English Español Deutsch Français Italiano Português (Brasil) Русский 中文 日本語

Advanced PPC: Win with effective search query management




Joel: This particular webinar is one that I'm very very excited about. I think our search advertising is sort of a core to PPC. I think this webinar, in particular, is one where we're going to learn some really good advice about how to level up our game a bit. 

There's a lot of changes going on with Google on the search side more and it has been for several years so we are all very, very fortunate to have Brad Geddes here who has been a long-time authority on advertising, on search advertising. He's going to really teach us all something new and I'm looking forward to it personally.

Like I said we have Brad Geddes, and we have Pinar and Robert and myself, Joel. By the way, so if you're watching and you're looking at the reports showing and if you feel intimated like okay I don't know how to do this, this Excel work is above me, don't worry about it. Don't any way feel intimidated. We'll have another session another time where I'll walk you through the steps and explain to you how to make pivot tables in Excel and do the analysis that Brad is explaining. 

This is actually you'll learn something that is going to be valuable not just because of the way Brad uses it, but just valuable for pretty much anything you do in PPC. Brad's going to present, we're going to pause for some Q and A between each section of the presentation. 

First off, I will introduce Pinar. Pinar is one of the co-founders of Cubics Digital. It's a search-focused digital and marketing agency based in Istanbul and Berlin, which is very very very relevant for what we're talking about here. Pinar has over a decade of experience in the search industry. 

Next up we have Robert who has worked on PPC accounts of all sizes across many industries and has a soft spot for really helping small businesses succeed with digital advertising which is amazing because digital advertising is a way to give a small company a power that was never before.

And finally, I'll get to Brad. Now Brad has been involved with PPC since 1998 which I think is about as early as anyone can go. He's a co-founder of Ad Analysis. At Ad Analysis; PPC recommendation engine used by thousands of clients and agencies to optimize the Google and Microsoft ads accounts  Brad you can share your screen and give your presentation.

Google Ads Match Type Changes Overview

Brad: What we're going to talk about today are the match type changes that Google's made, how that works in reality, the problems it's causing, how to diagnose those problems and ultimately how to fix them if you have them. The topic is a bit geeky, but we're talking about the search terms people use to find you and from a search standpoint they are sort of one of the most important fundamental things to look at in your account.

When we look at some of these match type changes, Google has changed every match type except broad in the past year, right? They started in 2018 with exact match. Before you could match to things like misspellings and stemmings and stuff. They started word substitution which is sort of going to be an important theme as we walk through some of these match type changes. 

And they made it so order doesn't matter. What has happened is we see at times words suddenly are showing for search terms that don't really describe them. Copyright logos and copyright symbols are not the same thing. Their example that they actually launched with on their blog was Yosemite camp sites and camping in Yosemite. 

In some cases, there's not a big change. New York City hotel, NYC hotel, hotels in NYC; you can probably write that 20 ways and not really change your intent. So not all these are bad. 

July of this year... Google made a huge change because everyone who had problems with exact match had moved to phrase match. We had clients who had actually paused all their exact match and were only using phrase and then Google was like well “we're going to change phrase match.” 

Words will flow. Semantic load. That's totally a Google term: Semantic load. Words with the same meaning can be substituted so we now see more word substitution. And word order is supposed to matter....there's so many exceptions, it might as well not matter whatsoever any longer. Phrase match sort of underwent a huge change and the leader of this is often a word substitution problem. 

You've got a keyword, Kenya Safari. Someone searches for Nairobi Safari. Nairobi is the capital of Kenya except usually when somebody searches Nairobi Safari there is a safari, a big zoo in Nairobi and from a travel agent standpoint, it's a very low order value. Kenya Safari...is a much higher value. The location makes a big difference. 

Someone who's searching for cheap laptops, Google says well discount, sale, affordable; they all match the same keyword. We can argue, well, the intent might be the same, but if you've got one ad group with discount laptops with the discount laptop ad, another ad group with “hey here's laptops on sale”. And you had different ads and landing pages for these. Google's often just picking which one they want to show for, ignoring your hierarchy. Now someone searches for laptops on sale, Google doesn't show your sales page, they show your standard laptop page. That's where these become problems. 

The biggest change with modified broad match was, again, word substitution changes. This can be quite extensive in how they are looking at modified broad these days to where modified broad today looks a lot like broad match back in 2007, 2008 before they went crazy with broad match. 

The root words may not be in there at all. We will see things like someone has the keyword, trademarking. They want to show everything related to trademarks and Google has showed them for logo design, which is not at all that same intent. From a searcher standpoint, what we care about, is the intent of the words a user typed in the engine matching our keywords or at our landing page? As soon as that sort of sense is broken, that's when we start to see poor results and when to take action.

If things are doing great don't fix what's not broken. It's when there are problems we need to sort of look more into this which we're going to do in a moment. When we look back at Pinar and Robert, what are some of the issues you're seeing as Google's rolled out these match type changes?

Pinar: One of the things I wanted to add is especially the language matching. We've seen that in some international accounts for example, we're targeting Germany, but the website is in English and we run English ads and we use English keywords, but then suddenly it matches, for example, leather jacket matches with leather and jacke. That shouldn't actually happen, but it happens so it makes the search term report a lot more relevant...we have to check it constantly actually.

Robert: Well I have a B2B client that they're a consulting business, they do services with Microsoft, Office365 package and so a lot of what Google treats as synonyms and when they do that word replacement, really throws the intent off. Situations where we're bidding on a server integration and they'll substitute the word services instead of server. I can see how there's a semantic root in common between these words, but the intent is just way, way off.

I can see the justification when you start looking at these huge datasets of oh yeah there's a relation between these words, but Google has gotten really loose with stuff. Their machine learning isn't picking up on a lot of those intent cues that I think as human managers you see that in a search term report and you instantly know oh that's a negative. But their machine learning hasn't figured that stuff out. 

Brad: In many ways, what we're really talking about here is fixing the mistakes of machine learning. If machine learning could figure this out we wouldn't need to have webinars like this one to fix the problems.

When is Google getting this right? Usually, when the industry has a lot of similar words or adjectives that don't affect how a user converts, then there's not a difference. Let's say you search for green Nike size 11 shoes and Google shows the exact same shoe, blue size 12, but on the landing page you can change the color and the size. It's just a little more work for the user, but it's not really affecting the conversion rates. 

Now, if, you said, on our site you can choose your size, but not your color, that's a category level. Now a user gets to a page where they search for green shoes, they see blue shoes on the page, they can't change color selection on the page, they have to go to category, find it, go back to a product. That's when conversion rates really start tanking. If there's some set maintained and the user can fix it, then that's still working.

Overall in travel, like flights; e-commerce is overall doing okay. Most people are seeing higher CPAs. In some cases, it’s due to there's more advertiser bid pressure. A natural five, six percent increase in CPA year over a year not that unusual and not necessarily always just match types as a problem. 

How Google Match Type Changes Can Cause Issues

Now there are cases... where there's not good solutions. This is a wholesale restaurant supply company. When someone searches restaurant waiter clogs, plural, that exact phrase, their checkout value is $623. If the user removes the s from the word clog, they want a single pair, it's a $91 order. They're willing to bid five times more if you include the s. 

With the new match type changes, Google's now showing them for these other words they were never shown for previously, clog shoe waiters, even shoes waiters wear. That's a phrase match by the way. Shoes waiters wear to restaurant waiter clogs is a phrase match, query match in Google's mind. 

The amount of negative keywords necessary to try to fix this where you only show for plurals versus not doesn't exist. Google does not allow that many negatives. In some cases, you now have this sort of new average order of how you're bidding to and you can't cherry-pick specific keywords. If you look at your account and you've got a thousand keywords and five really matter, which happens on occasion. If there's a lot of word substitution, you really have to go after those negative keywords which we'll get into.

If there's not a lot of word substitution then you're okay for now, but you're going to want to watch that closely in case Google starts suddenly changing the queries that used to convert really well for you.

Where we're seeing some of these problems stem from is if we're going to look at the word in green first here, they have an accurate, registered trademark with the exact match keyword trademark registration. Now when we look at the search term when someone actually searches for the word trademark registration. The exact match word registered trademark has 19 impressions. The actual exact match keyword, trademark registration has zero impressions and this modified broad match word that actually has a lower bid has 973 impressions. All the impressions in this count have flowed to the modified broad match terms even though the exact matches exist in other ad groups. 

This is where when we start looking, we may have the same search terms in our query reports, but the keywords and ads they're showing for may have shifted. That will often cause the downturn in conversion rates which is why we want to do this year over year comparison of match types to see are we seeing these problems even though our impressions by query haven't changed, where they're coming from has changed. 

And that's what we need to do on this analysis here. What it really means for you is it used to be if we were having ad serving problems we could just go to phrase match. We could say all right Google we give up, we're going to phrase. It's our sort of back up control and that's gone now because of word substitution, language substitution. 

If you didn't do the work, you were kind of lazy in your setup, you only have ten ad groups and each of them have a thousand keywords in it and you didn't really think about your keyword selection that much and you probably should have added more. You're probably getting all those impressions now. You're probably seeing a lot without doing the work. 

Now if you did the work, that's where the problems actually coming in. The more work you put in to really work on the query to the landing page relevance, that's where Google's really messing with it and so kind of like the more work you actually have to put back in. 

Duplicate queries which we're going to touch on in a second and in some cases a lot of poor intents. Overall how many match types are working for you? Do you like the changes? Dislike the changes? Do you have clients who are just killing it with these changes? How is that working, Pinar?

Pinar: We have some bigger accounts and some small accounts. For some small accounts, we don't really bother to change it that much. We use broad match modify a lot. We try the phrase match also, but we use quite a lot of exact match. 

We expand on the exact match first for the conversion rates and then, of course, it really depends on their KPIs as well. We are really still trying to figure it out at the moment because we have some huge accounts and it's really not easy because it was very well-structured and now it gets a bit messy with the close variants and the exact match. I don't know what's the golden rule for it, but I think it really is a case by case scenario.

Robert: Best case scenario with my clients has been kind of a no change. That's the best situation. It's just that things seem to be running similarly and there's no problems.

I think that especially with smaller budgets and high CPC accounts, it has been very problematic. Playing the whack-a-mole game with negative keywords still can suck a decent percentage of the budget out even if you're being very diligent and you are using a lot of negative keywords and you're trying to actively sculpt some of those queries to the right places so they see more relevant ad copy and such. I've found that it does take more time looking at search term reports, doing some of the analyses that we're going to see later in the presentation here.

Joel: I would think it would matter on how targeted your campaign needs to be. Like if you're going for a lot of long tails and you don't have so many groups and let's say you have a variety of types of searches that could work well with like few ad messages, then these changes might actually be beneficial. 

But if you're, let's say, the B2B client that Matt was talking about before, in that situation then you could really get hurt by that because people are searching for something very, very specific. I think in Google's mind maybe, this is better to get more people on board; more people to more easily get into doing ads.

Comparing CPA and Conversions by Match Type Over Time

Brad: What we're going to get into next are pivot tables from Excel. If you know pivot tables at all none of these take more than about 30 seconds to make. If you don't know pivot tables then look for the invite from Joel in a few weeks and he's going to walk through how to make some of these things and it's not hard. Once you've done this, I don't know, three times in your life, the next one takes seconds. First one can take a little bit.

The first thing we want to look at, do just a year over year comparison by match type to see how our CPAs and conversions have changed. These are never going to be perfect because you have new keywords, you pause some keywords. There may be times when yeah you had a thousand broad match last year, you have 500 now. It shouldn't line up exactly. 

We're trying to just say is “hey our exact match went from $20 CPAs to $41 CPAs. We need to go dig into what’s going on with exact match.” Or hey our modified broad went from 90 to 92. That's really not much of a difference. We're trying to just sort of narrow it down. If you've got lots of keywords and ad groups, you can't look at everything. 

Where do we narrow this down? If this were a massive account, we would then take the second layer, saying okay we need to look at exact match, but let's look at this by campaign real quick. Is this coming from one or two campaigns the big difference or is it account-wide? 

When you're doing these comparisons it's super tempting to go into Google's tool, choose their timeframe comparison, just apply the data. This is not going to always help you. 

The way this tool works is you have an initial timeframe. You're comparing to an old timeframe. If in the initial timeframe the data is zero, but in the old timeframe the data was 500 or whatever it was, Google won't put in the comparison because there is nothing to compare initially to something in the past. 

What we usually want to do is download timeframes, merge them together, then look at our analysis. It's a little bit longer. It gives you much, much better data than relying on Google's comparison tool. 

Analyzing Close Variant Data for Anomalies

Another analysis we need to do, if we say “okay our exact match is doing worse”. One of the things we then want to look at is our search term data that looks at here is the actual exact match keywords or phrase match keywords, here is their close variant data. What we're trying to see, right, the match type itself is when the query matches that keyword by that match type. The close variants are word substitutions, misspellings, words not in your account. We're looking to see is do we have a large variation between the match type and the close variant. That will tell us that the changes to these match types are affecting our data a lot. 

These are usually a couple good big-picture places to start and understanding how the match types are doing. But a big question is everyone thinks about close variants differently so let's flip. Robert, when you're looking at your query data, you're looking at variants, you're looking at close variants, you're looking at exact match keywords, what are you usually looking for?

Robert: Typically I'm usually filtering whenever I go to search term report. I always try to get it so that I'm seeing kind of apples to apples. I'm trying to look at just the exact match close variants or just the phrase or just close variant versus not so that I can understand a little bit more how diligent I need to be as I'm looking through those. 

Really what I'm looking for is first thing is I look for stuff with a lot of clicks without a lot of conversions. And same thing with lots of impressions, not a lot of clicks. Basically trying to identify anomalies. The stuff that you need to pay more attention to kind of creeps up there. 

Pinar: We started doing this especially for a big E-commerce client is we really do it weekly. We look weekly at the main campaigns that we see some trends that we catch them pretty early. We also use filters and then we use a very short period of time. If you do it regularly then the work is not that huge than if you do it monthly or something. 

Joel: I think basically if I could summarize the lesson you're giving from this in a very short statement it's that close variant isn't as close as it used to be.

Brad: For those of you who are super advanced, we can get into Levenshtein distance. It's a cool word, but all that really does is it looks at character differences between your keyword and the search query or search term. This is good for exact or phrase; modified broad or broad we expect to be at large differences. 

Finding Duplicate Keywords in Multiple Ad Groups

We've got this keyword Patagonia Tours. Google dropped the word tours and settles for just the term Patagonia. Because they dropped the word, our distance between the two is six.

This is a way to start looking at hey if we have exact and phrase match words and we see distances of five, six, seven, sometimes 15, 20 characters, that means there's location substitutions and big changes going on. It's really an easy way to identify where the big changes are happening and not staring at things that are misspellings by a single letter. It lets you identify some of the bigger things easier. 

Because what's happening...this search term Patagonia Chile has now been displayed in 17 different ad groups. When you have the same search term from 17 ad groups, your bids aren't right because your data is so splintered, your ads probably don't mirror this quite right. When we look at the comparison, we look at a month of time in 2018, we would see five impressions, zero impressions, 24 impressions. 

Right after Google made this change, like at the end of July of this year through August, the numbers shot up significantly to a point I asked Google did you change other match types and not actually get into it because they’ve become so much more aggressive in the modified broad and broad match expansions. And in this case, by the way, the word Patagonia Chile as a keyword is not in their entire account. They don't have that keyword anywhere, but they've shown 17 different ad groups for that particular query. That's how much Google's changing what they're showing for. 

We really want to look at duplicates because if you've got your really important keywords and five ad groups, one ad group probably converts worse. In reality, three or four probably convert worse than your best one. 

Your bids aren't right. You're ad testing isn't right. It causes a lot of issues. It's actually an incredibly simple thing to deal with. You download your query data, you put it in a pivot table, you say row labels by query, count of ad group is how many ad groups this particular word has been displayed from. If it's more than one, Google has shown you for more than one particular ad group for that query. That's all we're looking at. 

And we can even add like an impression filter to say if the impressions are three, don't show us anything under X number or whatnot, but this will let us see, do we have a problem where our queries are so disparate across our account that we're not bidding and managing queries correctly. 

What you should be seeing is primarily one ad group per query, occasionally two. And if that second one is five impressions who cares, probably not worth your time. It's when you start seeing across a lot of major keywords, ten, 12, 13, 14 different ad groups, Google is just basically sorting whatever they want to and you need to sort of recap this control. 

Now there are times the number should be greater than one ad group. This is a rental car company. They are in 154 different countries. They have a campaign per country. They should have the same keyword 154 times in their account. When we look, we can say, well let's do count of ad group and count of campaign. Just do a simple map that says campaign minus ad group, what's the number?  And so this is really if hey you're a plumber, you're in five cities, you have five campaigns using the same keywords, you should show five times. That’s is all this is trying to fix. 

Negative Keywords Analysis

Now the slightly more advanced analysis is when you start looking and saying okay so Google is showing our search terms in multiple ad groups, and we've seen that our conversions have sort of shifted in what ad group they're in. What we're often trying to find out is what were the search terms to that specific keyword that had the most conversions. 

Then let's do a timeframe comparison saying this query to this keyword had five conversions last month and now this search term to keyword is down to one, so a negative four. It has shifted to a different place in the account. This is getting super granular and trying to fix overall big problems when Google is really messing with your ad serving.

If they're not messing significantly with your ad serving, you don't need this level of data analysis. But it's one that can be super useful if suddenly Google is treating, like in this account, Google is treating the word trademark symbol, copyright symbol, trademark logo, copyright logo as the same word. Those are four very different things to an intellectual property lawyer. Therefore they need to go to that level analysis.

When you see hey we have this search term in three ad groups, what we want to look at is what's our conversion rate difference by ad group, what's our CPA difference by ad group and what is our row loss by ad group. In this particular search term, they organize match types by ad group. 

When this search term shows from their exact match ad group, they have a row loss of 520% with this particular search term. Now they use a different ad and landing page for the broad match. When this search term is shown from the broad match ad group, their row loss is almost 300%, but over 200% lower. They want to add exact match negative keywords to the phrase and broad match ones to force them to exact because they make so much more per click when it's shown from that ad group. 

Robert: I've found when you're thinking about negative keywords and I think the Patagonia Chile example, well, you have to put more thought into what level you're using those negatives at. If it's something where it's showing up in so many different ad groups and you feel like you're just playing whack-a-mole, really consider putting that in an account-level list just so that you can blanket the whole thing if you know it's not worth keeping around. 

And then the other thing and this is something I came across in a search term report the other day, there was a query over six months it had over 2500 impressions, but the query was over 20 words long and it's like okay I don't know how that's happening. The keyword that was getting matched to that query was like words sixteen, seventeen, eighteen. Very much towards the tail-end of it, but if you're adding negative keywords, remember that negative keywords don't register past word ten. If you're seeing a query and word eleven is the word you want to get rid of just know that adding a negative for that word won't do anything to that query. 

Joel: I have a question from Matt Gozola, he is doing some work for a dentist, for a dental surgeon. Here's an example, a lot of clients have been asking for dental implant grants, but they don't offer that so should he just simply add the word grants as a negative or dental implant grants to negative like all together? 

Brad: Well then my question first is does it convert? Because we'll see things that you don't offer that does convert. If it converts, do nothing. Now if it's not converting you don't want to show for it then yeah just add the grants as a negative, don't worry about the whole thing. 

You add the keyword to a different ad group with a lower bid and you make minus grants as a negative keyword in the first ad group so you have dental implants minus grants, dental implant grants lower bid.

Pinar: He says that we don't want it. 

Brad: Yeah. Well then just make it a negative keyword at the campaign level and you're done or the campaign negative list level and you're functioning. 

I'm going to hit the next ten slides quickly. We go to our search query data, we build our filters for how we want to add keywords and add negative keywords. 

What has been a common problem right now is someone says hey if it has got two or more conversions in the past month I want to select it, make it a keyword. That's another common workflow on top of that. This is sort of like a flowchart I've used for years except recently we've had to adjust it a little bit to say “here's our query, is it converting? -> Yes”, which means you want to make it a keyword.

We should ask ourselves is it already a keyword? Because if we don't it's easy to actually make more duplicate keywords and just sort of make the problem even bigger than it is today. Finding duplicate keywords if you already have some is the same process as finding duplicate search terms. Same pivot table, same everything, no difference. You're just using keywords instead. 

The difference is when we look at keywords one of the things we can look at that's sort of interesting is quality score differences. The same keyword in three ad groups and hey in ad group one it's a ten, ad group two it's a nine, ad group three it's a seven. What's the ad difference here? Fixing duplicate keywords isn't negatives, it's just hitting the pause button on the extra keywords. Super easy.

Now where the biggest problems are coming from. This is a company who is a rental car company. Google treats the word hire and rental identically. If you live in Europe you say car hire. If you live in the US you say car rental. Users have a highly different CTR if they search hire and see the ad says rental there are much lower CTR then they search rental and the ad says rental and vice versa. Google treats them the same. 

In a case like this you need to add the negative keyword to every hire ad group, all the minus rental terms and all the misspellings. And every rental ad group, all the misspellings of hire. In this case they had to reorganize to do this by campaign level because you run into negative limits when you try to add every rental misspelling to be two million ad groups.

Look first, see if there's a problem. Look at your data. Look at your exact match versus your phrase, versus your close variants. If they're all the same you probably don't have anything broken. If you see problems, that's when you want to dig into it, but do look for queries, look at keywords because they will just cause problems that underlie all the bid issues, testing issues in your account. 

Joel: It has been a pleasure. To everyone still watching like we said, we will post something later about the tutorial that I'll get about how to make the report and I think that's it so we look forward to the next time. 

Brad: All right. Thanks, Joel. Thanks, Pinar. Enjoy. Have a great day.

Pinar: Thanks, Brad.

Joel: You're very welcome. 

Pinar: Thanks, Joel. Thanks for everything. Bye-bye.


Check out other webinars from this series