Welcome to episode #77 of The PPC Show, where we interview the best and brightest in paid marketing. This week we're joined by Andy Taylor, Associate Director of Research at Merkle.
Brands may be at Google’s mercy when it comes to the price of branded traffic, but Andy outlines steps they can take to reduce their CPC pain.
Stay tuned to learn about:
- AdWords brand CPCs up in Q4
- Google’s history of algorithm adjustments as corrections
- And some actionable tips on what you can do to combat the increase
Listen to the Episode
Andy Taylor is a Senior Research Analyst at RKG, responsible for analyzing trends across the digital marketing spectrum for best practices and industry commentary. A primary contributor to the Merkle | RKG Blog, Dossier, and quarterly Digital Marketing Report, his 4+ years of experience have seen him master and provide valuable insights into topics that extend across paid search, comparison shopping engines, display advertising, SEO, and social media.
Prior to coming to RKG, Andy worked as an event organizer for a political campaign and dabbled in freelance writing. A graduate of the University of Virginia with a degree in Economics, he likes to spend his free time watching documentaries and selling homemade ice cream sandwiches at farmer's markets with his wife.
JD Prater: Andy, welcome to the PPC Show.
Andy Taylor: Thanks, JD. Glad to be here.
JD Prater: Yeah, man. I'm pretty stoked to have you on to be talking about some of the reports that Merkle has been putting out, as you are kind of the guy writing them all up on Search Engine Land and on the blogs, so for those listening, why don't you tell us who you are and what you've been up to?
Andy Taylor: Sure, yeah. I'm Andy Taylor, Associate Director of Research at Merkle, and so in that role I'm primarily focused on looking at how different trends are taking hold of the channels that we manage. My main focus is on paid search, but I also dabble a bit in paid social, display advertising, SEO, Amazon ads, all of that good stuff, and so I usually package that stuff, like JD said, into blog posts and then also speak at a few conferences here and there, just to give the industry an idea of what we're seeing. We have a pretty large data set that we can play with, and it's pretty easily accessible for us, so I like to slice and dice it and see what's causing the different trends that we're seeing for our brands.
JD Prater: Yeah, and if you guys haven't ever seen a Merkle digital marketing report, you really got to. I'll make sure to link it in the show notes, but these reports are fantastic. I know, I've been keeping track of all your write-ups and all of your posts now for quite awhile now and I've always used these, whether for benchmarking, but also just kind of seeing the change that we're maybe experiencing across the industry, so thank you guys for putting those out.
Andy Taylor: Yeah. Thanks for reading. That's what we're hoping for.
JD Prater: Nice. Well, let's get into this one. You recently published an article on Search Engine Land called What's Going on With Google Brand CPCs? Why don't you give us a quick summary of what you found?
Andy Taylor: Sure, so we saw a pretty big shift in Q4 in how much our advertisers were having to pay for traffic to their own brand keywords, so these are our advertisers' own trademarks, and what we found is that while we were seeing a decline in Q3, we actually saw brand CPC go up pretty meaningfully in Q4, up 23% year over year, compared to a 13% decline in Q3, and so typically, particularly with brand keywords, the price that you end up paying is more a result of the Google algorithm than it is a result of the competition for most of the brands that we deal with. I mean, certainly there are brands that have competitive brand keywords and who are jostling for position on the page, but the vast majority of the brands we work with, typically the price they pay is as a result of whatever Ad Rank Google has assigned their brand keyword, and then the corresponding CPC that goes with that.
What makes this increase interesting is that, so with the decline in brand CPC that we saw in Q3, it seemed to be aligned pretty well with a May change that Google made to AdRank, in which it came out and said that they were making adjustments to how they calculated both first page and top of page minimum Ad Ranks, so the Ad Ranks that brands are required to reach in order to show either on the first page or at the top of the page above organic results, and so at the same time they also said that they would, in some cases, be taking the bid into account more for some queries, "depending on the meaning of the query."
It was a pretty vague update, but what ended up happening is that we actually found that our ... Right after that update we found that our brand first page and top of page minimum bid estimates went down pretty meaningfully and that seemed to align with that CPC decline, and throughout Q4 we actually saw those minimums decline even further, and so we've been databasing this information for a few years, so we have a pretty massive sample set. We don't think it's the result of any kind of outliers or a specific client. Pretty much across the board we're seeing much lower first page and top of page minimum bids for brand keywords, but our brand CPC was up pretty meaningfully in Q4, so kind of running counter to what those kind of directional signals would indicate from those minimums. But certainly not the first time that Google has had stuff like this happen.
You know, like I was talking about in the post, back in the year 2011, we actually found that brand CPC started declining pretty quickly, and by mid-2012, they were 30 to 40% lower than what they had been at the beginning of 2011, and so at that time we actually saw CPC then rebound pretty quickly. When we reached out to our reps and asked them about it, they were pretty straightforward and said, "Yeah, well if you actually look at it over a longer timeframe, you'll see that your CPCs have been declining over the past couple of years, and that's because basically we've rolled out all of these ad extensions, including enhanced site links and other options that advertisers could utilize to increase their click through rate."
But they hadn't on the back end really adjusted how they were calculating Ad Rank for the algorithm, and so what happened was that they were receiving a lower CPC because their assessment of quality was going up because it wasn't baking in the impact of those ad extensions that had been loaded. I think this might be kind of a similar situation, in which perhaps Google may not have intended for CPC to decline at quite the rate we were seeing following those AdRank changes, and that something was adjusted in Q4 which caused them to rebound. However, it is certainly weird that the minimums kept going down even though the CPC went back up. I just word vomited a lot. Any thoughts on any pieces of that or any questions?
JD Prater: Yeah, man. Let's break it down. That's a lot of information that you just gave us there. I mean, super valuable information. Let's start with maybe even just methodology. I'm sure ... We get a lot of questions even around our AdStage reports. I'm sure you guys get a lot of questions too. How do you guys determine brand versus non-brand? Is this something that account managers are doing with maybe their campaign names, or are they using tags?
Andy Taylor: Sure, so we actually assign it at the keyword level within our database, and so each of our keywords is denoted being brand or non-brand, and so we can pull it directly from our proprietary system. When we're pulling out information directly from the UI, then we really on the campaign name, which does have denotations to signify which campaigns are brand versus non-brand, and then we aggregate that data across a fairly large set of brands that have been with us long enough that we can report on their results for the past couple of years.
JD Prater: Got you, so I mean, really pretty interesting stuff there too. We could definitely dive into how do you make sure that everyone is labeling their campaigns correctly? But that's always a problem coming from an agency side, but so now that you have all this information, it's in your database, what does that process look like whenever you're trying to figure out these changes? Is this something that you guys just have up, maybe in a real time-ish dashboard, or just something you guys are diving into and then finding this conclusion?
Andy Taylor: A lot of this stuff is really just the result of kind of one-off investigations. We really started to pay attention, particularly to the first page and top of page minimum bids back at the end of 2014. We had been databasing that information for awhile, but what happened at the end of 2014 is that we saw ... Kind of similar to Q4 we saw a really huge increase in brand CPC, and at the same time also an increase in non-brand CPC, and click growth kind of bottomed out at the beginning of 2015, and so we were kind of paying a lot of attention to how our ads ... Our position on the page, which actually moved further up the page.
We were paying attention to impression growth, because impression growth actually fell off a cliff, and at the same time first page and top of page minimum bids skyrocketed, and so basically we deduced that Google had updated their algorithm to actually feature fewer ads per page because our average position was going up the page. We were getting fewer impressions. All of these minimums were going up, so basically we just use these bid estimates as a signal for what's going on in the universe. It's not a perfect signal, as obviously the continued downward trend of brand minimums indicates. It's not a perfect one to one, this is what you can expect. But yeah, we usually, we just try to segment it as best we can and try to figure out what variable's causing the drop.
JD Prater: Nice, and within your current role, so within your own title, it's Associate Director of Research, how much time do you spend putting together these type of reports, going through the actual research versus maybe like hands-on management?
Andy Taylor: Sure, so for me, in terms of handling management, I do basically none of that anymore. I'm more called in on one-off projects and help our specific client teams with particular questions or problems they might have, but realistically I'd say I spend probably about 75, 80% of my time really just poking around in stuff like this and trying to figure stuff out, and then turning that into content.
JD Prater: Nice. Yeah, so now let's get into the fun stuff. Let's get into maybe some of the methodology. Whenever you put out these type of reports, I don't think a lot of people realize how much time, effort, analysis, weird questions you end up asking yourself happens, right? I notice, for example, you guys use median, right? I think a lot of times we want to, as PPC people, think about averages. Maybe talk to us about why you use median versus an average.
Andy Taylor: Sure. For our larger statistics like Google spend growth or something like that, we tend to use aggregate but for a very curated sample set for brands that we can feel very confident haven't meaningfully changed their goals over the past couple of years, who aren't massive outliers that are throwing the figure five or 10 percentage points by themselves, so when we do use aggregate we use a very curated sample set and I got to give a shout out to Mark Ballard, who does a lot of that good work, figuring out what the best client sample sets are, and is also a huge part of the DMR. But then when we use median, it's more to try to get at what your standard advertiser is seeing, and you know, it kind of prevents one or two advertisers from throwing the sample just by being large.
Certainly some of those bigger brands can spend hundreds of millions of dollars a year, and that could totally dominate a sample set, but if you're looking at median that basically makes those large advertisers just one data point in the set, and lines it up with smaller advertisers on an even playing field, so especially with faster analyses that we're trying to just churn out and figure out quickly, then we'll probably lean on median. But when we're looking at those big, overarching themes like how much advertiser spend is spent on Google, that's when we turn to more of a curative sample set and do use aggregate.
JD Prater: Nice. That's a really good explanation for all of you guys listening that can't remember the difference between average and median. You should go Google it because ... I know for AdStage, for 2018, our benchmark report, we're going to be switching over to median actually. Our product analyst who kind of helps me run these benchmark reports and helps put them all together, he's really been taking a lot more of these data science classes and so he's like, "Yeah, you know, we need to get a real sample set. We need to make sure we get rid of outliers. We need to make sure that we have median, and if we do get rid of outliers, when do we do it? When do we not do it?" For me, I'm just like, "Yeah, sure man. Whatever works."
Andy Taylor: Yeah, and that's the funny part. I mean, really at the end of the day it does just come down to making some of those judgment calls and going with what seems correct and what seems to be the most honest approach at looking at the numbers, and so as a data scientist, you can make the numbers look however you want to make them look, but at the end of the day we're just trying to be as accurate as we can. That's really our goal for the most part.
JD Prater: Yeah, I think that's what he is ... Josh, you know, hats off to him too, right? He is really just trying to be as honest as possible, right? You know, for me the advertiser's like, "Look man, I'm not going to second guess that, so just throw it out there," you know? That's why we've, in our benchmark reports, we've been showing more histograms lately, is to really show where you line up, if you're in the median, if you're a top advertiser, you're maybe in the lower end.
Andy Taylor: We have actually really considered using histograms in the past just for that reason, because it carries such a good range and gives people a better idea of where they fall, but somehow we haven't actually made it into the DMR yet, but yeah. We should probably consider something similar.
JD Prater: Well, nice. Yeah. You know, again, for me, it was our compromise.
Andy Taylor: Sure.
JD Prater: But anyway, we continue to get better and I'm sure you guys will continue to get better with yours, so whenever you're kind of pulling out that data, something that I've wondered for you, so why publish this information, right? I mean, don't get me wrong. I love it. I love seeing the information as an advertiser, as someone in the industry, but whenever you're thinking about maybe competitive advantage, do you think this is highlighting your skill set, or do you think maybe you're giving away too much?
Andy Taylor: Sure. There is that fine line. I would say, especially for the stuff that we include in the DMR, we really just try to keep it to what our advertiser's seeing, and so that can be helpful if you're talking to a brand that's not growing quite as quickly as their peer set, just trying to get to the bottom of why they might be lagging, and so it can be helpful in that sense, in talking to prospective clients or in terms of current clients, just giving them a benchmark to go off of.
But the magic bullet that's going to keep your brand CPC from going up, we're really just trying to tell the story of what's happening to most brands. I'd say it goes back to just trying to be honest as opposed to good or bad. That said, I do think the way that we end up curating our sample probably does have a positive impact on how the numbers end up showing, because obviously if these brands have stuck with us for a couple years, then they've likely been doing fairly well with us, so obviously that does have some bias inherent in it. But hard to avoid that when you're just trying to trend stuff over time.
JD Prater: I can tell you, as someone who's worked in an agency world, I see you guys putting out these reports and all I do is get really jealous. I'm like, "Holy cow. These guys are able to pull these kind of reports, and they're able to get this kind of insight." I can tell you beyond ... Against you, it's like, "Wow. These guys are way better than us."
Andy Taylor: It all goes back to really we came from a legacy agency called RKG, and back when we were setting that up, our founders had a really good, clear vision of what they wanted our database to look like and how accessible it needed to be, and so they kind of set us up for success for a long time just by sucking a lot of the stuff into our database and making it really easy to pull this stuff.
JD Prater: I know. It's one of those things where the word database now is getting more popular, but when you guys were doing this years ago, it wasn't as popular, so hats off, again, to those guys for being really forward thinking and understanding that realistically data is going to be the champion moving forward.
Andy Taylor: Yeah, for sure, much smarter guys than me, for sure.
JD Prater: Cool. Well, let's move on. One of the things I thought was really fascinating with this article was whenever you guys published this, so kind of in the vein of releasing this data, you guys put out a blog, and it was highlighting this increase in phone brand CPCs, and quickly, like after you guys published that blog, there was a huge decrease in brand CPCs. Talk to me about that.
Andy Taylor: Yeah. You know, with anything like this, Google is never going to come out and be like, "Yes, you found a problem and we fixed it because of you." But in this case, we're pretty sure that that might have been the case, particularly because it took about two days really for the CPC to start dropping as soon as we released a blog post on it. Then further into the following week, we actually had reps reaching out to us, trying to reimburse clients for spend that happened as a result of that increase.
Especially for me, as someone who's more in the weeds in the background looking at a bunch of numbers, sometimes I don't have a massive impact on our accounts day to day, so it's kind of nice to see that we can, even when it's just reporting on the trends and keeping people abreast of everything that's happening, you know, it can induce change from the search engines if you can give them a good story and explain what you're seeing, so yeah. I'm pretty comfortable saying that we caused it to go back down, but I mean, I'm sure Google would deny it.
JD Prater: Again, signals, signals.
Andy Taylor: Exactly, exactly. Directionality.
JD Prater: No, I think it's really good. I mean, talk about powerful reporting. Whenever you're able to ... This goes back to the database, having things in a database, being able to surface these type of insights and then being able to publish them in way that tells a story, that shows clear trends one way or the other, I think is just absolutely fascinating, so hats off to you guys for being able to put all that together, and then ultimately I get to feel the impact with brand CPCs coming down, so thank you guys.
Andy Taylor: Yeah, totally, yeah. Hopefully it helped everyone.
JD Prater: Yeah. A question for you that I had was within your database, so this is getting kind of more on your tech stack here. What do you guys use to kind of visualize your database? Are you guys using mode or something like that?
Andy Taylor: Yeah, so we have some automated dashboards created off of R and a couple other reporting possibilities, but most of the stuff that we put together is actually pulled using old-fashioned SQL commands. A lot of it's ... Particularly when we're trying to figure out what's going wrong with a specific segment that we haven't really dove into for awhile, a lot of it's very one-off, specifically looking at particular slices. While we do have a lot of it kind of automated for our analysts to quickly pull reports for clients on a day to day level, when it comes to really looking at these weird trends that pop up from time to time, there's just no way to automate the reports ahead of time because they're so ... It can pop up anywhere, and it's just not something we're looking at day to day, so a lot of one-off stuff as well.
JD Prater: Nice. You mentioned two things that are on my bucket list for 2018, and that is to get better at SQL and to learn R.
Andy Taylor: Yeah. I do not know R, so yeah, I'm right there with you. I think it's a really good skill to have in the quiver.
JD Prater: It's one of those skill sets, you know, where you ... As we kind of talk about this rise of the paid marketer, or the rise of the technical paid marketer, and really understanding the value that someone like an analyst can bring, where before they've always just kind of been in the background, but now you're able to bring them to the front with all of these insights and their ability to query a database and then maybe use R for statistical modeling or something like that is absolutely invaluable. I'm super jealous of those skills.
Andy Taylor: Yeah. I'm super glad that when I came on board, that SQL was already a big part of their training, because it was just a great skill to learn early on, and I think moving forward, that's only going to become more important. Like you said, it's just nice to have people who can do both, who can model the data, create it, and then tell the story.
JD Prater: All right, Andy, so let's wrap up here. In conclusion, right, with all of these reports that you guys put out within the Search Engine Land blog, wrap it up for us. What are some of the things that we should be looking forward to?
Andy Taylor: Sure. I think, especially for the first half of 2018, I think CPCs are going to continue to be up year over year, particularly for brand but also for non-brand. We saw those start to go up following the May AdRank change. The good news for advertisers is that we also saw a corresponding increase in the value that brands were getting, so it seemed like Google may have pulled some levers that actually resulted in better ads being shown to users, which is a good thing. It means that brands are getting a corresponding value for their increase in cost per click. At the same time, certainly be looking to figure out how you can limit the increase as much as possible, particularly on the brand side.
Most advertisers are pretty willing to just eat the increase in cost because they want to control that messaging and ensure they're getting as much traffic on their brand ads as possible, but that doesn't mean you can't play around with bids, try stepping down here and there to see if you really need to be bidding as much as you are, or if you can get most of the traffic for a lower CPC. Then at the same time, just I feel like ... We talk about all the new bells and whistles and the new trends and everything, but realistically, paid search management is pretty similar today as it was a few years ago. Just be churning through the different segments, making sure you're using the proper negatives, targeting the right keywords, using the right match types, all of that good stuff, is still super important.
JD Prater: All right. Well, Andy, thanks again, man, for coming on, talking to us about what you guys found with brand CPCs. Absolutely fascinating report, and I look forward to seeing more reports like this come out in 2018.
Andy Taylor: Yeah. Thanks so much, JD.
JD Prater: All right, everyone. That was Andy Taylor of Merkle talking to us about their latest report about brand CPCs on Google declining, so go in, check it out. I'll link to it in the show notes and we'll see you next week.