You're Doing NEGATIVE KEYWORDS WRONG

You're Doing NEGATIVE KEYWORDS WRONG! 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: Now, before we begin, if you have not yet seen episode 83, which was the episode right before this one, you have to go back and watch that one first, because this episode is piggybacking and building off of that episode. And you won't get as much out of this episode if you haven't seen it. Even if you have seen it, you may want to go back and review it if it's been a while. The reason why we're talking about this today is because we see Far too many people completely butchering the keyword negating strategy and really over negating their search terms, which is essentially just draining the life out of their Amazon PPC campaign. So the number one most common mistake that we see when it comes to negative keywords is really over negating and it comes down to usually three different criteria or filters where people will just say if a search term hits any of these three criteria, Just negate it. And what's even worse is they'll flip on an automation that's just always running. And then they end up into all these problems with why are my sales dropping? So we're here to fix that today. So the first criteria that people usually do is they'll just negate anything with a high ACOS. So if your target ACOS is 30%, they're just gonna say anything that's I mean, God forbid, they say anything that's over 30%. Hopefully, they have a little bit more of a window there, but they might say something with 50%, 60%, even 100%, whatever. They're saying, I'm just going to blanket negate all of the search terms that have hit that criteria. Now, the primary issue with that is, again, from the last episode, we kind of talked about this quite a bit, but a lot of times those are actually relevant search terms, where the only issue is that those bids are too high. And the fact that it has any ACOS at all means that it does have a sale, because if it didn't have a sale, there would be no high ACOS. The ACOS would be non-calculable or 0%, however it just displays in your view or your dashboard. Those are essentially terms that have a history of sales for you that you're just getting rid of where the primary issue was not that the search term, the search term itself was not the problem, it was the bids on that search term. Andrew, any other thoughts there on that item? Speaker 2: Nope. Just make sure you watch that last episode. You said it all right there. Speaker 1: Yeah. It's a little bit of a repeat from the tail end of the last episode. The second one is high spend, no sales. So this is a big one where it's wrong for a lot of reasons. But most of the time, the people, what they're using as their high spend threshold is wrong from the get go. So they might say anything with $50 or $100 of spend and no sales, I'm just going to get rid of because that's really high. Okay. Where did that $50 and $100 come from? For most people, that came from just, well, that's a lot of spent. That's a lot of wasted spend. That's the end of the logic. Whereas, let's just say your product has a $20 sales price, a $20 AOV, and even with a 50% target A cost, right, if that's, let's say you have great margins, that means you can spend up to $10 before you are overspending and this is a keyword that's becoming a problem because spend has exceeded your target cost per acquisition, which is $10 of spend to earn that $20 of sale to earn you that 50% A cost. If you're waiting until $50 or $100 to spend, you're jumping on it way, way, way too late. You should have been flagging that at $10 and been doing something about it. So then if you set your filters to then just $10 a spend with no sales and then you're just going to get all that well now you're that's probably even worse because now you're running into that same problem of potentially getting rid of a whole host of relevant keywords with good volume that the problem was not the keywords. The problem was the bids and you really just need to clean up the bids on those items. And then Andrew, what's that third final one? This one's probably the most egregious and the most offensive to our Amazon PPC campaigns. Speaker 2: One thing that you definitely should not do is just negate anything with no sales. We actually heard somebody say this to us one time that they just went through their auto campaign over the last like 60 days and selected everything that didn't have any sales and they just negate it. Blanket apply a negation across all of those. Now there's a lot wrong with that. You're likely negating a lot of really relevant stuff that maybe just got one click or it's a little bit low volume. And if your product requires You know, 10, 15, 20 clicks in order to actually generate a sale on average, then you just completely eliminated yourself from a bunch of terms without giving that keyword or that search term a little bit more time to prove itself. And so you're just completely taking yourself out of auctions and going to be nuking your sales on a lot of things. Speaker 1: Yeah. And the idea sounds really good on paper. Like I totally get why people do it because they'll say, you know, last 90 days, right? Anything that did not drive a single sale in the last 90 days, if I negate that, Theoretically, in this point of view, I will experience no sales loss is the idea because nothing in the last 90 days from what I'm getting has driven a sale. However, here's why that's incorrect. Let's say you have a phrase match keyword, right? And that phrase match keyword, let's say it has 10 clicks and one sale and you're getting a 10% conversion rate. Speaker 2: Great. Speaker 1: It's a good keyword. But on the other end of that keyword, you have 10 different search terms. Each of them are relatively low volume. And you got nine of those search terms have one click and no sale. And one of those search terms has one click and one sale. Now, maybe the next week or the next 90 days, It inverses a little bit. We're now another keyword from that list of 10 got the one click one sale and the other one got the one click and no sale, right? So they're kind of alternating. So let's say over the course of a year. Each one eventually gets one click, one sale. It's point being here that all those little volumes, they kind of balance each other out for every, for all of those one click, no sale conditions. You're gonna have a bunch of other stuff with one click and one sale and all that works together. It averages out at the end of the day pretty well. Whereas if you just negate everything, what you've done is you've lost 90% of all your traffic and at the end of the day, What sales is is a combination of total traffic and conversion rate on that traffic. So if you just destroy 90% of your traffic, Even though it was not converting in that time frame that you were looking at, all of that traffic is still contributing to the overall traffic to your listing, which is contributing to the overall sales of that listing. But you've now reduced it to only 90. You've reduced that traffic by 90%. You will see a sales drop of 90% because those other terms didn't convert right away, but they would have converted. And now you're left with only those other really, really low volume stuff that, again, kind of piggybacks off that last episode where if you're only ever harvesting These super, super low volume stuff and negating everything else. You're going to just end up with sure stuff that has a history of converting, but there's no volume behind it. And so you're just going to lose all the volume to your to your listing. So those are essentially the issues, Andrew. Speaker 2: Yeah, I mean, that falls perfectly in line with a client that we were working with that sells really high AOV items where that purchase consideration window is actually really long, maybe like six months. And so it might look like that keyword is getting a lot of spend and not converting in a certain time frame. But Amazon's attribution window is only 14 days. So if you're In a higher price point or product or whatever, you would be completely eliminating a lot of that traffic like you were saying to those products and that could really have a big impact downstream for a product that's more higher priced. Speaker 1: 1000%, dude, you are so accurate. Because something else I just realized is not only is for sponsored products, the attribution only seven days. So yeah, a lot of that traffic could have been, could lead to like add to carts and all these things that actualize outside of that seven day window. The other thing is the sales are last touch attribution. So if someone typed in one search, clicked on your product, it drove a click, they leave Amazon, they come back and search again. Maybe the keywords a little bit different instead of searching for like, maybe, yeah, they were looking for table tennis paddles. And then the next time they come through and they look like for table tennis paddle set, and they click on another ad and buy it. Well, your first search term for table tennis paddles Got a click but did not get the sale because that sale was ultimately attributed to the last click, the last search term that triggered your ad. So there's a ton of stuff in there that actually could be, that's why that traffic is important. Even if it's not attributing the sale, that doesn't mean that the traffic is completely worthless or necessarily wasted. So it can be. It can be. And that's why we're going to talk in a second here about when you should be negating what that criteria should be. So let's actually transition to that. But I also want to just say for the record, this whole episode is not at all talking about segmentation and isolation strategies, which is Basically meaning I'm trying to make two campaigns not target the same thing. That's not the topic of this episode. That's a topic for another time. We're really just talking about for reducing wasted spend or inefficient spend. What is that criteria that you should be using to negate? So let's dive into that. So when all is said and done at the end of the day when it comes down to it, the only two criteria for when you should negate A search term and we're really almost one criteria. The only criteria should be number one, it is irrelevant. If it is an irrelevant search term, negate it. And we could also throw in one extra bonus one there that it has volume. So number one, it's irrelevant. And number two, the search query has decent volume. So kind of wrapping up everything we've said up until this point, the biggest mistake we see is that people are inadvertently negating a ton of relevant queries with a lot of really good volume when instead they should be harvesting them. And what we're trying to tell you is that the only criteria for negating is really that the query is irrelevant. And in a second, Andrew is going to dive into the different filter sets you can use to help you identify those irrelevant search terms. But I want to talk really quickly just on that volume point. Why do I say that that volume matters? And it's mainly just a data hygiene kind of cleanliness thing where if you find a ton of long tail search terms, because there are a lot, there are a lot of really, really low volume, long tail items. I mean, you could literally find probably millions of low volume search terms that are not relevant in your account. And if you try to negate all of those things, a lot of those terms will never be searched again. They were like searched once and will never, ever be searched again. And if you try to negate all of those items, you're just going to end up having way too many negatives in your campaign, which isn't necessarily a problem. There are some limits with how many negative keywords you can have. I think it's like, is it a thousand, something like that. So It's a thousand, right? Speaker 2: I think it's a lot more than that. And I think you can actually get that increased if you talk to Amazon, but okay, it used to be a thousand. Speaker 1: It used to be a thousand point being you could run system limits, but where the real problem comes down to is the bulk sheets were at least at time of this recording. When you are downloading bulk sheets, you cannot choose to not download the negatives. So it's always going to download those negatives. And so Andrew, I already made this mistake, where we kind of over negated with an account, we end up having millions and millions of negative keywords across thousands and thousands of campaigns. Which basically meant that bulk sheets broke. We could not download the bulk sheets anymore because there was way too many rows because Amazon was downloading every single negative at the campaign and ad group level and Excel could not handle it. Amazon would even throw errors. They could not generate the report. So that's the reason why, you know, if that search was never going to be searched again, You don't have to negate it because it's not going to be searched again. So through the really, really low volume stuff is just because the volume is so low, it's better to just have a little bit of a cleaner account. That's just also easier if you're ever trying to review your negative keywords to be like, hmm, did I actually make any mistakes here? It's nicer to have a little bit of a shorter list of negatives to go through instead of having just millions of stuff that's impossible to search. Andrew, in terms of finding those irrelevant terms, what are the filters that you're normally throwing on in your search term tables to try to find those items? Speaker 2: Absolutely. Number one thing is conversion rate. So I'm looking for things that have a conversion rate that's 50 to 80% lower than the conversion rate that my product typically gets. I ran into this issue one time with the We're a company selling goggles and they kept showing up for prescription goggles. And like every once in a while that would like still convert, but we weren't selling prescription goggles and we were just wasting a lot of ad spend there. So I'd be looking for those terms that the conversion rate's really low on, significantly lower than the normal conversion rate for that product. Speaker 1: And the reason why that conversion rate's important is because We're not, and you'll see, we're not looking at ACoS here. I mean, you can look at ACoS. The reason why we don't like it is because let's say you average a 10% conversion rate. I've seen search terms with a 50% conversion rate and a really high ACoS. And again, piggybacking on the last episode, the ACoS is ultimately determined by a combination of conversion rate and CPC. So if the conversion rate's really, really good and the ACoS is still high, that just means the CPC is way out of control. which means you shouldn't negate this thing if anything you should be harvesting it and controlling that CPC better. So the reason why having a low conversion rate instead of high ACoS because high ACoS doesn't necessarily mean you're gonna have a ton of clicks. A super super low conversion rate by necessity requires a lot of clicks. You can't have a low conversion rate without having a ton of clicks. So that kind of guarantees a certain level of volume behind this thing and it's really not working so if you average a 10% conversion rate you want to find stuff that's Like three times or worse conversion rate or you could say like the conversion rates like 66% below your average. So if you're averaging 10% then something with a 5% conversion rate might still be okay. You might want to harvest that out and control the bids a little bit better. The CPC should maybe on average be half of what you're normally spending, but it's still decent if it's relevant. But if you start seeing things with 1%, 2% conversion rate, 3% conversion rate, those are probably pretty good candidates to take a look at and see if they're relevant or not. And again, if they're relevant, Different conversation. You don't necessarily want to negate it. You might want to harvest it, but if it's irrelevant, those filters will help you figure that out. Andrew, what about items with no sales? What are you going to be looking at typically for those ones in terms of the filter sets? Speaker 2: Whenever there's no sales, the first thing I'm looking at is how many clicks does this keyword have? And how many clicks on average does it take for a product to get a sale? So taking those two things into consideration, I might start looking for things that have exceeded two to three times that average Clicks to conversion and if they're falling in that category, they're getting clicked on a lot. They've got a lot of volume and they're not converting and those are potential candidates that you'd want to go through to identify if you need to be negating those or not. Speaker 1: Absolutely. Yeah. So setting your click threshold, you know, we have a metric that we call the ACTC, which is an abbreviation for average clicks to conversion. It's just the inverse of conversion rate. The only reason why we use that is because like If you're trying to filter for something when there's no conversions, you can't say is my conversion rate at 0% like that's not going to get you to that click threshold. So that's why we use clicks. So you take the inverse of your conversion rate. That's your average clicks to conversion for 10% conversion rate. That means on average 10 clicks to get that sale. So anything that has like 20 or 30 clicks, two to three X more clicks than your average clicks to conversion, that could be something you could look at. You could also look at those high spend no sales things. So same thing we said before, like look at what your target CPA is, your target cost per acquisition. And if the spend has exceeded those levels, those are all problem search terms, right? We're not saying you should just let those things fly. What we're really trying to emphasize is that doesn't mean you should necessarily negate them. Sometimes the solution, a lot of times the solution is getting the bids right and not negating. And if you take nothing away from this whole entire episode, this is the biggest thing and the only thing I want everybody to hear and really understand, which is you do not solve death by a thousand cuts syndrome or that problem. By blanket negating everything with no sales or high ACOS. That is not at all how you fix it. We'll do another episode on how to solve the death by thousand cuts problem. You can check out our episode on smart bid ceilings that starts to kind of touch on that subject if you're eager to jump onto it now before we release the next episode. But that's the biggest thing. People think that you solve death by thousand cuts by just blanket negating everything without a sale, regardless of the clicks. And we're saying you should at least set some click thresholds. And after all, after you set these filters, you are reviewing the search terms. If they are relevant search terms, And they're still getting really bad performance. You should harvest them out, control the bids. That's how you fix the performance. Only if they are not relevant, do you negate it. Hope that made sense. Leave comments if you have questions. Andrew, with those filters we just described, Would you automate it and have a system run through and if something has a low conversion rate or a lot of clicks in those sales or high spend in those sales, you just have this system run through the automation and negate it on a daily basis? Speaker 2: Same answer as the last episode. I like the idea of having something that scrapes out stuff that meets certain criteria and previews them to me so I can review them like you said. We're reviewing these terms for relevancy. We're just using these different filters to get us a list of things that are problems and we can diagnose what we need to do about that. Is it optimizing the bids? Is it harvesting them so we can control the bids on those a little bit more? Or is it just completely irrelevant and you can just do a quick skim through things that are falling in this criteria and figure out whether you need to negate those or not. I don't like the idea and unfortunately this is what a lot of people want is just an automation to run through it every single week and just negate anything that meets these criteria. A lot of times you're just negate, like we've already talked about, you're negating a lot of relevant things and things that you should be just optimizing. And you're mistaking what the solution to the problem is. You're not diagnosing effectively what the problem with that keyword is. It's usually a problem with your cost controls, your bids, and not necessarily the keyword. Speaker 1: Very well said. Or, I mean, we're not going to dive into this topic, but like it could be placements, right? Sometimes like that keyword is doing so bad because all the placements were on product pages. And like it was the placement that was the issue. And you just got to move this into another campaign where you can have more controls over those placements. So I completely agree with you, Andrew. The automation, oh my goodness, for last episode and this episode, the reason why that automation is not a good idea is because over the course of, let's just say over the course of a month, your sales start dropping, your ACoS is spiking. And you're trying to figure out what went wrong and you're looking at all these keywords and like these keywords like you you fix the bids but now all of a sudden like the ACoS is going crazy and you're like what's going on like and you're having to reduce the bids, reduce the bids on keywords that used to do so well. And you don't realize that like on a daily basis, you had some automation running that was negating a ton of stuff. It's impossible to track basically. When these things are running on a daily basis or weekly basis, you can't possibly know everything that's been going on all the time between all the keywords being harvested and all the keywords being negated. And it really becomes impossible to troubleshoot when performance starts to trend in the wrong direction. Versus what we are really trying to advocate for is this semi-automatic solution which is essentially the system will prep these sheets for you saying hey here's a list of keywords that are advocates for getting promoted, sorry here's a list of search terms that based on your criteria they are advocates for being promoted to keywords and here's another list of search terms which based on your criteria are advocates for being negated and then you get to review and verify are these relevant, are these irrelevant and only then do you perform the action to harvest or negate based on relevancy alone. It'd be great to have AI that can obviously read the queries and verify based on the product descriptions of your products, if they're relevant, based on the semantics used. But I mean, that's something that we're probably years down the road from doing. So between now and then, do not automate it. Get some safe filter sets, get some interns who are running these tasks for you, or stay tuned and hopefully AdLabs will have these solutions built out one day, these semi-automatic harvesting and negating solutions. In terms of frequency, how often you should be doing this, again, topic for another time. And we got one other topic that's also savings another time, but Ngram analysis, You may not have ever heard of that before, so definitely subscribe because we're going to drop a whole other episode diving deep into Ngram analysis. But Andrew, why don't you give people a teaser about these Ngrams, what they are and how it's helped you in the past or how you use them. Speaker 2: Yeah, this is super helpful for just identifying problem keywords or problem phrases. It basically takes a look at all your search terms, breaks them out by monogram, bigram, trigram, basically meaning like one word, two word, three word phrases that are, and then it gives you the performance data associated with them. So that makes it a little bit easier to spot like, okay, what are some common threads across the account where We're bidding or we're spending too much and not driving sales or we have really low conversion rates in certain places and we can diagnose from there. That gives us a little bit deeper look on it and that's kind of how I found the example I mentioned before with prescription goggles. That showed up as I was spending a lot on things that included prescription and I was like, oh, we're not selling anything prescription. So that'll be a great topic and a great episode to dive in and share. That's actually something in the masterclass that we have. A special tool that helps basically parse out all your search terms and show you this data in a really quick and seamless way. So stay tuned for that and maybe we'll give that away. True. Speaker 1: Incentive. Yeah. So and just to because maybe people don't understand what you meant by monogram, but basically Andrew has like a big list of search terms where there's no single search query that's like Causing a problem, but maybe there's a hundred different search terms that are all like one impression, one click, no sales. And they're all very different and they're all very long tail, long search queries, but they all contain the word prescription in them. So what the Ngram allows them to do is basically break up all these long tail queries into their individual words and then say, hey, out of all the words in your account in these queries, you have like 100 terms that contain the word prescription. And those terms in aggregate are getting like really bad ACOS or really low conversion rates or whatever. And then Andrew can quickly say, oh, that's not relevant, negate. And that's also something we can hopefully build out one day into AdLabs. But it takes time to build things. Make sure you subscribe because we do have a Google Sheet that can do the Ngram analysis. And like Andrew said, we might be giving that one away if we get enough subscriptions and likes and comments and all the things that make us happy. So anything else to add, Andrew, or is that a wrap? Speaker 2: That's a wrap. Come back and find out later if we're going to give that away. See you next time on That Amazon Ads Podcast. Speaker 1: Peace.