Why Rule-Based Amazon PPC Fails (And What to Do Instead

Why Rule-Based Amazon PPC Fails (And What to Do Instead) Speaker 1: Alexa, play That Amazon Ads Podcast. Unknown Speaker: Which one would you like to hear? Speaker 1: The best one. Unknown Speaker: Okay, now playing That Amazon Ads Podcast. These gentlemen are completely changing the game. Speaker 2: After listening to That Amazon Ads Podcast, my ads are finally profitable. Unknown Speaker: I also heard they're pretty cute. Speaker 1: All right, Andrew, we've got three primary reasons for why these rule-based automations don't work. And the reason why I think this is very needed is because we are just seeing this method of managing is becoming very, very popular. People have said no to AI, but yes to these rule-based automations. And we still see quite a few problems with it. And even though you can have a lot more control, these rule-based automations, by the way, are frequently going to be things like You have a keyword rule where if the keyword's ACoS is above or below some threshold, you're going to increase or decrease by X percent. And then they set up hundreds and hundreds of these rules for different campaigns, for keyword targets, for search terms, et cetera. But there's three reasons why those rules aren't that great. And Andrew, what are they? Speaker 2: Yeah, the three main reasons are number one, these rule-based automation tools limit you to just using fixed look back windows. Number two is they are always optimizing all of the account and that's not something we feel like you should be doing. And number three is that they limit you to only being able to adjust bids using an increase or decrease by X percent or simply pausing keywords. Speaker 1: So we're about to, you know, go through and explain each of those, but that first one there, the look back windows. Yeah, so most of these tools, you can pick that look back window. You can say, I wanna, if this keyword condition for the last 30 days or last 14 days or last 60 days or whatever, falls under this high ACoS or high spend no sales, then we're just gonna like pause it, archive it, reduce the bid, whatever. Lots of reasons for why that's limited, but first and foremost, and we have a whole episode on picking the right date range, so go check that one out. But those look-back windows are highly, highly dynamic components. If you look at my history, my change log within my AdLabs accounts and you look at the date range of selection for each of them, I almost never pick the exact same timeframe. It's usually somewhere between last 14 days to last 30 days. Sometimes I'm going really short timeframe. So if performance spiked, like things went crazy over the last seven days for whatever reason, Then you need to address that. You can't sit around and waiting for the last 30 days to average poorly before you start reacting to it, right? Like maybe these things look great in the last 30 days, ACoS is 30% on these keywords, great. Last seven days, ACoS is 500%. You're not going to want to wait 30 days burning through cash until the overall ACoS average is too high. You got to jump on it quicker. You got to use a shorter time frame. And you're also going to want a lot more controls, but you're probably going to be optimizing a smaller batch of keywords over a shorter date range. So yeah, you need to be able to be a lot more responsive. You need those dynamic timeframes. And I just got off a call before we hit record on this. With a seller who going into they sell like gifts for Valentine's Day is basically their category. And as you can imagine, for the first half of February, spend sales conversion rates were through the roof, massive drop off. And every other system that's trying to automate things would likely be continuing to increase. Because they don't use automations, they were able to manually decrease and protect their ACoS and all those types of things. But they were only able to do that because they were not using fixed look-back windows and they were not just putting things on simple automation. So, yeah. Speaker 2: You gotta be dynamic. You gotta be able to change and pivot those date ranges on a dime. And it would take a lot of work to go through every single one of your different rules that you have running, change all those date ranges to the proper timeframe. You know, not gonna happen. Speaker 1: Here's what's going to happen, and this wasn't even something we said, but either you're going to have way too many rules that you set up and it's going to be a major pain in the butt to manage, or you're not going to have that many rules and it's easier to manage, but now all of the logic in your account is like way too simplistic. And yeah, everything is always dynamic in the optimizations that we're doing, lookbacks, conversion rates, different products, et cetera. So that's used. Speaker 2: All right. Speaker 1: Andrew, second item here. Speaker 2: Yeah, should you or should you not be always optimizing your account? These rules, a lot of times people will come in and they'll set it up across all of their account. And so they constantly have changes going through, just touching every single piece of their account, always adjusting bids. It really muddies the water in a lot of different ways and doesn't really allow you to pinpoint what change is causing what to happen. And so we generally recommend Being more precise with the types of changes that you're trying to make. You don't always have to be running every single type of optimization. Sometimes you wanna skew more towards doing more bid increases. Sometimes you wanna do more bid decreases. And you have to be able to pivot and change what types of changes you're making based on what the client's goals are, what the campaign state is. Are these new campaigns? Are these really well-seasoned campaigns? You're just doing some balanced normal routine maintenance on, you know, there's going to be different types of levers that you're going to want to pull depending on what types of changes you're trying to make. Speaker 1: So if you have these rules set up where you're always increasing on low visibility, low ACoS keywords and always decreasing on the high ACoS keywords and those are just always going all the time, it's really difficult to hit specific client goals because there are some times where the client wants to be a little more focused on growing sales and growing visibility and other times where they want to be a little more focused on reducing spend and increasing profitability. So what Andrew and I really like to do is with any batch of optimizations that we're doing, we try to focus on just one of those goals for that round of optimization. So for this week, it's like, let's just focus on increasing the traffic. And then the next week, cool, we got traffic up. Now let's try to make things a little bit more efficient. And it's also very easy to Track the changes in performance. Where did it come from? Versus if you were if you had a hundred thousand keywords that were all getting 5% increases Oh like, you know, however, and the other thing too is like how frequently are you having these rules run? Hopefully it's on a once a week thing if you're having these automated to run daily We need to have a whole separate conversation with you but I mean for the daily situation like if people are saying I want to increase bids by 5% and And for the low visibility terms or whatever, low ACoS, if you're running that daily, those 5% increases add up really quickly. After a couple of weeks, you've doubled the bids. So you got to be super mindful of that in terms of that frequency. And yeah, it's way more efficient for performance and also troubleshooting to Really just focus on the top 10% or the bottom 20% of the account in any given optimization. So if you're trying to refine things, get the worst performing 20% of the account and really just focus on cleaning that up. Did I say worst 10% or 20%? 20. 20, okay. Yeah, take the worst 20% of your account, clean up the performance there, and that's gonna have a much larger and much more efficient impact on your overall account rather than trying to reduce everything That is underperforming simultaneously. You're going to lose too much traffic and then you're going to have to try to increase all the bids on those things again because you lost all the traffic and now you're spiking the ACoS again and you're just whipsawing back and forth. This is actually something I see all the time when people churn from agencies is they're like, yeah, it's like ACoS is through the roof and then sales are to the floor and it's just like back and forth, back and forth, back and forth because they're always trying to optimize everything and over-correcting and causing lots of problems there. Anything to add to that one? Speaker 2: No, let's keep it moving. To the third thing, so a lot of times with these rule-based automation platforms, They are very limited in what they allow you to do, how you can set up the automation. So most of the time, it looks like you can either increase or decrease by X percent based on the current bid or the CPC, or you can pause it. So there's a lot of limitations. You can only do so much. There's no formulaic bidding. You can't do revenue per click in a lot of these platforms that I know of. And they just pretty much allow you to only use that one data point. And go up and down from that. And that's not dynamic either. It's gonna be the same change across all of the keywords. If you're setting a 10% decrease, it's gonna be 10% no matter how bad the ACoS is or how high spend this target is. It's gonna really run into a lot of problems. Stephen, what do you got for me on this one? Speaker 1: Yeah, so this is the main issue and this is the issue we're basically going to camp on for the rest of this episode is saying why that increase-decrease by X percent method just simply doesn't work because that is the by far prevailing method for managing Amazon. It's literally like, I mean, I would venture to say it's how over 95% of PPC managers are optimizing their accounts if you include also people using AI because that's all the AI is doing too. So I would put maybe 99%. You know, Andrew, we had a t-shirt that was saying bid like the 1% that Andrew made, which is kind of, yeah. And speaking of t-shirts, you guys might have noticed the first official Adlabs shirt ever made We just bought it, so I'm wearing it because it just came in the mail yesterday, but we got team merch. Speaker 2: Looking sharp, man, for sure. But to your point, I was working at that agency we used to work at and this guy got hired on. He had some experience doing Amazon PPC and he came in and we had been using Adlabs to optimize the bids and all this type of stuff. And I was out and you were out. And he came in and set up this whole rule-based automation system, very intricate, very in-depth, covering all these types of things. It just didn't work and they were constantly trying to figure out what was going on, what the problems with performance were. And it was just because he had all these things running at the same time and really kind of threw things off. So I would say yes, it is pretty much the prevailing wisdom to utilize rule-based automations, plus or minus X percent, inching your way up and down towards the target. And like you said, a lot of automation tools do that as well. And for all the reasons we've already listed, those tools, Generally can get close to target ACoS, but there's a lot of inconsistency or it takes way too long in order for those tools to actually get on target because there's all this slow incremental adjustments that are happening. Speaker 1: Or they're doing it, maybe they're not even doing automations, but if you're doing increase, decrease by X percent just in the ad console or on bulk sheets, you're still going to run into some problems there. Kind of going back to what I was saying previously, though, or what Andrew was starting to say previously, is the limitations with these rule-based automations is just increased decrease of X percent. What would be great, sometimes they get a bit more complex and it's like you can actually set the bid to the CPC times a variable. And so, okay, that's kind of cool. Now we can get a little bit more sophisticated, but that CPC is Can be way off because you might only be bidding 50 cents and the CPC was a dollar, but that was because you had some top of search adjustments in there, right? So that CPC is like the actualized CPC after all the placements are accounted for. So if you're trying to, you can't really factor in the base bids for the keywords to account for the underperformance of certain placements. And then you can have some other rule-based automations that are increasing, decreasing placements, but we'll talk about that in a second for why that doesn't work. So you're just limited in the logic. Like you can't say, for example, the high spend no sales scenario. The correct answer should be, what's the high spend condition? If the spend is overspent on a keyword, if the spend is overspent, the target cost per acquisition, which is essentially the average order value of that product times the ACoS, the target ACoS. If you have a $10 product with a 30% target ACoS, then after $3 of spend, now you're overpacing that target cost per acquisition. So that's the formula you should be using. But these rule-based automations don't allow you to say if spend is greater than average order value times target ACoS, then calculate the bid because you don't want to just kill that keyword. It's hopefully relevant if you've been harvesting relevant keywords. Then that bid should be, okay, calculate it based on how many clicks we need on average to get that sale. And then, you know, given the total number of non-converting clicks plus the anticipated number of clicks to convert, that should be what the CPC is or the bid should be. And then you're also factoring in the placement settings. Like, that's the logic that's needed to properly optimize that keyword. You gotta pull in some different reference points. We're usually pulling it from the ad group level, assuming the ad group has that data confidence. If not, we're looking at the campaign level or group of campaigns or the campaigns with data. All of that information is how you get an accurate bid. Just saying decrease the bid by X percent is... It's way too simplistic. Now, it's going to also sound, go ahead, Andrew. Speaker 2: I was just going to say that really brings up the real crux of this whole conversation because it's not a cohesive system, right? You're optimizing pieces of it here, pieces of it there, and they're not really talking to each other. They're not referencing the placement modifications and things like that. It's like you're optimizing bids in a silo. You're optimizing placements in a silo, all that type of thing. So there's no cohesiveness between all the different components that actually make up what the final effective bid is. Speaker 1: And even those like spend thresholds and everything like they're all, all of those should not be fixed because every single week of the year has some different flavor to it, some different level of seasonality or search demand or competitors going in stock out of stock. Everything is a trend, right? So The conversion rates are fluctuating. So that's going to influence everything with these calculations, with the high spend on converting, with high ACoS keywords, all of that's constantly changing. The spend threshold is constantly changing, especially as you're shifting your goals around. So you really need something more dynamic. You can't just have these rigid percent changes for a rigid look back and then just say, optimize the whole entire account like that on a daily basis. Things are going to break. So now let's talk more about why the increase decrease by X percent does not work. And I'm going to give you some examples here. So this first one is hypothetical, just for easy math. But let's just say you have a high ACoS keyword. So let's pretend you're targeting a 50% target ACoS. And for the last 30 days, this keyword's ACoS is 100%. So what people typically do is they say, okay, well, if my target ACoS is half of what the current ACoS is, it stands to reason that I should reduce my bid by 50%. Technically, you are correct if you're thinking you need to reduce your CPC by 50%. So assuming that the average order value and the conversion rate stays the same, then yes, having a 50% lower CPC compared to what you've had in that date range would get you there. However, we're not looking at the CPC in this case. People are only looking at the bid. So they'll just reduce the bid by 50% from whatever it was, regardless of if that bid was at $100, And even though it was at $100, you were the highest bidder, you were only paying like a $1 CPC anyways. But when you go from $100 down to $50, you didn't actually change the bid at all. So you have one problem there. Or maybe that bid was already reduced to like two cents and you can't just keep reducing it anymore. So if you do reduce this bid by 50%, which a lot of these tools, let's just say, you set up the automation, you say if ACoS is over 100%, reduce the bid by 50%. You come back the next day or the next week, however frequently this is supposed to be running, if it looks back at the last 30 days, let's just pretend that for the last 7 days, since you made that optimization, you lost all the traffic, you now have zero traffic because you brought that bid from $1 to $0.50, so no more traffic anymore, but when it looks at the last 30 days, You still have 100% ACoS. And so it's going to try to reduce the bid again, another 50%. And then you're going to go down to a 25 cent bid. And then a week later, it's going to look at it again. And assuming that you still had some of that high ACoS on like that beginning period of that 30 day look back, now it's going to cut that bid in half again. And it's going to go out to 12 cents or 12 and a half cents, 13 cents, whatever. So you're going to run into some problems there. Speaker 2: That is exactly what I see on a regular basis. Basically, people are stepping themselves out of auctions completely. What you really need to do is use a shorter timeframe in those scenarios and adjust that date range to give it an up. A clear representation of how that keyword's actually performing since the last optimization. Like you were saying, as you're continually referencing that same date range, you're still just using that old data and making decisions based off of that rather than the data since your most recent optimization. So yeah, I've seen that quite a bit, especially people coming from automated solutions. Speaker 1: Yeah, but when you're going for a shorter timeframe, That does come with some trade-offs. You're now having way fewer clicks that you're optimizing from, where the last 30 days, all of that, whether it was low conversion or whatever, that's all relevant data that should still be factored into consideration. So what people will usually do is, like Andrew said, take a shorter timeframe, which does have some downsides to it, less data, not as good data. Or the other option is people will say, I'm gonna wait until a certain amount of time has passed or a certain number of clicks have passed. So I made these optimizations, I reduced the bids, whatever. And now, there's actually one tool in particular that does this, that someone was saying, oh, this is genius, you guys should do this in Adlabs, where it waits until a certain amount of clicks have passed on that keyword before it makes the next adjustment. So I don't know what that threshold is, that's one problem, is that whatever amount of clicks they're picking is arbitrary. But let's just say we're saying 20 clicks. We're gonna wait for 20 clicks until we make another adjustment. Okay, well, if it was bad, if it was a bad adjustment to begin with, We're now not optimizing for another 20 clicks. So you're either going to continue bleeding out and losing money for another 20 clicks or that first adjustment was too aggressive and or made too many reductions and now you're not getting any clicks. So you're never going to get the click data confidence that you need to make the next adjustment. You run into all these different problems with only increasing, decreasing by X percent compared to what we would recommend, which is just calculating what the bid is worth And set the bid to that amount. Those were from the first like three, four episodes that we've ever done. So if you are wondering what we're referencing, go back to those, but it comes down to like the revenue per click approach, which is just calculating the bid the right time, the right way, the right, the first time. I'm gonna give one more real scenario that this happened last week where there was a user, they were trying out Adlabs and there was a high ACoS keyword and it was showing that the bid calculation that we were recommending was to increase the bid on this high ACoS keyword. Now we do have a full other episode too on when should you increase bids on high ACoS keywords. Let's gonna dive into this if you want more information. But he was trying to say why would you increase the bid on this high ACoS keyword? He was previously using rule-based automations that would say you should reduce the bids. on High ACoS Keywords every single time, regardless of whatever the bid is. It did not factor in the current bids, it was just always reduced on High ACoS Keywords. Well, what we saw was that the historical CPC was $1, the historical revenue per click was $1, so it was 100% ACoS. And with a 30% target, ACoS, his maximum affordable CPC is $0.30. Well, the Keywords bid was $0.02. And so when you change the bid to two cents, when you use the look back window, it's still showing the historical CPC. Changing the bid down does not retroactively undo the spend from the last 30 days. So you're still going to have a high ACoS no matter how far you reduce the bid. And so any other system that's looking at that window is just going to keep saying, oh, we got to reduce the bid, reduce the bid. He's got tons of two-cent bids all over his account because this rule-based automation was, who knows how he had set it up, but however frequently it was doing it, just continued reducing, reducing, reducing the bids until he got to two cents. In reality, $0.02 is just the wrong bid. He could afford a $0.30 bid. The problem was not a $0.30 bid. The problem was not bid increases. The problem was a $1 historical CPC. So that was where the problem was. He could only really afford $0.30. That's what the right bid should be. And when you use the RPC formula, it does not matter how many times you calculate it, you're always going to end up with $0.30. Every single day that passes as you're collecting more data, you can always be adjusting that. You could run it a million times a day if you wanted to and it will constantly be recalibrating based on that 30% of the revenue per click. But that's just way better than constantly decreasing it until you get to two cents. I think that's the best example I could think of. Speaker 2: Yeah, that is such a great representation of why formula-based bidding does so much more and so much better than just plus or minus X percent. That's like probably the number one question I get in demos with people when showing Adlabs, like, oh, why is it increasing ACoS? Or we'll get like help tickets and stuff like that. Like, why is it increasing ACoS? Something's wrong with your bidding logic, but there's your answer right there. Speaker 1: That's such an important thing to know. You don't always decrease bids on high cost keywords because sometimes the bid is too low because you overcorrected at some point. And as long as you're continuing to reduce the bids, you're just going to kill the keywords. So anyways, and I think we've killed that point. We got one final example here of these rule-based automations. They'll increase, decrease, they'll take that same faulty logic that I think we just were beating a dead horse on. And they'll then apply that to placement adjustments, which is taking everything that was wrong and putting it on steroids now because Placements do not have bids. Like the default placement adjustments is 0%. They are just multipliers on top of the keyword bids. And we also have a full episode on how you should calculate placements. So you can check that out to go more in depth on this. But you can have all your placements be at 0% and they can all be high ACoS, right? So if the placement is 0%, how could you possibly have high ACoS there? Because the keywords Placements don't have bids. Keywords have bids. Keywords have high ACoS. They can have higher ACoS at specific placements, but those keywords are the CPC. That's where the actual CPCs are coming from. It's the keyword bid plus any multipliers on top of it. And these placements, because they can't go negative, If you're trying to reduce the ACoS on a specific placement by just reducing the placement to 0%, well, you can only get to 0% and then you can't go any lower there. So that's gonna be like one clear example for like why that doesn't work in addition to everything we said previously. And then the second thing is, and I see this mistake too, where there will be tools that you have all, let's say all three placements are low ACoS. If all three placements are low ACoS or low visibility, they're gonna say, okay, increase for all three placements by 25% or whatever. Okay, well, if you increase all three placements by 25%, that is the exact same effect as just increasing all the bids in that campaign by 25%. It doesn't make sense. The only purpose of campaign placements is relativity. And I need to make that extra clear. It is relativity between the three placements. So one placement is going to be the worst performer. That placement should always be set to zero percent. The other two placements are performing better than the worst by some relative amount, some relatively higher conversion rate. So those other two placements should be increased based on how much better they're performing relative to the worst. So it makes no sense to have all three placements increased because You're not taking advantage of why these exist in the first place, which is to only increase for the better performing placements. You don't want to just increase for all three of them because that's just like I said, increasing everything in the entire campaign. Yeah, that hopefully makes that clear where you have the same problems as before where it's like fixed lookbacks, limited logic, issues with frequency, and you're going to have all kinds of problems in your campaign placements when you're increasing all three placements because they're low visibility and also increasing all the keyword bits because they're all low visibility and all of a sudden Between 25% increases on all the keyword bids and 25% increases on all the placement settings, you've now doubled down and now you put 50% increases across the whole entire campaign and then your ACoS is going to spike. But then because of your 30-day look back window, it's going to take about a month until finally the ACoS is soaring on everything that then it's going to start reducing everything back to 2 cents again and 0% placements on the campaign. And then you're going to have zero visibility on that campaign and no sales. And then it's going to start The next 30 days are gonna start working their way back up. You can see where we're going with this, where you're just oscillating between like two bad scenarios rather than just using formulaic calculations to figure out the perfect bid for every keyword and the placement settings. Speaker 2: Boom. Very well said. Now, let's change gears just a little bit to wrap this episode up and talk about some areas where rule-based automation is actually preferable and it does work. I think we've mentioned this in a couple of the previous episodes, but for three areas of your account, I think rule-based automations are good and can be helpful. It's harvesting, negations, and budgets. Now, one caveat to that is that I don't think that these automations should just run I'm your host, Andrew Bailiff, and today I'm gonna talk to you about how you can set up a rule-based automation that scrapes out any search terms that have gotten more than one same-skew order and that could preview in a list for you so then you could go through and check and see which ones you wanna actually harvest into your destination campaigns or your exact match campaigns. With negatives, you could set some automation filters, whatever, to help identify any search terms that are 50-80% below your normal conversion rate so you can go through those and review them. Are they relevant? Are they not? Should I be negating these or not? I don't think you should just have anything just automatically running. And then budgets is another area where you could set some automations like if the budget is capping out early in the day and is meeting target ACoS within a certain timeframe. But now that I'm thinking about it, you would definitely want that to be a little dynamic in some capacity. So it might not necessarily work. But if budgets are, or sorry, if campaign is, is profitable, extend those budgets. Steven, anything else to add to that? Speaker 1: I would just say that the perfect world is you have a rule-based system that presents the recommendations for harvesting or negating or increasing budgets or reducing budgets based on different criteria that you've set up, these thresholds that you've set up. And it preps those for you. It's the equivalent of just having an intern or a VA that you gave some SOPs and you said, hey, I want you to, on a daily basis, go through my account, find these things, Send them to me so that I can give the final sign off on whether or not we push them through or not. You don't want them just going all the time and making these changes constantly because there's way too many things that can slip through the cracks. Amazon PPC is a game of exceptions where there are always exceptions to the rule coming up. You're gonna have irrelevant search terms that for whatever reason came up in the auto campaign and converted with a really low ACoS. But it was a weird one-off thing that doesn't mean you want to harvest it and now spend enormous amounts of money on it. At the same time, you're going to have some other search terms that are underperforming. And while they're highly relevant, you might just want to control the bids on them. You don't want to just automatically negate them. And then with budgets, you're going to sometimes have campaigns that are running out of budget every day and the ACoS looks really good and you want to extend it. Maybe for the last seven days, it was actually really bad performance and you actually are realizing you got to take care of something before you extend that budget or maybe you harvested way too many keywords on the last round. Even though it looks good in the last 30 days, this campaign, now that it's running out of budget, most of that budget from the last seven days was on a bunch of irrelevant keywords that were harvested from some automation rule and now the performance is bad and you want to double check stuff. It's always a good idea to just spend 30 to 60 seconds reviewing everything, just a quick once over, just to see if your eye can catch any anomalies and it's not a lot of time, but just that 30 to 60 seconds a week Of manually reviewing these automations before they actually go through is going to be a day and night difference with the actual performance in your account. Speaker 2: Yep, absolutely. Automation really just does a lot of the legwork for you. You still want to be the mind that's actually pushing through those changes and making the adjustments with the context that the automation can't necessarily take into account. So love that. Anything to add here before we wrap up, Stephen? Speaker 1: That's all I got. Speaker 2: All right. Well, everybody, thanks for tuning in to another episode. We'll be back next week. Make sure you like and subscribe to see the next episode. We've got a full content library out there. Make sure you go and watch the rest of those videos we have out there and we will see you next time.