Don't Make This Keyword Harvesting Mistake

Don't Make This Keyword Harvesting Mistake Speaker 2: Alexa, play that Amazon ads podcast. Unknown Speaker: Which one would you like to hear? Speaker 2: The best one. Unknown Speaker: Okay, now playing that Amazon ads podcast. These gentlemen are completely changing the game. Speaker 1: 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, you might be asking, why does this topic even matter? You guys already covered keyword harvesting in another video. What more is to be said? Now, the reason why we're making this video is because we see that most people in the Amazon space are doing this wrong. And there's one common mistake that most people are getting completely wrong when it comes to harvesting. It's that they're overharvesting. They're harvesting way too many keywords. They'll basically take a look at their search and report and they'll look for anything with one order and they'll say, ah, that's got an order. That's relevant for my product. We need to harvest it into keyword status. We need to be targeting that directly in manual campaigns. There's a few reasons why this is wrong and this is ineffective and it's not really necessary to be harvesting every single thing that gets an order in your Amazon campaigns. Number one is that a lot of times if a keyword only has one order on it, you don't know the volume. A lot of times you'll... You can see orders on keywords that are very low volume. If you were to go search up the search volume on Helium 10, Datadive, wherever, you'll see that that doesn't even register in most indexes for getting a lot of search volume. You'll see things that get one impression, one click and one order. These are just one-off sales that typically only happen every once in a while. These are things that only get searched pretty infrequently. Speaker 2: So the problem with what Andrew just described of overharvesting and particularly in this first issue with low volume search terms or low volume keywords is that after you pull in all these items with one order that perhaps only had one impression or one click, you end up having way too many keywords in your campaigns or ad groups, which from the previous episode or an episode a few weeks ago, it's not that having too many keywords is a problem per se, But when it's a lot of low volume keywords, it can tend to result in the death by a thousand cuts scenario in which most people's bidding systems are always going to be increasing bids on low visibility terms. You just increase it over and over and over again. And pretty soon you're up to $1, $2 bids on all these low visibility items where the problem was never the fact that you weren't bidding high enough. You're just never gonna get enough clicks on those because they just don't get enough search volume in the first place. And so you just end up having on average a lot of keywords that are averaging a higher CPC than what you can usually afford. And that is the primary reason for how you end up with thousands of one-click keywords with no sales that is basically like a vampire just draining the life out of your PPC account. And while there are some tools There's only one tool that I think can actually solve the death by 1,000 cuts problem, which is AdLabs, but for most people who are doing things on bulk sheets or with any other AI software, it's not gonna catch that and you're just gonna run into a lot of problems. What's the second issue, Andrew, with overharvesting and how that might happen? Speaker 1: Yeah, if you're just going based on that one order, you could also run into the situation where you're harvesting what we call just like coincidental purchases or maybe irrelevant search terms into your campaigns. So for example, I have this client that sells peanut butter and we have this auto campaign running and it started getting sales on a bunch of irrelevant stuff. Like it got a sale on the search term bananas. And so like just cause it got that one sale doesn't necessarily mean that I need to be pulling that out and putting it in a manual campaign. It's not relevant to what I'm actually selling. It might be complimentary and we'll pick those things up in our auto campaigns and our Broad match stuff. But yeah, you don't want to be just pulling everything because you might end up with these irrelevant terms. Right. Speaker 2: So it's not something that you would want to negate necessarily because, hey, it did convert. But it's also not something that you want to harvest into a new keyword that you're really trying to invest in and grow the sales and scale on. That's not a scalable keyword. So great example there, Andrew. And what about the third one? Speaker 1: Yeah, the third one is A brand halo sale. So sometimes you'll get a click on a search term and you'll see that it has a sale, but if you actually dig into it, you look at the same SKU orders, it's not there. It actually was a brand halo sale. It went to something else. So if you're selling peanut butter and almond butter and somebody clicks on almond butter, but then ends up purchasing peanut butter or whatever, it could, uh, did I say that right? Speaker 2: I think you said it right. But yeah, if you're selling both almond butter and peanut butter, someone searches almond butter, they end up clicking on the peanut butter ad, but then they end up purchasing your almond butter. So the click, the peanut butter ad, ad group, campaign, however you have it set up, is the one that got the click and the sale attribution because the orders We are attributed to your entire catalog and the almond butter, peanut butter thing is not even that extreme of an example, but we've seen some really bad ones where people are looking specifically for a small item and they click the large item, whatever, and that large item is now continuously bidding on the keyword small, small, small, where it's just not going to drive that sale. So this is similar to the coincidental purchases, but it is different because it's more talking about, it wasn't a sheer coincidence and also the bananas thing. The fact that that was even a complimentary item makes that not as much of a coincidental. I've seen some things where it's like you're selling, it's completely irrelevant. Like I'm selling toothbrushes and someone was searching for like gifts for men. And for whatever reason, you know, like some completely random thing that's not relevant that just so happens that that person, even though the search term, your product has nothing to do with it, they just happened to also want your product. That doesn't mean that you want to promote this keyword, the search term to a keyword and continue targeting it again and again. And then that brand halo effect is somewhere where it's like, it was a different product from your catalog that the person wanted. And so, yeah, just all these things to keep in mind. When you have very poor criteria for harvesting or if you're not checking the keyword relevance as you're harvesting things over, the chances of you... Essentially just ending up with these campaigns that we see when we audit, which is, you know, really low spending campaigns, or maybe the entire account is only spending $1,000 or $2,000 a month. And yet somehow this guy has 200,000 keywords because he's been harvesting everything that's ever gotten a sale or anything that's ever gotten a click. So you are going to run into some problems there. And that's why we're talking about this episode. So with that, let's now talk about what should your harvesting criteria be. When all is said and done, there are only two criteria for harvesting a keyword. Number one, the keyword is relevant. And number two, the keyword has volume. So that is all that matters, that the keyword is relevant and that the keyword has volume. Volume is pretty easy to determine. Relevancy requires a little bit of mental power because you have to think through, is this keyword actually relevant to what I'm promoting? So we're going to talk about a few different filters that can help you We're here to help you discover those search terms that are both relevant and have volume, primarily through filters that you can use in your search and reports. So Andrew, what are these filters that we're usually looking at to basically just take a list of hundreds of thousands of search terms and drill it down to just a few hundred search terms that we can quickly scan through and see are these things relevant? Speaker 1: Definitely. So the main filters that we're using are usually Is it getting same skew orders? Like is that keyword driving sales on the actual product that you're advertising? We also look at more than one order. So usually if you're looking at the number of orders, if it's more than one, it means that that search term is getting searched more frequently. Like we said before, if it's just getting one click, one sale, those things are probably not getting searched that much on Amazon. And so just that filter of more than one order will get you to those keywords that are getting a lot of volume, getting searched frequently and driving orders. And then the third one is just the number of clicks or the search volume that this keyword has. You can quickly go and check a list of keywords in any tool out there for search volume and just verify that those are getting actual traffic. And then a fourth one that I'll add is just go search it. Like just go look on Amazon and see what is actually displaying on that detail page or on that search result page and see if the other products that are ranking there organically and are kind of saturating that page are very similar to what you're selling. That's an easy way to quickly go and check. Now, you can't necessarily go through and do that for every single keyword, so use those filters first, and that's usually a good place to start and gets you most of the way there. Speaker 2: Yep, absolutely. And a couple things to note on the same-skew orders topic. Is that, well, first, same-ski orders is really going to ensure you have a lot of relevancy there. It might be a little bit too restrictive. So, for example, if you have a green variation and a red variation of the exact same product that are both relevant for the exact same keywords, just they're different colors, and the search terms aren't necessarily looking for, they're not color-specific in their search intent, that same-ski order, if someone clicks on the green one and buys the red one, That would not count as a same-skew order. So it could be very restrictive in that sense. So you're going to be really, really relevant at the trade-off of potentially, you know, missing the opportunity to continue promoting it for, if you were just doing at least one order, you could have also captured a few more search terms. So going only for same-skew orders is going to be a very, very tight window. And that's why we'll sometimes look at having at least one same-skew order We're more than one same skew order or even you could say more than two orders. If it has at least two orders or more, including brand halo, then that's going to have a bit more confidence that this is not just a coincidental purchase or a brand halo purchase. So let's just say on your peanut butter, if you're finding that when you're selling that peanut butter, promoting that peanut butter, your almond butter has like five orders on it that are not same skew, but it's still from the brand. For whatever reason, it's driving traffic to your catalog and people are ending up purchasing from you anyway. So if something has at least, if it has a high order count, then it's worth considering. Something else is when you're limiting to only one order, whether it's same skew or brand halo orders, If it's only ever one order, you're very likely going to have a lot of stuff that's low volume, where it's one impression, one click, one order. So the reason why we like sometimes going up to two orders or three orders and everything above that is that you're pretty much never going to see something that has three orders, three clicks, three impressions. Like this search term was only ever searched Three times in the history of Amazon and all three times your product got all of the 100% click share and 100% purchase share. So, you know, the more orders you have, the more confidence you can have that this keyword has volume. Or you can just go with things that have a lot of one order items because there could be stuff that's you got one impression, you got one click, you got one order. That doesn't mean it's a low volume search term. It's likely a low volume search term, but it could just be that it's a competitive search term. You were never bidding high enough to get more impressions than that. So there's two ways to check for that volume, right? There's a lot of orders or you've got at least one order, but you've verified in the Amazon search query performance reports or in the search term impression share reports or the third party tool that this search term does actually get a decent amount of volume. What that volume means, like how many clicks that keyword should be getting or how many searches that keyword should be getting, That's all going to be a little bit relative to your account. So I would just say, you know, go through some of your best performing keywords to pull in some averages. Like what's the average number of clicks that you're getting on a keyword that is doing well for you? What's the average amount of search volume on keywords that are doing well for you? Pull those in and reference them. But you should probably be getting stuff that has at least like a hundred or a few thousand searches per month. If you find something in there and you can, you know, go through that list and you don't have to read every single search term, but you should be scrolling through the list and just be seeing if anything catches your eye. Like if you're selling a plastic product and you're scrolling through and you see some searches that have orders that are searching for like steel or titanium. You can, as you're harvesting these search terms, you can deselect those or delete those rows, whatever. Remove those. Just a quick once-over is going to ensure that you have the most relevant keywords so that when you harvest things over, you're not running into a lot of wasted spend or ACoS spikes. All of that, I think, Andrew, begs the question with what are our thoughts on automating this process? So we've determined our criteria. Let's just say for the sake of this example, we're saying it needs to have at least two orders And at least 10 clicks. Okay, so that's the criteria. Now, would you feel comfortable, Andrew, with just turning on some automation and just having this thing run on a weekly basis, pulling in everything that meets that criteria for research reports? Why or why not? Speaker 1: I would say if it's just putting it into like a preview state to where I can just see it and review it before those changes are actually applied, before those keywords are harvested into their corresponding campaigns, and I can make sure that things are relevant, things are good enough volume, then push them through, I think that's fine. I like having a tool that kind of just does the manual work for you of just scraping out the The stuff that is converting well and getting good click volume. And then I can look at that and make a decision as the manager of the account to whether or not we're going to add these keywords. So I do not like having something that just completely runs automatedly. Automatedly. Is that a word? I don't know. Speaker 2: Automatically. Speaker 1: Automatically, thank you. And just adds everything that meets his criteria. Usually those filters do a good job of weeding out a lot of the stuff that shouldn't get pulled in, but it definitely still happens. It definitely still slips through. So I like that extra stage of just a manual review before things actually get pushed through. Speaker 2: Exactly. So what Andrew's saying is, Imagine that you had some intern and you just said, hey, every week I want you to prepare me a bulk sheet that's ready to upload with all my new keywords that matches this criteria. And rather than, you don't want this person to just upload a ton of stuff, because let's just say this person doesn't even like speak English, so they cannot determine what's relevant or what's not relevant, right? So they just send you this list of basically a prep doc or a prep sheet saying here's all the keywords that have met your criteria that are I'm eligible for promotion and it's a list of like, especially when you start having really good criteria, you're not gonna have tens of thousands of keywords. Maybe when you first launch an account, but definitely not after a few months and most of the keywords are in there. Every week, you're only gonna have a few dozen, like 100 max. So it's not gonna take that long to just give a quick once over, make sure everything makes sense. And if it looks good, then you can say, okay, cool, ship it. And you are just giving the sign off, right? And so we haven't seen a lot of tools that do that. There is one tool that we've used in the past that can do that really well, which is essentially, it does exactly what we described. It was like an auto-manual process, which AdLabs will soon have, but essentially it just preps that document for your review. And then you can give that review and then approve or reject any of those items. The reason why that's so powerful is number one, it's going to stop irrelevant terms from slipping through the cracks. And number two, it will just keep you a tiny bit more engaged with your account. You have now seen every single keyword that's ever been added to your account. There's not a single keyword in there, in your entire account, that has ever been added without your express approval. And with you looking at it saying, yeah, that is a good term. And the performance difference that you will see by spending just 60 seconds on this, We'll be life-changing. Performance of your account. So I hear way too many people complaining about how they don't want to spend 60 seconds going through this and then they'll spend hours and hours diving into all these crazy black hat tactics and black magic tactics, I should really say. Things that make no logical sense but some, you know, they weren't able to identify the difference between correlation and causation and so now they're going spending hours into stuff that is more hurtful than helpful. So this is what we're trying to really advocate. Now, We also need to talk about frequency. How frequently should you be harvesting these keywords? Stay tuned. We'll do a whole other episode on that in a couple weeks. So we've got one final topic for today, which is should you be harvesting high ACOS keywords? Because you might have noticed We did not say that the filter should be low ACoS below a certain level. So let's talk about why that wasn't included and answer the question, should you ever be harvesting high ACoS keywords? Speaker 1: So if you watch a lot of videos on YouTube that talk about keyword harvesting, they'll tell you that your filter should be cap the ACoS, like 60% ACoS and below, then we can harvest it. We're going to tell you why this is wrong and why you should be harvesting high cost keywords in some scenarios. So sometimes you'll have an auto campaign, a broad match campaign, whatever, showing up on targets that aren't delivering good performance. A lot of people, what they choose to do, they'll look at the search term. They're like, ah, it's not performing. Let's negate it to clean up the performance of that keyword. This is wrong because You're basically just completely eliminating yourself from the auction on that particular keyword. And if you're targeting it in a broad match or auto campaign, it's probably relevant and you are just bidding too much basically. And the reason why you can't optimize that bid is because if you're looking at a broad match keyword, if you change the bid on that, it's actually going to change the bid across a larger selection of keywords. So you don't want to have like a really big impact on that. So what you can do is harvest that high A cost keyword, put it in a manual campaign where you actually can control the bid on that exact match manual campaign and control the bid on that keyword to product relationship. And then negate it in that other campaign where it's not performing. This will allow you to calibrate the bid on that keyword, lets you still be showing up for it because it's relevant. It's by nature already generated a sale because it's high ACOS. It's just you spent more than you wanted to to drive that sale. So it is relevant and you probably should be targeting it. You just need to be able to control the actual investment levels, which is your bids and your spend on that particular term. Speaker 2: A hundred percent. Yeah, and I cannot stress or emphasize this enough. So hopefully you've made it this far because this is very, very important. And the next episode, by the way, is going to be talking about negating criteria. And so these two episodes will go hand in hand. But when you have a high ACoS keyword, let's say your target ACoS is 30%. I see way too many people who say they go through their entire search term report, anything with 45% or 50% ACoS, they just negate, right? All of this stuff where the ACoS was too high. And they did not at all consider what we said, which is, is the keyword relevant and does it have volume? So if there's a keyword that has a ton of volume and it's a highly relevant keyword, and it's currently getting a 45% ACOS in your broad match or your auto campaign, but it's super relevant. First of all, the fact that it has any ACOS at all, that the ACOS is not like dash, dash, can't be calculated, dividing by zero, means it has at least one sale. So it does have at least one order to it. That's why there is any ACoS at all. So if it has a 50% ACoS, it has at least one sale. Otherwise, you could not calculate it. So it has an ACoS. It has a sale. It has volume. It is relevant. The only reason why the ACoS is too high is because the bid is too high, right? Your CPC, as Andrew said, and I'm just restating, has not been properly calculated for that specific query based on that query's individual conversion rate. So if you have an auto campaign, let's just say you're bidding $3 on Close Match, and you have a $3 CPC on Close Match, and Close Match is doing great, but you end up realizing 80% or 90% of all those sales are happening on really, really low funnel stuff, or maybe even some branded terms that are driving most of the sales. And then there's a few other really good keywords that you should perhaps even be ranking on, but they're not worth a $3 bid. You can really only afford a $2 bid. What would you rather have? More sales at an affordable A cost with a $2 CPC and continue working on improving that BSR, getting more reviews, generating more sales momentum or negate that thing entirely and lose all of that sales velocity and as people are also doing the same problem that again next episode we'll cover off on but like Automating all of these negative rules where you're just negating everything with a high a cost when you should actually be doing the opposite harvesting those terms. That's where you're going to start seeing just the traffic on your on your product pages starts dropping your sales start dropping and you're trying to figure out what's going on and it's because you were not. Doing what is a bit counterintuitive and harvesting those high ACoS keywords. Speaker 1: Yep. Very well said, Stephen. For everybody listening, please, please, please make sure you're harvesting those high ACoS keywords. Don't listen to the fake gurus out there. They don't know what they're talking about. You're going to get much better performance out of your account if you're targeting those things that are high ACoS. All right. With that being said, Anything else to add, Steven? Speaker 2: Just make sure you subscribe so you do not miss next week's episode when we cover the same topic, but with the criteria for negating keywords. See you next week on that Amazon ads podcast.