Amazon PPC Placement Testing Strategy [The PPC Den Podcast]

Michael Facchin: What is going on, Badger Nation? Welcome to the PPC Den podcast, the world's first and longest running show all about Amazon advertising, to make your Amazon PPC life a little bit easier and a little bit more profitable. We have been podcasting now for over 350 episodes, so if you want a checklist of a lot of our topics organized in a nice Google sheet so you can find what you are looking for no matter what stage of the Amazon PPC game you are in, please check the description. Today on the show we have friend of the show, Mike from Adverio. We are going to be talking about placements, top of search, rest of search, product page, something that is still even today a little tricky to wrangle for so many. We are going to talk about what it might look like to have one campaign where you bid aggressively for top of search and the exact same campaign in another where you push a little more aggressively for product pages. And we give you a nice action item toward the end about testing this in an auto campaign. Let us jump in.

Michael Facchin: Welcome back to the PPC Den podcast, Mike. Great to have you. I am a big-time runner, and recently you got into hybrid training. What is this and how has it impacted your Amazon performance?

Mike Danford: Interesting. A hybrid athlete is basically an event or competition where if you are a runner, you are not going to excel at all the events, and if you are a strongman or strongwoman with power, you are not going to excel at the running portions. So you have to be a hybrid, somewhat relatively strong for your body weight, and do cardio, which nobody likes. I have been pretty active since grade school, bodybuilding, powerlifting, and CrossFit since 2017. This is a totally new thing for me. Now I do Zone 2 training, which I had never done before, it was always go, go, go. Zone 2 seems boring, but when I realized what it does for fat burn, I wish I had done it years ago. It has been a fun journey. I am not sure how it applies directly to my Amazon experience, other than always looking for that additional optimization and how to get more out of what I am doing.

Michael Facchin: I agree with that exactly. If your mindset is that you are an optimizer, it carries over to the rest of your life. And the healthier your body is, let us get real, all of us listening are sitting behind a keyboard for a big chunk of the day. If your body is wrecked outside of keyboard work, it is easy to feel depleted. So a little endurance, being able to push through the slog of a workout, does apply to business, digital marketing, PPC optimization. I actually think exercise is the one non-business thing you can do that makes your business life much easier.

Mike Danford: A hundred percent. I could not imagine just sitting at the desk and not doing everything else I do. Now I have a sauna at the house, we have the ice plunge, I do that every morning. You mentioned coffee earlier, how much coffee can I get today, and I am like, I like the ice plunge instead.

Michael Facchin: So today we are touching on a really interesting topic, one that is still not fully understood by most people, only because Amazon makes it a little tricky to fully pin down: placements and how they operate in sponsored products campaigns. And there is a degree of logic failures that come into placement settings. You mentioned something really interesting, taking sponsored product campaigns and duplicating them targeting specific placements, either product pages or searches, top of search, rest of search. Let us start with the basics. We are talking about sponsored product campaigns. What are placements, and what is the intention from Amazon for them?

Mike Danford: Yeah, so it is different types of campaign targeting. Sponsored products, brands, display, DSP, on Amazon, off Amazon. This is more about where does this version of the ad show up. Different placements have different impacts, different visibility, and you want to optimize based on that. Certain campaign types let you do that by adjusting the bid potential for a different placement.

Michael Facchin: It makes sense that if there is a placement at the very beginning of the search results versus on page two at the bottom, the conversion rates can be different, the competitiveness and the price you will pay will be different. So where a piece of ad shows up on Amazon is really indicative of the results. Some general trends: top of search is going to be more expensive, probably convert better, matter more to organic ranking, potentially more competitive. And deep on the rest of search, it might be softer competition, you might pay a little less, conversion rates might be a little lower. I also want to mix that up, because there are so many factors on Amazon. I have seen scenarios where that is flipped, where certain products behave differently in one category versus another. The point is performance can be different at these placements, and it is important to understand what they do. What would you say Amazon's intention is, why did Amazon create these? Because at one point there were no placements, then there were placement bid modifiers where you can bid more aggressively for a particular placement. Notoriously Amazon does not allow decreased bids for some funny reason. What would you say the intention is?

Mike Danford: I think in the early days, when they were trying to understand where Amazon made more money, they started realizing there was a big delta in different placements, top of search, rest of search, banner ads, different placements on their own properties. So if they are running analytics for these, they figured, if we share these performance metrics and people want another way of optimizing their ads, they will pay more for these or less for those and see what happens. That would be my take on how it began.

Michael Facchin: And again, notoriously you cannot bid down on these. You cannot say, decrease my bid on product pages by 40 percent. You also used very specific language when describing placement modifiers. Basically it is a bid modifier up. If I go into a keyword bid and set it to a dollar, then set top of search to a 100 percent increase, that ends up being a two dollar effective bid. And what I have generally found is that your base keyword bid matters a lot, meaning you cannot simply have a 10 cent bid and then a 900 percent increase to get a dollar effective bid, because Amazon appears to look at that keyword-level bid to assess if you are worthy of the auction itself. Have you noticed that as well?

Mike Danford: That is a great question. We have several brands, and I will go with the bedding brand because it is the easiest. Higher AOV, so 100 to 150 for the comforters. Ironically, most of our bids for those are anywhere between 5 and 10 cents base, and then we have pretty high relative placement multipliers on top of that. Two things come into that. One is the typical CPC for that particular category and product. We will have another brand where it is a 15 dollar product, one tenth of the AOV, however the CPCs are six times that of the higher AOV products. The biggest thing is the conversion, the scrolliness for a category. For bedding, it is very common for people to click around, look at different designs, styles, densities, fabrics, comforter versus quilt versus duvet, so there are a lot of clicks and the clicks need to be cheaper. For the apparel products we have, they are very specific and the search intent is much higher once you get into those search terms, so the number of clicks to a purchase is much lower, so you can pay a higher cost per click. So it is relative to CPCs and category. If we had a very low bid of 10 cents in a category where it is a 1.50 average, I do not think we would get that visibility, and it would be hard to get a high enough actual click with that base bid so low.

Michael Facchin: Big time. And that is so true with everything, which is why it is really difficult to have hard and fast rules like always do it this way, never do it that way. It really depends on the conversion rate of the category, the pricing of the product relative to its peers. A good Amazon advertiser is aware of that. It is really important to test in a controlled way, test certain products or campaigns a particular way, and see how it behaves for your product in your industry. That is such a vital skill, because if there truly was one best way for every scenario, there would be no agencies and no confusion, you would just turn it on.

Mike Danford: I am going to interject a little. That is why we double down in soft lines and larger catalogs and in certain verticals we already know and have experience in, as opposed to dipping into something where we have less experience. We have learned what works at a category level, and that is what applies across multiple brands. It has been great for us.

Michael Facchin: Even Amazon does it with their own reps. You talk to a rep and they will often be in a particular category so they can gain skills there.

Mike Danford: With Walmart, they had the category specialist, but we are often teaching them what works and does not work in the categories.

Michael Facchin: Now that we are aware of what placements are, I want to touch on this idea where you take a campaign, duplicate it, and focus on certain placements. How did you come up with that idea, and what does it look like in practice?

Mike Danford: Absolutely. We are really big fans of negation and heavy negation, and going to a more general or reach targeting, broad, auto, phrase, as opposed to exact.

Michael Facchin: Can you pause there, that is really interesting. Just to clarify, you like broad type targeting and then you are heavy into negating. I did an analysis earlier in the year of really high-performing accounts spending a lot with a very low ACoS, and those were the characteristics I found time and time again: heavy negative keyword counts, many more times negative keywords and negative targeting than positive targeting, oftentimes 20 times more. So they had 2,000 keywords and 40,000 negatives, an intense amount of negatives. Is that what you are saying?

Mike Danford: Yeah, we have actually developed a metric specific for that, with correlation factors to understand it. In general, for a lot of the categories we have, we have large catalogs, so we have a lot of products competing for the same placements on search, off Amazon, product detail pages, so we have to be very strategic. The clicks are much cheaper on a category target or expanded ASIN target or a broad, or even an auto, and we still lean heavy into autos, which seems anti, but with that you have to be extremely diligent and more vigilant with your negation. A year ago I would have never said I would negate a term with only four or five clicks in a 30 or 60 day timeframe, but for us with this larger reach targeting, we are okay to do that, because we have thousands of other search terms we are converting on, so we do not need to go for that one. And we still want the search terms we are converting on to get those really cheap clicks, because we get the same click-through and conversion rate just cheaper if we can optimize for those broad, expanded targetings.

Michael Facchin: Well said. Back to my initial question, where did you come up with the idea for placement-specific campaigns, duplicating a campaign and having it focus on a certain type of placement?

Mike Danford: I think you have it, it is called the ACoS power ratio, I believe it is your metric. We have one that is very similar in house that we call GEAR, growth efficiency and advertising ratio, with some weighting tool with incrementality. Long story short, 85 to 95 percent of some brands' spend is going to clicks that are not converting yet. We still have 3,000, 5,000, 10,000 converting search terms every month. So we thought, why do we not just do a full stop switch and go all exact on these terms, because we have 100,000 search terms each month and only about 5 percent are converting. Well, that does not work, we kind of knew that, but the brand wanted to, and we have to answer to the brand, we have to explain why we spend 90 percent of our spend on terms that are not converting. Long story short, that resulted in CPCs going through the roof and missing out on tons of things.

Michael Facchin: I think that is one of the most annoying things in all of PPC. Why can you not just do that? Imagine you went to your account and said, show me all search terms with one order in the last 60 days and just turned everything else off. That is another strange quirk about the way Amazon works.

Mike Danford: Absolutely. So doing that and confirming with the brand, in their own money, that it does not work, and there are parts of it that do work, so do not get me wrong, use all the match types. When we go back to the broad and automatic campaigns, we are like, okay, why do we not be more aggressive on our negation, we have our thresholds and formulas depending on relevancy, clicks, conversions. Let us get rid of everything that does not convert but keep the 5,000 or so search terms active and hopefully trigger in the broads and categories. Well, one thing that really impacted that is when we negated ASINs in bulk, we actually stopped showing up for search terms in certain campaigns, and that was the quickest thing we noticed. Product detail page placements dropped, the conversion and traffic impressions on those placements dropped overnight, very instant. We did it at scale, and we want to do it at scale so we can understand quickly what is happening. So we reversed it, and then we dug in and found out, when you have a campaign going after ASINs or triggering for ASINs and you want to negate it just because that product detail page for that ASIN does not convert well, you also block the search terms that ASIN triggers for in that campaign. There must be a way where you are negating enough ASINs enough times that block the same keywords that it says, okay, you do not want to show up for these keywords in this campaign, so it blocks that keyword. That is the theory we have. That is what started all this.

Michael Facchin: One thing Mike mentioned is the amount of non-converting spend in an account. We actually rolled out a non-converting spend dashboard where you can track for any PPC account the amount of non-converting spend over time so you can see how your ACoS changes when you spend more in relation to your non-converting spend. It is one of my favorite new tools we have released, because it lets advertisers advocate for themselves and claw back some ad spend to dedicate toward spend that actually converts. I always say you do not have an ACoS that is too high because of the things that do convert, you have an ACoS that is too high because of the clicks that do not convert. So you notice a high amount of non-converting spend, you flick on only exact matches, CPCs go through the roof, and then when you backtracked you noticed something interesting about adding negative ASINs and how that impacts search terms. Just to call out a quick point, on Amazon PPC you can target ASINs and then see search terms in your search term report, and vice versa. This is a strange thing on Amazon. Break that down again for people who might not be aware this is happening.

Mike Danford: Sure. Some brands may not even know how to use an ASIN or category or product detail page targeting. If you have search terms you are going after, Amazon is going to say, hey, there are products relevant to your search terms, and we think you will convert for it, so we also want to give you placement visibility on that ASIN and its product detail page. You are not targeting that ASIN, but we think you should, so we are going to force you to. That is a good thing, but we also have to take some of that control back.

Michael Facchin: So you notice this is happening, that you are targeting keywords and showing up on product pages, which convert differently than keywords, and that inspires you to say, what if we had a campaign focused just on search and another focused on product pages? How did you architect that, are you literally just copying a campaign, same keywords, one gets a higher product page bid, the other a higher top of search bid? Talk me through the structural mechanics.

Mike Danford: Let us take one step back to clarify that you have keyword targets and ASIN targets, and both of those have product detail page placements or search result placements. When you have an automatic, keyword, or ASIN targeting campaign, if you negate an ASIN in those campaigns, some at ad group level, some at campaign level, you are going to block search results. We do not want to do that, we only wanted to block that particular ASIN's product detail page because it did not convert at the level we wanted. So how do we prevent that? You mirror the campaign. It can be the same exact auto, the same exact keyword campaign, the same exact category or ASIN campaign. And you set the base bid, to your point earlier, low enough that it does not show up on the search results. Then the inverse is, if it is a search results campaign, you have a really high top of search, probably about half of that on rest of search, and no placement multiplier on product detail page. Once you get that figured out where your base bid is low enough, you should not have any ASINs or product detail page placements in that campaign. Then you do the exact opposite on another campaign that is identical in targeting, the only difference is the placements are inverted, so no placement multipliers for top of search or rest of search, but a high product detail page multiplier. Your base bids are usually pretty consistent and close in the campaigns when you start, but detail page CPCs, conversion rate, and click-through rate are much different than on a search result page. This gives us two things. One, it is very clean in the campaign manager, and if you are using third-party tools, it helps with bundling and grouping so you can see performance over time. And it gives you another level to pull without over-negating or over-blocking some other placement.

Michael Facchin: How do you ensure things do not eventually converge? If you are optimizing for performance, what does the team keep in mind so they do not accidentally change bids in a way that undoes what you intended?

Mike Danford: Great question. We have it built in where if it is a search results campaign, zero is the hard set on the product detail page multiplier, and it can never be greater than zero. And you are more constantly optimizing the placements than the bid, because when you move the bid you do not want it above a certain number. We know the average CPC for a brand or campaign, and we say this is the threshold, usually half of whatever the average CPC is, and that keeps it where you can move the base bid up and down without eventually commingling and going back into a placement you did not want. There is also a monthly reset to make sure. Even as vigilant as we are, we still have random product placements in a search campaign and vice versa, but it is a very small percentage, and the ACoS is single digits for that, so it is not moving the needle.

Michael Facchin: You are doing this primarily on sponsored products, is that right? And in what percentage of your campaigns are you doing this, have you noticed areas where it works better and areas where it does not?

Mike Danford: Great question. Where we see the bigger impact is automatic campaigns and category campaigns. When you get into the more specific keyword versus ASIN or expanded product attribute campaign targeting, it gets a little different. But if you want to start somewhere, it is those. If you are not running automatics, go back and run automatics and figure this out, along with the negation we talked about. That is where we see the biggest efficiency created through this strategy.

Michael Facchin: Awesome. So one action item worth experimenting with is to duplicate an auto campaign. I have done a lot of testing over the years with auto campaigns, splitting the targets, and I have generally found that works well. My general rule of thumb is more segmentation is typically better unless it becomes so unmanageable that you lose your breadcrumbs. So it is worth segmenting until it becomes a burden or the algorithm does not agree with you. Do you do this with all four auto targets turned on?

Mike Danford: Yeah, interesting. We did that about the same time, one or two years ago, where we split subs, comps, close, and loose match and had mirror campaigns. During all of this we tested it, and now both versions of the campaign have all four of those active, because we do not care how it shows up on a product detail page or a search result, we just want to do it. This segmentation gives us better control. And to your point, that worked better than not doing it.

Michael Facchin: So one auto campaign, all four targets on, then you give one campaign the top of search boost and half of that on rest of search, focused on the SERP, and another focused on product pages that gets the product page multiplier, with the inverse set to zero. That is something for the good people in Badger Nation to test with. I always like to give people something to do, that is their homework for this episode. I love these kinds of experiments where you take something and play with the settings and see how the algorithm reacts. It is frustrating you cannot downbid, because I wonder what this strategy would look like if you could. Closing thoughts, how long have you been doing this?

Mike Danford: At scale, this happened around Prime Day. We did the heavy negation prior to Prime Day because spend was going to scale 5X for a lot of these brands for that event, so could we take a dent out of that non-converting spend, what would it look like? I am so glad we tested it a couple of weeks prior and had time to turn it back on. So about six months, a little over five months, is where we are at, and we are applying this to every brand we work with and every brand we audit. If you do not end up working with us, do it anyway. We have had comments like, wow, that alone gave us a boost we were looking for. So when you do this, try your best to give it a couple of weeks and not do much other volatile updates or optimizations and see what happens. I would wait until after the holiday, do not do it around a tentpole because it will not give you the best results, wait until you have a stable time.

Michael Facchin: Normalcy, yeah. Well, thank you so much, Mike. If someone wanted to get in touch with you, what is the best place?

Mike Danford: Sure, adverio.io. And I will see if we can spin up a webpage where you can submit your results from this test, so hopefully we can give a link update when the podcast releases until we can get some good case studies.

Michael Facchin: Right on. Well, thank you so much, Mike. Have a good one, and everyone else, I will see you here on the PPC Den podcast.