#112 - How Long Should You Wait After Optimizing Until You Optimize Again?

#112 - How Long Should You Wait After Optimizing Until You Optimize Again? Speaker 1: After you optimize your bids, how long should you wait before optimizing again? Speaker 2: And what date range should you use? Should it only be the data since the last optimization? Speaker 1: We're answering those questions today on That Amazon Ads Podcast. 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: Andrew. Speaker 2: Stephen. Speaker 1: We got a question. From a user, an AdLabs user, this comment came through the podcast on one of our other YouTube videos from a Craig. So I'll read the question and then we're going to answer that question on this episode. So just so you guys know, you should always be dropping your questions on the comments of YouTube or on Spotify because we are reading those and answering those and sometimes even doing a whole podcast episode just about your question. This one says, Craig is asking, let's say I have an ACoS of 35%. All right, 35% ACoS. And I use a 30-day window for the first optimization with a target ACoS of 30%. All right, so trying to pull that ACoS down a little bit. I then come back in one week and my ACoS has gone down to 33% with no irregular spikes. That's good. So it's coming down, but not yet where I want it to be Do I then run the bid optimizer again with a target ACoS of 30%? And would I then use another 30 days look back window from that day? Or would I just use the last seven days of data, i.e. since the last optimization? Really good questions. Andrew, anything before we jump into answering those questions, is there anything you want to that jumped out from you from this that we want to like add any additional context or clarifiers to? Speaker 2: No, I think that's pretty clear. This is a question we get very often from a lot of people. So yeah, let's dive into it and answer them questions. Speaker 1: All right. So before we do, Two critical episodes that you should watch in addition to this episode is episode number 35, which is on date range selection, and also episode 47, which is on optimization frequency. So both of those episodes, I mean, that's kind of like the two questions that are going on in this main question, which is just like, you know, after I've optimized, does that basically change the frequency? Because we did that whole episode on optimization frequency. Basically those two episodes, nothing changes. Those two episodes are basically the answer to this, but we are going to dive in and provide a bit more nuance in this question. So that first question is basically, he's seeing his ACoS at 35%. He's pulling it down to 33% within the last seven days. And he's kind of wondering, I think two things. Number one, Was that enough of his last optimization? Did he do enough or does he need to jump in and optimize it again? So Andrew, what would you say to that? Someone's looking at the last seven days, ACoS is 33%. What do you think? Speaker 2: Yeah. So in general, I think there's a lot of context that needs to be taken into consideration here. Usually, the first thing is knowing whether those optimizations were good or not. You can usually tell pretty quickly after pushing through those changes if you're directionally moving towards what you were trying to do. If you push through changes and then in like a day or two, you notice your spend is spiking or where your ACoS is really going out of control, you know that that didn't directionally move you in the right direction. If you give it a little bit more time, I think there's a little bit of a sales attribution delay to kind of really have reliable data after those changes. So you need to allow for a little bit of those sales to come in, in order for you to kind of really diagnose and see whether or not those changes were effective. But you can usually tell pretty quickly whether those changes are moving you in the right direction or not. Speaker 1: After a round of optimizations, What is going to be very accurate is your spend trend. So if your goal was to increase spend or decrease spend, obviously usually your goal is to increase or decrease sales and increase or decrease ACoS, but obviously how you do that is going to be primarily through spend. If the goal was to reduce your ACoS and the very next day you see that spend spikes and increases, then you're going to know that there's probably something wrong there that needs some kind of adjustments. But if you are seeing that that spend is coming down and that was kind of the goal was like to lower spend and lower ACoS, like Andrew said, you're moving directionally in the right direction. I did the same thing that you said. I just said something redundant. But if that spend like let's just say gets cut in half or some drastic measure, then that's going to be an indicator that like, OK, I wasn't quite trying to make those types of changes. So, you know, that can looking at the spend is going to be your core variable to interpret if you are moving in the right direction. The sales and the ACoS, because of that sales attribution delay, were always a little bit delayed for the last 24 to 48 hours in the true sales that are coming through, sometimes even lasting up to a week. There's still some sales actualizing. If you're just ever looking at today's ACoS or yesterday's ACoS, it's always gonna look a little bit high. And so that would actually be a good thing for you to just be aware of in your account, because every account's different. You know, different average order values are gonna have different consideration windows. So if you're selling like a $10 product, most of those sales are probably happening same day. So if you're looking, if you normally have a 30% ACoS, yesterday's ACoS might be like 32, 33%, not that high. But if you have like a $200 product and you're normally at 30% ACoS, yesterday's ACoS might be like 40%, 45% if there is that longer consideration window before making that purchase. So it's important to just bear in mind that the ACoS is always going to look a little bit high for the last couple of days. And if you're just looking at a seven day timeframe, then two of those seven days, which is 28.5% of that data is We're averaging slightly higher ACoS due to that sales attribution delay. So you do need to be, you need to really take that into consideration because if he's at, if this user Craig is at 33% ACoS the last seven days, it's possible that by end of next week, This week's ACoS actually comes down to 30% just with some additional attribution. Speaker 2: Yep. A hundred percent. I see that a lot on accounts. And if you're managing an account really closely, you kind of have a pulse on that. Like, you know, within the last two days worth of data, your ACoS is like typically between a certain range. If you're shooting for 35%, maybe it's like, it's around 40 or, you know, 38, 40% on those last two days. And typically if you look at a longer timeframe and you see that it's kind of dropped down. You usually have a pulse on what the typical ranges are there and you can kind of pinpoint and see if your current ACoS is deviating from what's normal, even in a shorter timeframe. Speaker 1: And the other thing that's critical to address and ask in this case is, I don't know what Craig's optimization settings were when he pushed them through. So we did another episode that would be great for you to watch is episode 104, just came out a couple of weeks ago, but on micro versus macro adjustments. And we kind of talk about, we should probably do a whole separate episode on the, you can control like, So whenever you're optimizing, whether it's on a spreadsheet, in the ad console, in AdLabs, you're usually thinking through what's the max increase decrease percent that I'm willing to do as I'm calculating these bid changes. So if he was trying to reduce ACoS, but he set the max decrease to only 10%, Then yeah, that sounds about right, that your ACoS would only come down to 33%, which is, we would say generally you don't want to like, if you're trying to go from 35% to 30%, it's probably a good idea to like work your way down over the course of around two weeks. Cause that is a, I mean, that is a pretty big adjustment. That's, you know, you're trying to reduce it by around 14%. So, That is a good idea to kind of incrementally step it down. So like max 10% decrease this time. And then next week do another like max 10% decrease. But if it's imperative that you get that ACoS down from 35 to 30% like today, then you're really gonna wanna open up those max decrease settings. So going down to 25% decrease or 50% max decrease to basically accelerate the pulling back on that ad spend for those high ACoS keywords and just really making sure you get drilled in. That's just something else to keep in mind is the tighter, if you're trying to significantly move ACoS or spend or sales, then if you're having tighter limits, max increase, decrease limits, you're just going to need to have a lot more frequent optimizations to continually move it down as opposed to just doing like one single, you know, big sweeping change. Yeah, on that episode 104, that macro versus micro adjustments, if you're looking in the last seven days, And you are just seeing that your ACoS stayed flat or maybe even increased for whatever reason. And you think something went wrong in the last seven days. That could be a good opportunity for a micro adjustment, which would essentially mean you're analyzing just those last seven days and you're just looking for the anomaly in those last seven days. So what's like the one campaign or the, you know, two campaigns that, happens to be way off. Maybe the ACoS went down on everything else, but there was just one campaign where something was weird and it caused a spend spike and an ACoS spike just on that single campaign. And you're just gonna drill into that one campaign. It might even just be one or two keywords in that campaign. And you're just gonna do a little micro adjustment to pull that one in and get it cleaned up. Speaker 2: And this ties really, really well into how to actually select the timeframe that you should use to make those optimizations. You were talking a lot about, So if this person, Craig, ran a optimization looking at the last 30 days, and then is coming back looking at it again, we don't know what those bids actually should have been pulled down to based on the targets based on the data that's coming in. So like Stephen was saying, if you're pulling it down 10%, maybe it actually needed to be pulled back. Maybe you have some targets in there that are at like 80% and they need to be pulled back like 50 60%. One thing you can do is just go in We're here to help you turn off the limits so you can get an idea of how far some of those bids are actually needing to be pulled down. You're using 30 days of data. That's pretty high data confidence. And sometimes you need, like Stephen was saying, stepping it down, doing multiple iterations of incrementally moving towards that goal of where that keyword should actually be based on all your target ACoS and everything. If you start using a shorter timeframe in there, you're basically ignoring a lot of that historical data in the last 30 days that is highly valuable keyword data that should be fed into those calculations. And if you're doing revenue per click bidding, like we do in AdLabs, then you want to give more context, more More data to be fed into the algorithm for it to really work well. And you're not solely needing to use just the last seven days because that's the You know, the only relevant timeframe at this current bid level, like that other data is still very valuable. Speaker 1: Yeah. And I kind of get the thought process behind, um, you know, only using the last seven days cause they're like, it's a different bid. So that's like a pretty big variable to like how the performance is going to behave. So I should really just kind of focus in on that. I get the logic, but it is flawed because the variables, We're not that big. It's still the same keyword and it's the same product. And that keyword and product are going to have a specific conversion rate in their relationship that is relevant, whether it was at a $1 bid or a 50 cent bid. All of that's relevant because all of that is traffic. All of that is conversion rate. You want to include all of it. And so having that longer timeframe Still using last seven days data is going to be a good mix of like, here's some of the data since after the optimization, but also we have like thousands of clicks from before that, that are all still super relevant and accurate. And I mean, if you did want to be a little bit more biased, if you're, if you're kind of default date ranges like 30 days and you want to be a bit more biased towards what happened since that last optimization, you can do something in between. And do like a 14 day timeframe. So that will, you know, half of that data is since the last optimization, but then the other half of it is still before that optimization. Cause it's still just a lot of relevant clicks and, and traffic. So in general, um, I would say that only optimizing from days since last optimization is a bad idea unless if it's been, We've been here for 30 days since your last optimization, in which case, go back and watch the episode on optimization frequency because you're probably not... Speaker 2: You should be optimizing more. Speaker 1: Yeah. You're probably not optimizing enough. I think most people are probably optimizing like once a week or every other week. Typical. And in those cases, that's why we keep on saying, watch those other episodes because on the date range selection, we talk about shorter versus longer timeframes and there's no right answer of what's the right date range. What is the goal? What are the considerations for each of those? And once you take into account all those considerations, then you'll have your answer. Speaker 2: If you're using a shorter timeframe, you're going to try to make some of those micro adjustments on the campaigns. One thing that I would recommend, I don't know what Stephen thinks about this, but I typically avoid optimizing placements when I'm doing that. I just go in, as long as placement performance is relatively stable and is being allocated, the spend's being allocated how I want it to. I usually don't like using those seven day timeframes to optimize placements. Usually it gives a lot more sporadic results, wild results. I like a little bit more data confidence. I like longer timeframes, especially if, you know, things are consistent for the account, looking at like a 30 to 60 day timeframe to optimize those placements. So when I'm going in, I'm making those micro-adjustments. I'm usually just trying to isolate the keyword that's the problem, pull it down, not really messing with the placements, just trying to address the core issue there. Speaker 1: And then one other thing that I have heard, there was another tool that had a logic built into it that after making a bid change, It wasn't a time-based element. It wasn't like I'm gonna wait 7 days or 14 days before we optimize again. It was saying we're going to wait until we get X number of clicks before we optimize again. And I was also talking with another prominent thought leader and agency owner who was telling me He was saying this is the best logic I've ever seen waiting until, so let's just say you've got a high cost keyword, it's going to reduce the bids and then it's going to wait until it gets another like 50, 100 clicks before it optimizes again. And he was like, you definitely need to include that in ad labs. Unfortunately, this person, this individual, didn't understand the revenue per click method. And I was trying to explain it to him, but he wasn't really interested in learning. So I wasn't able to convince him to change his mind. But waiting until X number of clicks is Not great. It works if the bidding logic is simply to just decrease bids on high cost keywords. If that's the end of the logic, it's just like, oh, ACoS is high, reduce it by, you know, 10, 15, 20%. Then, yeah, if you then, the reason why this other software is doing this is because, let's just say you reduce the bids. Okay, the very next day, the last 30 days, the ACoS still looks high. So if you were running that on a daily basis, always using the last 30 days, then after you reduce the bids, the ACoS for the last 30 days hasn't changed yet. So it would reduce the bids again. And then the next day, it looks at the last 30 days, ACoS is still high. So to reduce the bids again, we do also have another episode on titled, I think, like why automation doesn't work or something where we explain this problem, where Yeah, just a simple decrease by X percent for high ACoS keywords. Looking at the last 30 days, running that on a daily basis is going to just kill the account. It's going to end up making the bids go way too low and it's going to cause a lot of problems. So waiting until you get another 100 clicks or whatever sounds good because it's like waiting until you get that data confidence again. But the problem with that. Speaker 2: What if ACoS is still high? Like you're just wasting money that whole time. Speaker 1: Exactly. So you can, you can lose a ton of money that way. Just waiting for that click confidence to come in where like, you already have the click confidence, like from the last 30 days of data, like we definitely have enough clicks to know what the bid should be. And this is why we use that revenue per click. Another episode you can watch is why the revenue per click is king. Sorry, I don't have the episode numbers on those, but you can search our channel for them and find them. But the same problem kind of applies to like maybe that when they reduced the bids, it was too much of a correction and now it has zero clicks. So now it's gonna be waiting until X number of clicks, which is not gonna happen again. So it's gonna, you're gonna have to wait basically 30 days until it hits this other condition, which is like no clicks in 30 days to start increasing it again. Yeah, if you're not using revenue per click, then you're going to have to do all kinds of gymnastics to try to make the logic, you know, work. And it's just a bad overall, it's inconsistent is what it is. It's inconsistent. It's relying quite a bit on just waiting too much and getting bad performance while you're waiting for that data confidence. And then you optimize it again. And then like resets everything in like the learning curve or whatever. In general, that's why we're saying you can optimize based off of the last 30 days. You don't need to, you know, one week later, you can jump in and optimize again using the last 14 to 30 days. Just be mindful of your max increase decrease limits that you're doing. It's totally fine to, yeah, if you're gonna use a really short timeframe, like just the last seven days, you just wanna be treating that as like a micro adjustment. You're not doing placements in those cases. You're just drilling into a few problematic keywords or campaigns and doing a small cleanup right there. Speaker 2: Yep, absolutely. Very well said. Hopefully this was helpful, Craig. And to everybody else that's listening, again, Stephen already said it, but leave your questions down below. We're happy to put together some thoughtful responses to those questions, help you out along your journey. So leave those comments, make sure you like, subscribe and come back next week for another episode. Speaker 1: That's it for today, guys. We'll see you next week on That Amazon Ads Podcast. Speaker 2: See you there. Unknown Speaker: I love you, papa. You know you're gonna find me.