PPC Detectives: Solving Amazon’s Most Mysterious Performance Crimes

PPC Detectives: Solving Amazon’s Most Mysterious Performance Crimes Speaker 1: What's going on, Batra Nation? I'm here for a special episode recorded in Miami. CSI Miami, you might say. Conversion scene investigators. When you get to Miami, everyone here is selling something, from coconuts to conversions. You just hope that you're not the one getting sold. Today on the show, we've got my dear mastermind partner, who I've masterminded with many Wednesdays for years, Steve Patenaude from Montreal. Today, we're going to be taking a look at five crime scenes. Five crime scenes inside Amazon advertising accounts. Being a good PPC manager means being a good detective. Today, my friends, we're going to investigate five crime scenes that you might have inside your PPC campaigns. My friends, let's jump in to the special edition of The PPC Den podcast, CSI Miami. All right, Steve. I remember it well. It was August 22nd, Miami Beach. It was a cold night and we were hired to come in and improve some PPC campaigns. Little did we know, things weren't exactly what they seemed when ACOS started to go down. All righty, Steve, what do we got? You were hired to work on a PPC campaign and it started to go well. But there was a little mystery here, wasn't there? Speaker 2: Yeah. Yeah. So basically everything was shiny. But when looking at more in depth in the campaigns and everything that was going, I started to look outside of just the PPC and saw the brand was doing a push with subscribe and save coupons. So, if I would have just stayed in my lane of the PPC side, I would have probably raised a bit and tried to push more to get more. But if I would have done that and they stopped the subscribe and save coupon push, then everything would have collapsed, basically, and probably ACOS go up and stuff. So, by investigating further on why that Shiny numbers, those shiny numbers were there, managed to learn that and then we could adjust. So I could push when basically they do the push and then lower back when they're not doing it. At the end of the day, it gave us a tool for that specific brand that we just did. So that was a few months ago and we just used the same lever because we had that tool out that They found out the right percentage they had to do and subscribe and save coupon. And then on my side, I was ready to rock with that. And we had good results. Speaker 1: You know, it's interesting because we often do those kinds of investigations when performance is bad in the sense of why is my conversion rate dropping? Why is my ACOS going up? Let me go investigate what the market's doing or if there's been any changes for the brand. And I thought it was really keen of you to investigate when performance is doing better than expected. It's like, wait a second, what happened out here? And what you found was that the customer was doing a subscribe and save push. And as a result, that made all the performance look better. Had you not done that investigation, you would have leaned into that in your PPC campaigns and potentially eaten up some profit. Or once that subscribe and save push is over, you would have been looking at a very high ACOS because everything would have been increased as a result. So that's very Now, you mentioned that, how do you thread that needle? How do you think about threading that needle when it comes to subscribe and save and PPC performance? Meaning if there's a certain percentage of customers for every order you get on PPC, knowing that there's going to be some recurring revenue. So it's almost like you move from a pure ACOS-driven model to a full business Cost to acquire a customer, because you know that if you acquire a customer, some percentage of them will become subscribe and save, making the overall business a little bit more successful. How do you think about goal setting in terms of PPC performance? Like when you're inside and you're looking at ACOS, total ACOS and overall business revenue and maybe overall business profit, how do you factor in subscribe and saves? Like how does that Do you set goals by setting a higher ACOS target, a higher total ACOS target, because you know that some people will be coming back, or how do you navigate that? Speaker 2: So in that, just before to answer, in that specific case, the ACOS went down. So that's kind of a, wasn't really the goal because often with repeatable products, the ACOS, we can Look at, and to answer your question, we can look at a ACOS that is higher. And the two main, for me, the two main numbers, there's multiple numbers, but two that I look a lot with the client is the lifetime value of a customer for that specific product. So the profit for that product, for that customer is different because they're going to purchase two, three, four, five, six, 12 times, depending on How long they stay and the number of subscriber. So those are the two. If your number of subscriber are always going up and you have your lifetime value, then you can really have A sense of the impact of getting a new customer. So how much can I spend to get that new customer? How long is it going to stay? How much profit that customer is going to bring? It's always averages because subscriber and save. In that specific case, I think it's, if I remember correctly, it's like 50% don't do the second order. So you need to figure out that part also is how many and we have more and more data in Amazon on that front in terms of how many do purchase like just the one with the coupon and then they don't purchase anymore or purchase two, three, four, five times. So you need to know those numbers. So those are not PPC, but they are very valuable Information because then we can discuss how much we can spend or invest. I prefer investing to invest and acquire a customer. Speaker 1: So the takeaway here is to never optimize PPC in a vacuum and that the best PPCers are always looking at what factors are influencing PPC. For example, subscribe and save, both from a goal setting standpoint and a performance standpoint. So that's really interesting to hear. Well, it looks like this account just got optimized. I showed up in Little Havana. There's always something magical in the air here. The air was thick with Cuban music and confusion. Now Steve, you set up campaigns perfectly based on all the best practices. However, little did you know, a truck was barreling your way and you weren't quite ready for it. The issue? Listing suppression masquerading as poor PPC performance. And let's dig into this, Steve. This is almost like the opposite of case number one. In case number two, you had been doing everything in this PPC campaign. However, ACOS started to increase and sales started to go down. Another lesson of not optimizing in a vacuum. What was your first instinct when you noticed that, hey, why are all these best practices making performance worse? What did you first think? Speaker 2: The first number before, like, of course, when a cost goes up or sales go down, then it's kind of a red flag. It's a 911 call that you get. Speaker 1: That's why you showed up to the scene. Speaker 2: Yeah. So, conversion rate really dropped. So, something that I like to look at is graphs. So, looking at numbers one by one, sometimes you don't see trends or you don't see So by having graphs, it helps sometimes, and in this case it did, because I could see the drop exactly when in conversion rate. Of course, if I put the same in terms of ACOS or sales go down, like they all go together. It's just one influence, like conversion rate dropping. Usually will impact the rest of the sales and ACOS going up. So they're going together. They're kind of suspects that goes together. So by looking at conversion rate dropping in campaigns, then had to go deeper. In that specific case, It's apparent with multiple listings. In some cases, the campaigns were just that specific SKU. In other campaigns, they were all the full family was in the campaign. For different reasons and different reasons, it was set up that way. So I had to go deeper. The campaign that performed the best was the one with multiple child in it. That was the one driving most of the sales. So that's why that one kind of triggered the 911 call first. And going deeper, I saw that the best ASIN in that campaign stopped to perform. So I had to go Further and further to basically go in the account and not the business, not the PPC side, but the business side and saw that the ASIN was suppressed. And if I put everything together in terms of timeframe, they all were linked together in terms of when that happened, then the rest happened. And then By that one being suppressed, another reason started to get traction in terms of traffic was not converting as well as the other one, because the other one was the best converter in that campaign. So that's by knowing that, then I had to change everything in the account for that specific family because need to kind of relearn once their product was back live with a new image, relearn the conversion rate and everything for that specific situation. Speaker 1: I have one question for you, which is, would you recommend anytime someone wants to change a main image that they use manager experiments, A-B testing? Speaker 2: I don't know. What I would like is to know on my side, and that's something, to be honest, that I've never thought about, is if we use the experiment of Amazon, how will that impact Because if Amazon does 50-50 in the tests, then 50% of the time, your data in the PPC side is going to be influenced by maybe a bad image, right? So yes, on the business side, it's super valuable. But to be honest, I never thought of the impact on the PPC side because maybe 50% of the time, But it's impossible on the PPC side to know which image was served when that click was done. Or impression was done, then we don't have the click because the image didn't bring the click, right? So yeah, then in the experiment side, I never have that. Now I do with that experiment and even another challenge similar was the main image was changed by the brand without me knowing and similar situation happened, right? So if you know it in advance, you can explain the data that you see. Much faster, so. Speaker 1: And the second question has to do with, so the first thing, I actually do like to use manager experiments a lot. I generally find that people that I talk to, situations I'm involved in, if they're using manage your experiments a lot, they're generally in tune with the market a little bit more because they're sort of saying, Whatever image that I think is going to be better, I need to let the market decide and determine that, which I find really, like I said, that's a healthy digital marketing ethos to have. So that's fantastic. Now, so the lesson here is that There are so many factors that influence PPC, and by tracking conversion rate per product in PPC, you should be able to see those changes. And when you see those changes, the first thing to always ask is, what influences this? And one of the things that influences that is, of course, the main image. So this case is closed, and it seems like the only thing getting clicked tonight is the main image. Case three, Wynwood, Miami. Crime scene. The air is still and we show up on a rainy afternoon where I see a campaign and a chalk outline. You had a good life, kid. I see a one word or small keyword campaign with some exact match keywords in there. The keyword doesn't convert and it was killed. But Steve, you noticed something about this particular campaign. Essentially, the situation is you saw a single keyword campaign or low keyword count campaign, something like five exact match keywords in there. And before you went To take this campaign out and maybe pause it so that the traffic can flow somewhere else, you did something that yielded great results. What did you study when you notice a keyword not performing? And instead of optimizing the keyword, what did you do first? Speaker 2: I've looked at the placement. So where did Amazon Send the traffic to, so it can be product pages, rest of search, top of search. When I see that Amazon is sending the traffic, if we don't do placement at out of the gate, right? So if no placement- They're all zero bid modified. Yeah. Yeah. So, because in some cases you do want out of the gate to push with the first search, for example. In those specific cases, nothing was done To bring Amazon to bring the traffic specifically on one of them. So the placement was the go-to to look at and saw that all the traffic went to one of them. So basically lower it a bit. And then push for one or the two others, depending on what I want to test. But if I would have had just rules in bulk sheets or software, whatever, that keyword would have or those keywords would have been paused. And we would not have known that they could work In another placement, so I've seen all placement work for keywords. So top of search, rest of search, product pages, even if it still doesn't make sense in the process of my mind, like product pages, product placement should be just more for ASINs and stuff. So that's still something. But rest of search and top of search for me is I saw all of those work in the past, so that's why now I look more at those before killing anyone or any keyword. Speaker 1: There's a lot of nuance when it comes to PPC. Another piece of analysis that I do as well when determining, like, is this a good keyword or a bad keyword is also what comes up a lot is, well, I'm looking at the last 14 days for a keyword. What did it do in the previous 14 days? That helps me identify any short-term change in the performance. Maybe historical performance is good, but just recent performance isn't. So always looking at the right length of time is helpful too. And in the same vein, also looking at the placements for how did the placements perform. You sometimes will look at a keyword and see the combined performance, And you might notice that a certain placement is doing well, another one's doing poor, so sort of doing that placement reset in a way where you look at the campaign's influence on it. Speaker 2: In the same vein, if you don't have like segmented keywords or isolate keywords, so something else that I saw a few times is the same search terms The same search terms, sorry, perform better in one spot than the other. So if you have the same search terms in different campaigns, so let's say auto campaign, broad and exact match, for example, if you do just a pivot table and learn about that specific search term, it might get an okay type of results in terms of conversion. But if you look at them one by one, maybe your broad is converting a lot better than your exact, and then you can learn more about placement. On those specific ones, which one brought more of that kind of versioning? Speaker 1: Yeah, I was just going to add, I was about to also say the same thing, which is like duplicated search term analysis, where you like look at each search term in aggregate, like everywhere it appears, and then you dig in and see every single place, every single ad group that that search term got a click. And then you can learn even more about this Particular search term before you make a final judgment on it. So yeah, being a good PPC manager is being a good detective as well. So the verdict here wasn't a bad keyword. It was just placement distribution. They thought that they were bidding high, but it was only their hopes that got raised. Crime Scene 4, Coconut Grove. Case number 458A. Unsolved. Alrighty, Steve. It's a case of a runaway. A runaway auto campaign. What happened here? Speaker 2: On a slow night, a speedy car just came. But anyway, what happened is an auto campaign. An auto campaign just started to spend five times more Out of the blue, if you look at the graph, steady, steady for weeks and weeks of spend, and then 5x the spend without any change, because that's the first thing I've looked is why so did Did someone did something to the bids or placements or whatever could have impact that? Speaker 1: It can happen a lot. A lot of people have access to an account. You never know. Someone might just get in there and change your budget, change a bid without telling you. Speaker 2: Yep. And no, the only thing that changed that it's a case unresolved so far because it's impossible to make sure it is the only element, but the main image was changed. So that's basically price was not changed. Like the rest was not changed. But because if the price was changed, that could have explained if the price dropped, could have explained the Amazon bringing more impressions and love to that product. But no, the main image was the one. Speaker 1: You know, it's interesting with PPC, you'll be inside a campaign looking at it down at the most granular level of a search term and then the target and then the campaign. And then you get even further up, which is like the individual product. This sounds like a case of maybe a competitor, perhaps. Something happened to a competitor. Maybe a competitor that gobbled up a lot of sales went out of stock, perhaps. And they going out of stock resulted in you getting more traffic. So it's almost like the, this comes up a lot, like being a good PPC manager is being a good detective and like knowing like, was there a change in the campaign? Was there a change to the product? Was there a change to the competition or the market? Each one of those layers is really Vast, and figuring out where the change happened, it can be really tricky, but it's worth it to investigate. Is that your hunch currently? Do you think that a competitor, something happened to the market that resulted in this? Maybe the product type was on Shark Tank or something, and now people are searching the product more? Speaker 2: Really, the only thing that I could figure out was the main image, but I've seen all that you've mentioned. The Shark Tank, The As Seen on TV, I've seen that one where The product that I've seen on TV was not on Amazon. So a really similar product just started to get, and by questioning the customers that purchased, the customer told the brand that they were searching for that product I've seen on TV. So I've seen it all in terms of what, and that one would have been really difficult to find out as an investigator. A customer would have told us, but yeah, right now the only suspect. That I have is the main image. Still don't get why, but... Speaker 1: Another place I would investigate is, of course, search query performance report, the summed value of all the impressions. I know this is an auto campaign, so you're probably getting a lot of non-search-based impressions, but I think that's interesting as well, because those swings, like if you were to download a search query performance report and sum the Total market impressions. There's some huge swings there from one week to the next sometimes. Sometimes they have no impact. Sometimes they can have a big impact. Another point of investigation might be one of my favorite tools, the Product Opportunity Explorer. They have that graph of how many competitors were advertising in a particular niche. I'm sorry, were available to purchase in a particular niche. They also have, you know, weekly conversion rate and weekly sales, which can be helpful as well. So the data doesn't lie, but it sure knows how to cover its tracks. Alrighty, our last case. Unfortunately, we're in downtown Miami. And unfortunately, we've got a case of domestic abuse. They always say it's someone in your life that you know, who might be an abuser. And this case was no different. It was branded spend driving up our cost and hurting performance. So we decided to put a restraining order on branded spend. Alrighty Steve, This gets talked about so much. Every brand, this crosses their mind, which is, I'm looking to boost profitability. I want to ideally spend more profitably. And they look at their search term report and they'll put a filter for branded spend, you know, show me all my search terms with My brand name in it, and it can sometimes be a huge sum of money spent on terms with a brand's name in the search term. Lots of different companies will have different philosophies on this at different life cycles, at different stages, at different levels of market share of voice or market penetration. And it's an interesting thing to wrestle. And in this case, a brand came to you and said something along these lines, which was, can we reduce the spend to increase profitability? And what makes PPC such an investigative exercise is because you can do this in brand A and have it work great. You can do this in brand B to poor result. And sometimes you can do this in brand C and it's a little bit murky and you need to dig even deeper. What did you find and what have you found over the years of trimming down branded spend? Maybe it doesn't go to zero. Maybe you're just trying to reduce brand spend by 50% or 20% and see what happens. So what have you noticed recently? Speaker 2: The first for me is it depends on the size of the brand awareness or known outside of Amazon. So in that specific case, we're working because it's still right now ongoing because of some of the first results we got out of the test of lowering the spend on the brand side. So we didn't get the results that we were expecting, basically. So I can maybe just share a bit, because one element out of the key, we put some elements Knowing that we were doing that, so we had some cameras a bit everywhere to watch some suspects around that time. So we used search query performance. One of the elements was, of course, tacos, profit, and all of those, because that was the main goal, is to get more profit. But at the same time, We wanted to make sure that we didn't lose too much of market share on our brand. So search word performance is a good place to look. Speaker 1: Because it'll tell you your brand's share of sales for branded terms. Speaker 2: Exactly. So basically, you looked at the search query performance week by week. So it took six weeks before the start of the task. And then did an average per week. So we had enough weeks because week, if I just compare with one week, it could have been like just a fluke week that didn't had really the right numbers. So the goal was to have enough data to really have something to compare off. right now we have six weeks of past and five weeks of the goal to lower. And after the first two weeks, already we saw something and now it's stabilized, but still we're looking to go deeper in that front. Quickly, search volume didn't go down that much. The total of impressions didn't go down that much overall. Even that, for me, is kind of something. So why would we lose 5% of search volume in impressions? But we did lose that in the average, so the six weeks versus the five weeks. The clicks went down about just a bit above 5%. Until there, it's kind of a, okay, it went down by 5%. Why? It's kind of a question mark. But where it starts to get Interesting is at the cart went down by 9%. So we're starting, we lost some search volume impressions, a bit of flake, but at the cart went down further. And then the purchases went down even further. The overall purchases went down by 15%. Speaker 1: And this is just for branded terms. Speaker 2: Terms. Yeah. Speaker 1: Wow. Speaker 2: So that's not we're not looking if we're losing share in the market. This is just numbers to see if we're losing share in our own brand. Right. So now the question is and kind of now it's the investigation is even I would say more fun if you like to investigate because now there's multiple suspects that you can look at. Did the brand, and that's something that we need to go deeper, is did the brand lower their effort outside of Amazon? Because that could be one suspect, right? What type of, or they spent the most, the same, but in different ways. So that could have an impact. So everything outside of Amazon could have an impact on all of those. The one thing that I would not then look at that suspect as the main one is the fact that yes, we did have a lower impressions in search volume, but it's the purchases that went down the most. So conversion rate overall dropped. So then, anyway, that's kind of where we're at. So I have other suspects, but I would like to have your insights. Speaker 1: I mean, it's going to be hard to, I mean, 15%, let's just say 15% fewer branded purchases. It's going to be hard, like that means that spend would have needed to go down as well in such a degree to offset the reduced sales. Speaker 2: All the numbers that I've mentioned is not our, it's the overall market, not our, because we did keep about the same percentage. We lost, I think, I think 2% of the market. So let's say we were at 90% of the purchases on the brand terms were ours before the test. We were at 88% after the test. So we didn't lose that much because that was kind of the element that we were afraid is by not spending as much on PPC, the competitors would take more market share of our Brand, basically. So that's something, you know, I would not want to see as a brand if I'm investing a lot of money outside of Amazon. It's kind of the funnel, the end of the funnel, the bottom of the funnel is the brand terms on Amazon, basically. Speaker 1: Because worst case scenario here is that people, because the assumption of cutting branded spend is always, I can cut branded PPC spend because they will buy from me anyway. So yeah, and then also during it, overall market versus your share of the market is also really valuable as well. So while branded purchases went down, your share of your own branded purchase, your own branded search purchases only changed two percentage points from 90 to 88. And I guess the It's really difficult to test things in Amazon for all these reasons, meaning you could do nothing, change nothing, do nothing, and your performance will change from one day to the next, one week to the next, even if you don't touch a single thing. And it's always difficult to look back and be like, well, let's say you touch nothing and your performance gets better. You always have the feeling of, well, what if I did make some optimizations? Could I have even better levels of performance? Or if your performance goes down, you can always look back and be like, well, if we made changes or didn't make changes, it would have been even worse. So it's always difficult to piece this together. I think the things that I'm mostly concerned with when cutting branded spend is, of course, my share of branded purchases. Because that's really the worst case scenario that you're trying to mitigate against, which is any sales that I lose when somebody searches my branded terms. Will I make up for that by spending less? And depending on how much you were spending on branded spend, how much was cut from branded spend, those two percentage points of lost branded share of branded purchases may be fine or they might not. So it really does, you'd have to look at how much you were spending on these terms. You could almost try to approximate like What those two points were worth, so if you didn't cut brand spend and you stayed at about 90% purchase share, how much would that have cost you? Because you can do historical analysis, so you can assume whatever you cut, add that back, you lost two points of that market share for your branded terms. How much revenue would that have been? And then you can sort of get your analysis there. That's like a fun piece of analysis to write. Speaker 2: To finish on that case, the two percent, if everything else would have stayed the same, I think at the end of the day would have been, well, we went down on the tacos by 2% so we have more profit. That 2% is worth the profit that we got out of it. But the concern right now that I have in that case is the 15% of total purchase for the brand terms that went down on Amazon. Was that because we were not as much shown? So people Looked at the other ones and finished by not purchasing at all on our brand term. So that's 15% drop in total purchase. That's the percentage that I'm looking right now and investigating why that 15% happened. Speaker 1: That's strange. The times where I've seen cutting brand it's been Work well, it's usually very cut and dry, meaning like we get rid of 25 grand of branded terms and we can clearly look at our, you know, profitability and find those extra 25 grand of spend that we didn't spend. We now see that as profit and our share of purchases for our own branded terms has not changed very much. And that's usually because the brand has such an incredible market share that The odds of someone stepping over a competitor ad to go to their organic listing is incredibly high. Not every brand can do that. For a lot of brands, if you appear number one for your organic search result, but there's a competitor there with a better offer, people will go to that ad. So you don't want to do your competition any favors by letting their potentially better offer be shown first. So it's a delicate balance, and I've usually seen it to be Cut and dry. I have a sense here that cutting the branded spin might not be a good idea for this brand. Speaker 2: Yeah, in their case, it's the high-end type of product in the market. So linking of what you've mentioned Maybe the offer of other ones brought the customer to another journey than purchasing our product. Speaker 1: Amazon shoppers have wandering eyes and Amazon knows that well. That's why there's tons of other products on any singular product page for sure. Well, as they say, sometimes you optimize the bids and sometimes the bids optimize you. Alrighty, Steve. Thank you so much for coming on The PPC Den. I hope everyone out there in Badger Nation enjoyed this episode shot and filmed in Miami. Have you ever been to Miami? Speaker 2: Yes, once. Speaker 1: Quick, quick. A quick one. Do you have any memories of Miami? Speaker 2: No, it was too quick to, just a business trip. Speaker 1: Got it. There is a restaurant here in Austin that serves amazing Cuban sandwiches. And every time I go, I'm like, how do I not go to this place every week? It is like the best Cuban sandwiches. So Steve, thank you so much for going over these five cases. I think the big takeaway is to be a good digital marketer on Amazon. You must be a good detective. I've always thought that to be true because when performance changes, The best marketers know why it changed and how can I influence this better in the future. So Steve, there's links to your LinkedIn in the description of this video. Thank you so much for coming on the show. And we mastermind every Wednesday together. So I loved interacting with you and those mastermind calls and looking forward to learning even more from you in the future. Thanks so much, Steve. Speaker 2: Thank you. Speaker 1: And everyone else, I'll see you next week here on the BBC 10 podcast. Unknown Speaker: I've launched campaigns and picked keywords. I've got my bids, set placements too. I've made a few mistakes. I've had my share of prank keywords. I'm The PPC Den, my friends. And we'll keep on the music. We are the PPC Den. We're talking about Amazon. No time for medicons, cause we'll fix the game.