept investing the money into Display is because their Google Ads conversion data seemed to suggest that the campaign was generating hundreds of leads per month for them, all from Display traffic. In this particular example it was live chat requests which were their most valuable source of leads at the time. To cut a long story short because I know we only have limited time on these podcasts, the leads from these Display campaigns turned out to be spam. Every single one of them from their Display traffic! Which is such a sad tale when you think about the amount of money they've spent on it, and as I investigated it in more detail the types of placements that were generating traffic and spend were mostly really dubious sites like Virtual Private Networks, flashlights, chats perhaps that no one has ever heard of before with really high click-through rates.

A pattern began to emerge as to the types of sites that were generating all of the spend and all of these strange live chat conversions. Now because the client hadn't been tracking their lead quality very closely or importing their after-sales information back into Google Ads, all they did was focus on these conversion figures from Google Ads that they were seeing every month.

MC: So when you say lead quality, normally in a Google Ads campaign it's really easy just to check when you're getting an inquiry or a phone call or a live chat request and by quality I assume you're defining this as ‘you've got a lead and then further down’. So if it's tracked in a CRM you're tracking whether it turns into something valuable as a sale or customer, right?

RL: Yeah absolutely, so for a lot of campaigns a lot of clients don't have that visibility they can only see that a lead was generated which is what was happening in this case. They weren't firing information back into Google Ads to say ‘this lead was qualified’, it was further qualified as a potential sale and then it turned into a sale. They didn't have that ability.

MC: So that’s probably something worth thinking about if you're listening to this podcast and you're trying to generate leads and you're using Paid media to do so that it is possible to hook up your CRM with Google Ads. You can track which ones turn into actual sales rather than just track quantities that should be on your checklist of things to look at.

RL: So for this particular client they had spent half a million pounds on conversions which it turned out didn't generate any actual leads at all, it was just spam when we investigated the actual leads from their customer relationship system. It was just filled with gobbledygook essentially and just to add, that the types of Display campaigns they were running were automated Smart Display campaigns which is a fully automated solution that Google offers wherein it automatically optimises the accounts to find more conversions at the lowest cost.

In this case it had focused entirely on conversions, just quantity of leads, it didn’t care about the quality of the leads. There was no human decision-making involved it was a fully automated solution.

MC: Assuming that would work better if they had set up the goal as a quality lead rather than just a lead right?

RL: The short answer to that is yes but not necessarily.

MC: Okay I'm going to stick with the yes.

RL: So I mean to me it's a very serious matter how the client’s spend had been able to get so high but they hadn't examined the lead data and the quality of those leads, and they hadn't investigated it so they or their agency should probably have been monitoring the quality of those leads and questioning it before they continued to invest another £250,000 into Display Advertising. They knew they had a spam problem they just didn't know where it was coming from. So these experiences with the Display Network - I have quite a few of these experiences unfortunately, they make me ensure that I'm always very cautious about spending money on behalf of clients on the Display Network.

Like I say it's a great channel and it can work but you have to plan it carefully and you have to assume that there's going to be a lot of low quality traffic.

So moving on from there I'd like to tell you about a new bidding method that Google introduced last year called ‘pay for conversions’.

MC: Just before that, so I mean I've used other Display Networks apart from the Google one and you know you're absolutely right, I take it as part of a course is going to be spam traffic but do you think on average I get the feeling that while there is definitely fake traffic on Google, it seems to be a bit better in terms of a slightly lower percentage than other networks I've used. Because other networks have had, you know 50+% fake traffic, which makes my first port of call always the Google Display Network and sometimes just the Google Display Network.

But then that means you put all of your spend in there so even if you've got 5/10/20% fake traffic that ends up like a big number doesn't it?

RL: Yeah absolutely it all adds up in the long term, you know how much money do you want to invest in fake traffic? Well none at the end of the day and I think any marketing activity you're carrying out, whether it's Display, whether its social media advertising, always make sure that your analytics setup is accurate. That the traffic that you're logging reflects the actual traffic that's arriving on your website, that way you can look at the actual volume of people who have visited your site and analyse their engagement and make sure that actually you are driving the traffic that your Display Network is reporting to you.

MC: So pay for conversions pay for conversions!

RL: Pay for conversions. It’s probably old news actually for a lot of people listening but last year Google introduced ‘pay for conversions’ as a bidding model which allows you to drive traffic for free on the Display Network, and you only pay when a conversion is generated. Essentially you set the amount that you want to pay for each conversion and Google does the rest, and it will honour that cost that you set.

When I heard about this I was so over the moon because it's a no-risk display test so we can come up with a display advertising campaign and it's completely risk-free for the client. They will only pay when conversions are generated and they will only pay the amount that they want to pay, or so I thought. So I decided to try this out for a lead generation client of mine in the tourism industry and they already had a really successful Search campaign which comprised the vast majority of their leads, but I wanted to find a new source of leads for them and this was just the perfect opportunity to use a Display Network in a risk-free manner.

So the client and I agreed an initial spend and we set a cost per conversion and I put together a very tightly focused Display Campaign that targeted the core audience and I used Google's custom intent audience approach, which we've discussed before, in order to do this. 24 hours later we were very excited when we looked in the campaign and we noticed overnight they had generated many conversions at the agreed amount of spend which was about £1,000 for the initial test and the cost per conversion target we set was £50, so we generated 20 leads from that - cost per conversion of £50, which was on target.

You know the results when we looked at them looked fantastic.

MC: Good news right?

RL: No it wasn't Mark. We decided to look at the actual quality of the leads before we carried on spending any more and every single one of those 20 leads from the display campaign was spam. I looked at the users in Google Analytics and the behavior of all of these users was really questionable. They just seemed to arrive on the website from the Display ads and then a few moments later they tended to just head straight to the ‘Thank You’ page which is what we were using…

MC: As you do!

RL: Yeah exactly! There was no intermediary step, they weren't even researching the product. They just arrived on the page, go to the ‘Thank You’ page, leave the site and in many cases they just kept interacting with the Display campaign over and over again. Now what was really annoying aside from that was that the spam that came through was just totally nonsensical it wasn't even selling SEO it wasn't selling anything it was just gobbledygook spam and the only reason I can think of the lead being generated was for users to carry out some ad fraud.

MC: So this means basically the motivation could only have been that someone's getting money for generating those conversions, those clicks.

RL: That's how I look at it, that's how I take it yes absolutely. So I'm not happy about this because I make use of the Display Campaign for many clients and herein we set up a campaign on a pay for conversions model and the client paid a £1,000 for conversions that weren't actually conversions. It was just fake traffic and fake leads so I decided to get in touch with Google. So first of all I collated all of the spam and enquiries which came through, all of which were derived from this one Display Campaign. I logged the GCL ID for each enquiry.

MC: So the GCL ID is the unique identifier for the click that came from Google Ads.

RL: That's right.

MC: From that ID you can tell which ad, which keyword etc. they clicked on.

RL: Yeah and I'm pretty certain Google can track back the ISP data of that user, where they're based in the world and all sorts of things. I made sure I took detailed screenshots of the spam traffic that I had uncovered in Google Analytics and I also included the ISP data of those visits as well, and note that nearly every one of those spam leads came from the same ISP service - a really dubious looking one whose name I shan't mention. So I had this information and I emailed Google's click-fraud investigation team and I assumed that given that the actual spam itself provided very strong evidence of ad fraud that they would welcome the intelligence refund, the cost to my client and block that particular ISP from doing it again.

Unfortunately that wasn't the case and I have had a very long email exchange with the ad fraud team, in which they told me that “Google's Network is 100% safe and it automatically removes such fraudulent ad costs”. The conversation is still ongoing but as of yet there has been no refund.

MC: There's very few things in life that are 100%.

RL: This is true.

MC: This reminds me of that car company that went on Twitter recently and said their car was a 100% unhackable and I think it was like 48 hours before it'd been compromised. You're just baiting people and you say that.

So I know you use Google Analytics to try and ascertain what traffic isn't useful, what is actually fake or fortunate for your clients and you do that via a segment if I'm right? I mean we have got refunds from the Google Ads team before where we've proactively identified fake traffic for clients so I think if you want to share how you do that, it might be quite useful.

RL: Sure absolutely. Well I mean obviously the first indicator that you're getting fake traffic are dubious leads. If you get all sorts of spam come through and it makes no sense, always have a look and find out where it's coming from because if it is coming from your pay per click activity, just remember that’s spam that you're paying to receive which you don't want. So following on the the pay for conversions incident I decided to take a look at the actual patterns involved, so I use a segment which helps me to identify potentially fraudulent traffic from Display Network and what I found was that with a lot of these potentially spammy visits that come through Display, they all seem to have the same type of behaviour. Which is that they generate multiple visits from Display, normally at least three visits, from the same user. So that user has clicked on a display ad for at least three times and they seem to interact with no other channel other than Display which in itself is highly suspicious.

When I drill down into these particular users with that segment an interesting pattern seems to emerge, which is that many of them have an unnaturally high average session duration, which is a metric pay per click managers look at as an indicator of traffic quality. So the higher the better, unfortunately in this case certainly it is not so, when I look at these people what happens is that many of these bots or these people who are clicking on a Display ad will just visit one page on the sites and then several minutes later refresh the page which removes the bounce and increases the session duration.

MC: So this makes it look on the surface at least like you're getting good traffic.

RL: Yep. So people were at the very top level in Google Analytics, they will look at their channels and go “well our site's average session duration is four minutes...”

MC: But the bounce rate is low.

RL: Yeah exactly. So that's something you have to bear in mind as well. The third observation was that the spammy traffic often head straight to the ‘Thank You’ page after arriving on the site without any intermediary steps and this is particularly so if there's no capture form in place, and the user doesn't have to fill in all of the details on our contact form. It may be that they're just literally scouring the website for a Thank You URL and heading straight there as well.

Unfortunately when I look at this segment I've found some accounts where up to 30% of users from the display campaign comprised people from within this segment. Now I'm not saying that every visitor from this segment is fraudulent, but it's a very high indicator and certainly I haven't seen so far any valuable leads generated from this particular segment when I run it.

As you said Mark you know Google does proactively remove fraudulent clicks so it might be that some of the traffic that you're looking at, if you decide to run such a segment has actually been refunded. To summarise what I've been talking about: you just have to be hyper vigilant, especially if you're a lead generation website, always make a point of importing your after-sales data back into Google Ads so you can see which campaigns have the lowest lead to sell rate and I'm afraid so most of the time it is Display that has the lowest lead to sell rates. So always question the quality of the conversions that you're getting from your display activity which is obviously a lot easier if you have a CRM system in place which can capture the source of the lead for you.

MC: Wow that was really useful! We've gone on a little bit longer than usual but hopefully you've got some value from it. Our next episode is going out of course next Monday on the 3rd of June. You can get all of the links, screenshots etc. of everything we talked about on the show notes at search.withcandour.co.uk. Of course please subscribe if you're enjoying it, or if you hate it and think it's rubbish give me some feedback and I hope you all have a great week!

More from the blog