YT YouTube2Text

YouTube2Text → Transcripts → Managing Google Ads accounts in AI era – Adalysis & the Paid Search Association

Managing Google Ads accounts in AI era – Adalysis & the Paid Search Association — Transcript

by Adalysis · 11,268 words · 1,654 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:02Hello and welcome to our paid search
0:05association webinar. I am extremely
0:07excited to have Brad Gettys here. If you
0:09do not uh know who he is, you're going
0:12to learn now on this webinar. An amazing
0:15paid search um practitioner and owner of
0:19Adalysis. Um and today we're going to
0:22get into AI. And without further ado,
0:24I'm going to drop off the screen because
0:26we have a lot of content to get through
0:28and let Brad take over.
0:30So, we're only going to kind of get into
0:31AI, right? In in reality,
0:34account management sort of went through
0:36a split um about five years ago where
0:40you had old school who grew up managing
0:43accounts one way and you have a
0:44different group who who grew up managing
0:46accounts under recommendation systems
0:49and we're going to look at how to sort
0:50of merge the management while overseeing
0:53things. So, like I was said, Sarah said,
0:56I'm Brad Gettis. been doing this for 27
0:59years in December. So I' I've seen a lot
1:04of trends over the years, but in reality
1:08they all go through the same evolution.
1:11So like we look at bid automation,
1:13right? Bid automation started all manual
1:16and then the first automation for bids
1:18was actually Windows 2000 supporting
1:20macros and then we had ways to automate
1:24manual bids and then Google launched
1:28conversion optimizer later called CPA
1:30bidding in '07. You know what it was
1:33terrible and it was terrible for about
1:36three to four years until it had the
1:38data and suddenly wasn't. And then you
1:41know Google launched row us bidding in
1:432013 it was bad and after four or five
1:48years it suddenly became good right and
1:51this is the the problem with machine
1:53learning is it doesn't get 5% better
1:56year over year or 10% better it's it's
2:00consistently good or bad or whatnot
2:02until suddenly it has a data density
2:04level that it does great and and we can
2:07apply this to almost everything. So we
2:10think about new automation, right?
2:12Google brings out something new and they
2:15publish case studies that are clearly
2:17cherrypicked on how something works and
2:20and Google is great about talking about
2:23things in the way they wish it worked
2:25versus our reality. And you have your
2:27early adopters and testers of something
2:29new and then you see a pullback, right?
2:32People say, "Wow, this is not working
2:33for us or we don't trust this yet." And
2:36often you hear like the Google tax,
2:38right? This is essentially you spending
2:40money on a Google product that's
2:42terrible because once it hits good data
2:44it'll be great and then over a few years
2:47and it's usually a few years Google
2:49learns right and adoption is pretty slow
2:51then suddenly it's really good and
2:55that's when adoption picks up and we can
2:57put pretty much any Google feature
3:00against this and see that same thing
3:02over the years like if we look at things
3:04right now right AIAX brand new wouldn't
3:07say good or bad about whatsoever yet. Um
3:09I have some thoughts but that's that's
3:11for later.
3:13And then we think about pullback, right?
3:15People have tried this and they've said,
3:16"Uh, no." Right? Autocreated assets, um,
3:19RSAs without pinning. Um, PMAX for lead
3:23generation, right? This is still still a
3:25pullback. And then when we look at PMAX
3:27for retail, it's doing well for most,
3:30even not everyone. There's some search
3:31issues, right? Broad match and phrase
3:34match. This is interesting. There's been
3:35a pullback on some of these for good
3:38reasons. Um, you're all expansion. But
3:40then we say, what is finally working?
3:43Well, all of Google's automated bidding
3:44is great for 99% of people, right?
3:48Display, audiences, remarketing, most of
3:51video, um, it can be bitconsistent,
3:54right? And demand genen somewhere in the
3:56middle here, right? So, a as we look at
3:59adopting automation, we have to think
4:02how does it work today? What do we need
4:05to look at to make sure it is going to
4:07work for us or do we need to be cautious
4:10right now and wait for it to actually
4:12get enough data? And so, you know, when
4:15we sort of look at account management,
4:17just day-to-day account management,
4:19right? When I started, I again this was
4:21a couple decades ago, this was
4:24considered advanced. I had a task list.
4:28This is clearly not advanced, but for
4:30the time it actually was, right? And
4:33then over the next few years, you we
4:35went to
4:37project management, which is still great
4:39today, right? Or calendars. I made a
4:43calendar system 14 years ago and 10,000
4:46people by 11 by 2011 were using this,
4:49right? So, and this is how accounts were
4:52managed for years. And you had either
4:55your task, your project management, your
4:57calendar. You knew what needed to get
4:58done. And then Google said, "Hey, we'll
5:02just tell you what needs to get done."
5:04And so we have this other group of
5:06marketers who grew up with this being
5:09their homepage, right? They didn't need
5:12tasks. They didn't need project
5:14management. Google told them this is
5:15what needs to get done. And and the
5:18other groups are sort of very deliberate
5:20in what they did. But you know, one of
5:22the issues is there's a lot of dark
5:24patterns in how Google creates things.
5:27So, like if if we just glance at this
5:29recommendation, right? It just looks
5:31like, hey, we're going to make 9,000
5:32more if we spend 5,000 more. That sounds
5:34great, right? Pretty easy. That's an
5:37easy yes. And how we look and say, hold
5:40on. We're going to spend 9,000, get
5:449,000 more per week, but it's going to
5:47cost us 5,000 a day. Well, that's not
5:51good math, right? And and this is where
5:54with these recommendations,
5:56you need to know why it's being
5:59recommended,
6:00what the outcomes expected are. And
6:03while Google's good at making
6:05recommendations,
6:06they don't really tell you what their
6:08criteria is. Right? We'll often see, you
6:11know, Google will say, "Switch the bid
6:13method to, you know, CPA bidding and
6:16without a reason why." And then a week
6:18later in the same campaign, you'll get a
6:20recommendation for I switched to
6:22rorowass bidding. And then a month
6:24later, it'll be like you'll do better
6:25with max conversion value bidding,
6:28right? And you'll make it'll make
6:29recommendations that doesn't tell you
6:32why they're recommending it. Now, I'm
6:34not going to rail against Google
6:36recommendations. Um the repairs category
6:37is amazing. It's fantastic to use. It's
6:40just a different way people manage
6:43accounts. So when we kind of think about
6:46the pros and cons of recommendations,
6:48right? The pro if you want to spend
6:49more, Google will help you. Now that
6:52isn't necessarily a negative thing,
6:53right? Many people want to scale, they
6:54want more impressions. So Google's
6:57looking for how to increase, right, your
6:59reach. Now, some of these are going to
7:01be good, some aren't, right? But it
7:04doesn't take revenue into account.
7:06Google wants control of your ad serving
7:08and it ignores what many of us consider
7:13essential tasks in the system, but
7:16you're never going to say Google's going
7:17to recommend you add negative keywords.
7:20Not a recommendation, right? But we all
7:22know that it's pretty important to
7:23management.
7:25So, you know, when you look at people
7:27who have been managing accounts for
7:28usually more than five or 10 years,
7:31right? They use taskbased systems. Now
7:33the good, you don't miss anything. The
7:36bad, it takes a long time to find out
7:38what needs to get done, right? And
7:40that's the bonus of recommendations is
7:43you're focused on actionable data,
7:46right? But the rules are predetermined
7:48and not all systems can you overwrite
7:50that that sort of recommendation. So
7:53because in in many years for many years
7:56management is about control, right? In
7:58reality, paid search optimization meant
8:01we take control from Google for
8:03ourselves. That's no longer the case,
8:06right? We realize we have to give up
8:08some control. Maybe not all of it, we
8:09got to give up some control. And it's a
8:12lot more about guard rails. Now,
8:15when match types changed about four
8:18years ago and and Google got much better
8:20at rows bidding and so forth,
8:23you saw a lot of people speaking about
8:26guardrails. Hey, you need to put
8:28guardrails around your account. And one
8:31of the problems was this word was tossed
8:33about, but it was never defined. And and
8:37some of the reasons for that is a lot of
8:39it's really obvious of what a guardrail
8:41is, but that was years ago when people
8:45were more focused on control. And so
8:47what we see with a lot of newer managers
8:50is because they didn't grow up needing
8:53control, some of the inherent ways
8:55accounts are built aren't sort of
8:57ingrained in them. And so, you know, I
8:59want to talk a few minutes about
9:01guardrails. And some of these are
9:02obvious, some are not, but your biggest
9:05guard rail is actually just your
9:07campaign structure.
9:09Are you only on search? Are you on
9:11display select? What's your budget? How
9:13do you split up budgets by campaign? Are
9:15you in the right locations? What's your
9:16bid method? These are guardrails right
9:19now. I mean, they seem basics. You have
9:21to choose all these things, but they
9:23dictate a lot about how AI works within
9:26your framework. Now, where it gets
9:29complicated is in Google Ads, there's a
9:32lot of ways to do the same thing, right?
9:34So, let's say I have an account and I
9:37have 50 locations,
9:40right? That's a lot of locations. So, I
9:42can make one campaign. I could use ad
9:45customizers. You know what? Easy to
9:47update ads. I've got one ad per ad group
9:49or two ads per ad group. Um, scaling is
9:51easy. I want to change budgets. I just
9:53change a single budget. But I can't
9:58change budget by location. I can't
10:01easily set up a special by location. If
10:04a manager of one locations were say,
10:06hey, can you just spend a thousand bucks
10:08more just for us? I can't do it, right?
10:10But I have great data density and
10:13aggregation for bidding. I can go the
10:16other way. I could say 50 campaigns, you
10:18know, one per location. So I can easily
10:20manage budget. Any specials ads can be
10:24easily done by location,
10:26but I have less data aggregation for
10:28bidding and I have a much larger account
10:31to manage. So there's not a right choice
10:34here, right? It's based upon what are
10:37the goals of the account and how do you
10:40want the automation to work or how much
10:41control do you want versus the machine
10:44and we can mix and match automation. Now
10:46I I see a lot of people who think we've
10:48got to use the same bid tech across our
10:50account. Um this is a company sells
10:52electric fireplaces
10:54in the US in the northeast. Most people
10:58have fireplaces built into their house.
11:00They don't make a lot of sales in the
11:02Northeast, but they want to, of course,
11:03show up there, but they don't have
11:05enough sales to use return on ad spend
11:07bidding, so they do CPA bidding. Now, in
11:09the south, they do great. They have lots
11:11most of their sales, so they do realize
11:12bidding. For a brand, they use target
11:14impression share. Um, now what makes
11:17their sales really go up is a forecast
11:20of snow. So, they keep a campaign
11:21around. Hey, it's going to snow this
11:23weekend. They quickly change the
11:24geography in the ad. They change the
11:26geography for location. They launch it.
11:29It has no previous stat. It's been
11:31paused for a while. So, they manually
11:32bid. You can mix and match this all you
11:34want to. But your first audit, your
11:39first thing you should look at to make
11:41sure you have your guardrail set up
11:44correctly is your campaign settings.
11:49It is super common to see accounts
11:51where, hey, campaign one is Puerto Rico,
11:54US, Virgin Islands, and 213 exclusions.
11:58Campaign three is the entire US then
12:00they go back. So they have two
12:01completely different ways of reaching
12:02users but all these campaigns want to
12:05reach the same people. Um in other cases
12:08ad scheduling is Monday to Friday 9 to
12:107. Other times it's all day 7 to 10.
12:13Other days it's all day long. Right? So
12:15the ad scheduling was set up for that
12:17campaign when that campaign was built.
12:19But when the next campaign was built, no
12:21one went back and looked, are we
12:23consistent with everything else? And and
12:25so consistency across campaigns as it
12:28makes sense. You may have some campaigns
12:30obviously using different bid method,
12:32different locations, maybe ad scheduling
12:34on purpose that way. But if they're not
12:37consistent, that's usually the first
12:39step to saying this is why these
12:42keywords aren't displaying properly
12:45because the way Google does their rules,
12:47we have campaigns that are restricted in
12:49their ad serving.
12:52Now, the the most important guardrail
12:54that exists besides, you know, your
12:57budget, right, is ad group organization.
13:00And this is a simple concept that is
13:04poorly done in in a lot of accounts and
13:06hopefully most of you have great ad
13:08group organization. When we think of
13:10search, right, someone does a search and
13:13if the search term matches your
13:14keywords, then an ad can show that
13:16should be relevant to your business. If
13:18someone clicks on it, they go to your
13:20website, that should be relevant to the
13:21keyword in the ad, right? I mean, that's
13:23a pretty simple concept.
13:25And then, you know, Google introduced
13:27responsive search ads. Um and that was
13:30seven years ago I think now and with it
13:34they changed the story of ad group
13:36organization right Google came out and
13:38said give us your keywords give us your
13:41assets we'll make this work for you and
13:44and the story sounds great right so what
13:46we saw and continue to see are a lot of
13:51campaigns with many many fewer ad groups
13:54than we would have seen 10 years ago
13:57now this is actual keywords and assets
13:59from an account and you know they this
14:02is a single ad group and their
14:04assumption was if we have New York City
14:06dog friendly hotel as a keyword and we
14:09have dog friendly hotel as an RSA asset
14:12Google would of course show that asset
14:13with that keyword. Now, what actually
14:16happens is that the New York City dog
14:19friendly hotel had 19,000 impressions.
14:22The keyword the asset had 4,000.
14:26Most people who searched, you know, dog
14:28friendly hotel were either seeing Times
14:31Square Hotel or downtown hotel. And
14:34their their big stick is they're pet
14:36friendly, right? They have special rooms
14:38for cats and dogs and so forth, right?
14:40That it's a big thing to them. and
14:43Google doesn't match us up. And so this
14:45is where, you know, reorganizing to say,
14:50does every asset, and I mean every
14:53asset, work for every keyword in that ad
14:58group? If not, you need new ad groups.
15:00And and while you know this sounds
15:02that's time consuming, it is. It's
15:05evergreen. And and when I say evergreen,
15:09I mean this is a slide from 2005. This
15:13is a 20 year old slide. You can tell how
15:15bad my design skills were 20 years ago.
15:18And it's a simple story, right? If you
15:20have an ad that says John's plumbing
15:22service, serving London, call for an
15:24appointment, and you have the keywords,
15:25you know, London plumbing, emergency
15:27plumbing, overflowing toilet, and then
15:30that was a common ad group. And then you
15:32ask a simple question. If someone does a
15:35search emergency plumbing and they see
15:36these two ads, what are they going to
15:38click on? Right? It's really obvious.
15:41You're going to click on the emergency
15:42plumbing service. You search that term,
15:44your basement's flooding right now. You
15:46don't want to just call for some random
15:47appointment, right? Who works weekends?
15:49Who's here now? And and this is that a
15:52group organization that if you set this
15:55up that way 20 years ago, we see
15:57accounts that's still the same group
15:58organization 20 years later and it still
16:00works. And it's time consuming. It's so
16:03important. And if you're not sure where
16:06to start, quality score ad relevance is
16:09actually fantastic recommendation here
16:12because if you have an ad group with
16:16some keywords below average ad relevance
16:19and and some keywords above average ad
16:22relevance, Google is telling you that
16:25this ad is not going to work for that
16:26keyword. It works for these, which means
16:28it's a it the ad has potential, but it
16:31won't work for these. It needs to be a
16:32different ad group, right? And and
16:34that's what's important as we sort of
16:35think about this is we're not just doing
16:38this for keywords to ads that are great
16:42for users and and that relevance is very
16:44important for the user bases. It helps
16:46our quality score which also does a lot
16:48of positive things in the account.
16:51Now second big guardrail targeting and
16:53bidding. So
16:56when you have multiple campaigns in your
17:00account, Google goes through a hierarchy
17:03system to determine which ad and
17:07campaign and so forth will show an ad
17:10for that search. So what happens is
17:13someone does a search and then Google
17:14says how many different campaigns have a
17:18keyword that matches this query. If
17:20there's only one, that one shows. If
17:23there's multiple campaigns and there's
17:26not a Pmax campaign, then Google looks
17:28up what is the most restrictive
17:31campaign, right? That means that the
17:34campaign that serves to the smallest
17:36geography, the smallest budget, right,
17:37that's not going to show all time is
17:39probably going to get the impression um
17:41there are exceptions if one of the
17:44campaigns has an exact match keyword and
17:46one one doesn't. Google does try to
17:48favor now exact match over other match
17:50types. Um second second one would be
17:53most restricted but if a Pax campaign
17:56exists then the competition continues
17:58that's not used. So then Google says is
18:02there an exact match keyword identical
18:05to the search term. So and Google says
18:09identical means spell corrected which is
18:11a little bit iffy but it does not
18:14include singular posal synonyms. So if
18:17someone does a search and you don't have
18:18the actual, you know, word that someone
18:21searched for and you have Pmax, then it
18:26goes to the next step. Next step is you
18:28have a phrase or broad match keyword
18:30identical to the search term. And if you
18:33do and you do not have a PMAX search
18:35theme, then the the search ad shows in
18:38most accounts. What really happens is
18:40you don't have a phrase or broader exact
18:42that's identical and most accounts don't
18:44have search themes. So then what happens
18:47to the ad survey? Google gives priority
18:50to what they feel is most relevant,
18:53right? And that means Google can decide
18:56what to what to pick in your account.
18:58And if there's multiple relevant
18:59options, then Google uses, you know, ad
19:01rank.
19:03So what often happens is someone does a
19:05search and you don't have that identical
19:09word, right? So Italy tour guide does
19:12not match to Italian tour guides.
19:14They're different words.
19:16So in that case, different campaigns can
19:18start to show what the what the um the
19:21ad is. Now the reason this is important
19:23is
19:25if you look at only a search ad slot, we
19:27ignore all other ad slots, only search.
19:30When a search campaign serves an ad to a
19:32search ad slot instead of Pmax,
19:3684% of the time it is a higher
19:38conversion rate.
19:4084%.
19:42Google wants Pmax to do well. They they
19:44wanted to spend money. They want to
19:46report to Wall Street. It's, you know,
19:47how much money they're making with this
19:49new AI campaign type. Um, the person
19:51took over for Jerry Drestler. She was on
19:53the PMAX team. So, it looks makes makes
19:55the person in charge of Google ads look
19:57good when PMAX spends money. So, often
20:00PMAX may have worse results, but it
20:03still is being favored.
20:05And this even goes for brand exclusions.
20:07You know, this company, they're not Papa
20:10John's, but they're basically identical.
20:13They have brand exclusions for, you
20:14know, Papa John's and Papa John's pizza.
20:17Pmac shows for all these other terms.
20:20Many of us would consider those spell
20:21corrected. They're they're not
20:23considered spell corrected, right? So
20:25that's why this this old habit of adding
20:28your top search search terms as exact
20:30match keywords, it died with
20:32recommendations because Google doesn't
20:34recommend anything but broad match. It's
20:37still essential to accounts, right? You
20:39could add them in different in different
20:40match types, but negative keywords is
20:43also an essential thing. Still, your bid
20:46method is another guardrail. And and bid
20:48method there, I'm going to go through a
20:50math here. It's it's sort of
20:52misunderstood in how some work at times.
20:56So max conversions, right, or max
20:58revenue, it tries you to get the most
21:01possible. It doesn't care about the CPA
21:04rows. it wants the most possible
21:07regardless of CPA rows. Now, the target
21:11bid methods, you're given a goal and
21:12their goal is actually try to hit that
21:14target, right? So, that that's
21:16important. Now, what I also want to say
21:18is I'm going to go through some some
21:20stats here, but
21:23phrase and broad match are working much
21:27different than
21:30a couple years ago and are working much
21:32different than than most would predict.
21:35next week, probably next Tuesday, maybe
21:37Wednesday, we're going to publish a
21:39crazy amount of data on match size by
21:41bid method. Um, which will get even
21:43deeper into some of these things. So,
21:44look for that next week.
21:47But let's say I've got a $105 daily
21:49budget. So, max bidding just says, "Hey,
21:52I want the most conversions possible."
21:54It makes mistakes. It might miss one,
21:56but it it gets six of them at a, you
21:58know, $15 CPA. Target bidding, you know,
22:01it's saying I need to average to about
22:03$15. It's often a little over or less.
22:06It misses some, too. So, maybe it's $16.
22:08They look almost identical, right?
22:10There's not a big difference here. Let's
22:13say we double our budget. Right now, Max
22:16Bidding says, "You want all of them? I'm
22:18going to go get all of them." Target
22:20starts to say, "I I can't get some of
22:22these because they'll push my target CPA
22:24too high. So, I have to skip some of
22:26these." So now we start to see the
22:29difference of max has more conversions
22:33but a higher CPA. Now if we take this
22:36even further and we put some really
22:37expensive conversions at the end, right?
22:40CPA bidding says, "No, no, no. That's
22:42way too much for me." Max says, "You
22:44gave me the budget. I'm going to go get
22:46it." Right? And and so there's this big
22:48big difference. So as you add often, you
22:51know, broad match or phrase match is
22:52crazy expanding right now. Even phrase
22:54match, right? Google's like, "Well, you
22:56gave me budget. I can expand these match
22:58types a lot. I'm going to try to find
23:00these additional ones." Which is often
23:03why you might see max betting doing
23:04great. And then as you increase budget,
23:08the CPA scales upwards a lot. And and
23:11this is, you know, why when you sort of
23:13look at your data by match types, if we
23:15look at, you know, someone who's doing
23:18max bidding, this is a max revenue bit
23:20account. Well, broad match is like three
23:23rows. for phrase and exact are roughly
23:25one. And it's not that broad match is
23:27bad here. It's that the goal was saying,
23:30get us the most possible. So, it wants
23:32to spend the money because you gave it
23:33the okay. Now, this account's target CPA
23:37bidding. You know what? Broad match is
23:40the same CPA is as phrase match. I'm a
23:44little higher than exact. That's pretty
23:46common. Broad's got a better conversion
23:48rate than phrase match, right? So, th
23:50this is where people might say, I like
23:53broad match. aren't like broad match.
23:54It's actually about bid method, not
23:56match types.
23:58Now, last one. When you sort of think
24:00about how to again build for the
24:01machines audiences, this is kind of like
24:04the ad groups. Someone will say, okay,
24:06we're targeting people who want to take
24:08a vacation in Africa. So, Kenya Safari,
24:12um, see the pyramids or Wanda Gorillas,
24:14these are all things that people do in
24:16Africa. They make this giant list that
24:18they think is great, but then when you
24:20run it, you get stats for the whole
24:23list. You don't realize, well, Ronda
24:25Gorillas is doing terrible, but the gaze
24:27of pyramids is doing great. So, you
24:29can't adjust it based upon things. And
24:32so, you're trusting the machine to make
24:34those decisions, but you don't know if
24:36it's doing it or not. So, there's no
24:38audit of is it serving the better
24:41audiences more or less than the worst
24:42ones, right? You're sort of blind to the
24:45machine. Now we have to audit the
24:46machine. So this is where even granular
24:49audiences hey let's make one that is
24:51only you know maybe Masamar and Kenya
24:53Safari and then let's make it for custom
24:56interest duplicate the same thing make
24:59it searched for and now we get all these
25:02audiences which we can say okay this
25:04isn't doing terrible we shouldn't just
25:06you know we should just pause it or this
25:09one's doing great do we want to put more
25:11budget behind this right so you need to
25:14have that data to make those decisions
25:16you know Same thing goes for audience
25:18characteristics, right? Do we need to
25:20exclude audiences or or not? If you have
25:23a low impression share, can you remove
25:25your worst times? Right? So, Google
25:29wants to spend your budget and they're
25:31going to do within your bid framework,
25:32right? They're not going to completely
25:34ignore your bid framework and and how
25:35this works. But if you have a low
25:37impression share and you'd have your
25:40best conversions between, you know,
25:42seven and four, then you could shrink
25:45the times you show your ads because
25:48you're not you're not spending your
25:49budget or you can't have all the
25:52impressions if your budget's too low. So
25:54things like loss impression share budget
25:56tell you you need to shrink your ad
25:57serving, right? If you have zero loss
26:00impression share budget and you could
26:01spend more, then you want to expand your
26:03ad serving. So impression share can
26:05really help think you think through
26:08should you expand or contract um the
26:10impressions you're serving for
26:13by regions we do the same thing right
26:15how are you doing by region and do you
26:17have regions doing poorly that maybe
26:20this needs to be another campaign so you
26:22can split it out and manage the ads and
26:25keywords differently by a region and
26:28this includes you know interest in
26:29versus um people in a location so you
26:33can't get this in the interface. I don't
26:35know why you can pull this with a
26:36script. Um, we can see it in our our
26:38software, but you know, we see in a lot
26:42of places
26:44that are larger cities that interested
26:46in outperforms targeted location. It's
26:49commuters. They live in the suburbs,
26:52they go into the city, they search well
26:54in their city. And if you're only
26:55targeting in location, right, often they
26:59might not see it. And so this is, you
27:02know, important data to look at too as
27:03you're making your optimizations.
27:05The most controversial pinning, you
27:08know, should you pin your your RSAs and
27:10this is a math problem too, right? With
27:12machine learning, data density is a very
27:16important word, right? Machines need
27:18data to function. if they have low data
27:21density, meaning if you have an an RSA
27:25with 15 headlines and two descriptions,
27:28there's 47,000
27:30combinations.
27:32So if you ad group gets a,000
27:35impressions a month, after 47 months,
27:39four years, each combination could have
27:42one impression,
27:44right? So you know what data density it
27:46doesn't exist. So, you know, Google is
27:49good if you ha if it has enough data.
27:52With low data, Google's terrible with
27:55RSAs. And this is why they default,
27:58right, to relevancy. I mean, Google used
28:00to to promote their three Rs, right?
28:02Reach, relevancy, and and rows as you
28:06know, why you use Google ads.
28:08And this is where so relevancy, they
28:10say, well, if we don't know what works,
28:13we have to pick what's mostly related to
28:15the keywords, right? So in this case,
28:18our ads are terrible. We're basically
28:20our headline ones and twos are the same.
28:22In fact, you know, Google a few months
28:24ago, you know, came out and said, "We're
28:26going to make it so ads at times can
28:28only serve one headline. We're not going
28:30to serve two, we're going to serve one."
28:32And you, in my opinion, it's because of
28:34ads like this where the second line adds
28:37nothing and they actually perform better
28:39without the second line. It doesn't mean
28:42they don't have good assets, right? This
28:44company has some they've been in
28:45business for over a hundred years. They
28:48are a dry cleaner who picks up and
28:50delivers for free at your location in
28:52New York City, right? You often don't
28:54want to run a dry cleaner there. And
28:56from their ETA data, they knew that when
28:58those lines were displayed, their
29:00conversion rate is 357% higher than when
29:04they weren't.
29:05But those lines aren't related to the
29:08keywords. And this company doesn't have
29:10a lot of data, so Google just never
29:12showed them, right? And this is what
29:14pinning does. It's it lets you say,
29:16"Well, we know it's important that
29:18headline two um shows and it needs to be
29:22one of these two highlighted lines.
29:23Let's just pin those two lines to
29:25headline two. Let Google play with the
29:26rest." Right? And so what pinning does,
29:29it reduces combinations. Just a math
29:31issue, right? So as you pin, you reduce
29:34combinations, meaning each ad
29:37combination has better data density.
29:40Machine learning works better. Now,
29:43there's a lot of falseeness out there
29:45about panning, right? Pinning reduces ad
29:47strength. Ad strength does not affect
29:50quality score or the ad group's
29:52impressions or your ability to enter the
29:54auction, right? In fact, lower ad
29:57strength ads usually have better
30:00conversion rates and clickthroughs than
30:01higher ad strength ads. Um, so the only
30:05time it matters is if an ad group has
30:07two ads of different ad strength. The
30:09higher ad strength gets more
30:10impressions. Don't be afraid to pin. And
30:13remember, you can pin multiple lines to
30:17one place. You could say, "These three
30:19assets are pinned to headline one. These
30:21three assets are pinned to headline
30:22two."
30:24Now, the one trick we have right now is
30:27that if an asset's not pinned, it can
30:30show a site link. So, some people love
30:33this because all those site links go to
30:35the same landing page. Um if so if you
30:39have assets you do not want to be turned
30:42into site links. You can just pin all
30:44your assets in any position and they
30:45won't turn into site links. And this
30:47this line second ad here we have you
30:50know two lines to one two lines to two
30:51lines at three. If we make one more
30:53asset it can only show as a site link.
30:57So we could have additional site links
30:58we control the text for that only go to
31:01our landing pages. If you want to get
31:02more into this, we did a whole video
31:04session, which is more than I can cover
31:06right now, on testing assets, how to
31:08work with with ads and RSAs and so
31:10forth.
31:12Now, negatives, as I mentioned, are
31:13incredibly important, and they are, but
31:16one of the issues with with negatives is
31:19most accounts have more search terms
31:22than you could ever hope to read. You
31:24have 10,000 search terms, a million
31:26search terms, no one's going to read
31:27those. So, again, humans need data
31:31density, too, right? We also want this.
31:32So this is where engrams come into play.
31:35So so for instance, if I have, you know,
31:37these search these search terms and I
31:40could have thousands of these. I don't
31:41want to read them all. But what I do
31:43want to know is, hey, the word San Diego
31:45is in two queries, the word plumbing is
31:47in three queries, the word services is
31:48in four queries and understand right the
31:52words used across queries. So we could
31:55aggregate the data by those individual
31:58phrases across queries. So for in this
32:01account they're a lawn company, right?
32:04And and they keep seeing things and
32:06leisure. Now the word and leisure is in
32:10265 queries. It has a total of 99
32:14clicks. No human's going to spot this
32:16because most of the times this this
32:19search term is in the query report. It
32:20has zero clicks. But turns out when it's
32:23there, they've spent $1,000 with no
32:26conversions. And that's going to say,
32:27"What's going on?" Well, there's
32:30something called Lawn and Leisure
32:31Summit. They didn't know this existed.
32:33They had no clue. They spent $1,000
32:36advertising for a conference for a
32:38competitor they weren't even going to,
32:40right? And that's where the engrams can
32:42sort of bring this up.
32:45So, in reality, we need humans still,
32:48right? AI is great, machine learning is
32:50great,
32:52humans override the machine, and AI
32:55doesn't always understand machines. This
32:57is this is one of my favorite studies.
33:00This um research institute, they took
33:02Reese's monkeys and they had three food
33:05containers up and a monkey walked up to
33:07the food container, pressed the button,
33:09they got food. All three containers gave
33:11the same amount of food. So for the
33:13first container, they put up, you know,
33:15a picture of a dominant monkey and some
33:18logos. The second one, they put a
33:20picture of subordinate monkey and some
33:22logos. and then a pina monkey, which is
33:24a sexy monkey and and some logos. And
33:27then, you know, monkeys were free to go
33:29get food where they wanted to. And most
33:31of them went to the the pina line, the
33:33sexy monkey line. The second most to the
33:36dominant and the subordinate barely
33:37anyone went to and and even after the
33:40monkeys got removed later, the logos
33:42associated got the same number of clicks
33:45pre as previous when the pictures were
33:47there. Now, we can see how humans would
33:49sort of behave like monkeys in this
33:50particular instance.
33:52But a machine would act the opposite.
33:57So a machine would say the most
34:00efficient line is a subordinate line. I
34:03should put actually more behind that one
34:05because I'll get there faster. And the
34:07second best line is the dominant line. I
34:09should put, you know, my second most
34:10amount of data there because that's the
34:12second fastest. So the machine would
34:15choose the exact opposite of humans.
34:20Now often we say with low data accounts
34:22you're better than machine and and
34:24there's there's a reason for that right
34:26and that's this concept of prior
34:28knowledge. So this was a a a company did
34:32a study
34:34where they had this you know a game here
34:38old 16bit classic you know ' 80s game
34:41and they told people right go finish
34:43this game and you you get a reward and
34:45they also gave this to 40 machine
34:47learning algorithms and said go finish
34:49this game you'll get a reward. Now you
34:52didn't get any feedback until you finish
34:54the level. So, humans, it took about
34:563,000 keystrokes or roughly a minute to
34:58win.
35:00Only one machine learning algorithm
35:02finished it. It took 37 hours of
35:05gameplay in 4 million keystrokes. 3,000
35:08for a human to finish it because a human
35:09looks and say, "I jump over fire. I
35:11climb up and down ladders. I can't go
35:12into a blank space." Right? Easy to do.
35:15So, the game is modified to use
35:16textures. All right? So, this confused
35:19humans that okay, the black one is the
35:21ladder and and you know, obviously it
35:23was it was different. It took humans
35:25about 20 minutes to finish this game.
35:28Now again the computers 37 hours one
35:32finished. The game was modified that
35:35every time you move towards the
35:37completion of the level a reward point
35:40was sent back. Now computers everyone
35:43finished much faster than a human
35:46because it had constant data feedback.
35:48Right? So if you're managing accounts
35:50without a lot of data, your instinct is
35:53probably better than the machine because
35:55it doesn't have enough information to
35:56work from and you have prior knowledge
35:58of how marketing and how humans work,
36:00right? Machines don't have that yet.
36:04So you know things like task list, PM
36:06systems, third party organizations,
36:08they're still incredibly useful today,
36:11right? So when we sort of think about
36:14working with machines, you have your
36:16account built, right? Right? And this is
36:18just how are you going to build the
36:19account. We've been doing this type of
36:21strategy for for years. But a
36:23consideration that's come up, you know,
36:25more recently is data density. So Google
36:29seems to be using more account level
36:32data for bidding as opposed to campaign
36:34level data for bidding. They haven't
36:35said this yet. I've just seen enough
36:36instances. I I have a feeling this is
36:38starting to occur.
36:40So, but you want to think about your
36:42data density and then your bid tech,
36:45right, determines your match types and
36:47then you're going to look at your
36:48recommendations and say, "All right,
36:50what do I always say yes to?" You can
36:53automate that. What do I say sometimes
36:55to okay, we need the humans still? And
36:58then you have to audit the machine,
37:00right? This is such an important step
37:01today. So, when we think about what is
37:04auditing the machine, what does this
37:06mean, right? Well, it means, you know,
37:07if we like our recommendations, which is
37:09usually scripts or thirdparty tools, we
37:11look at it every day. It's Google's
37:13maybe weekly. We look at our KPIs every
37:16day, our budget pacing, our
37:17disapprovals, right? Often weekly, you
37:20might look at search terms and and
37:22automatic placements for negatives and
37:23bids and then monthly call score and so
37:26forth. Now, you're going to modify this
37:28based upon an account spend, how it
37:31functions, its client levels. You know,
37:33we have have some clients who have
37:35millions of audiences but only hundreds
37:37of keywords. They focus on audiences way
37:39more than search terms, right? But this
37:42is is the audit machine is those tasks
37:45that were really important to marketers
37:4710 years ago and often get ignored in
37:51users who grew up with with
37:52recommendations. So because when we
37:54think about management today
37:58right we have recommendations but it's
38:00come from whatever source you want to
38:02use now you recommendations you never
38:05use right so those you dismiss or you
38:08change the criteria to be something more
38:11useful if something you always do well
38:14automate it if you always do it automate
38:16it the majority are the it depends
38:20favorite PPC answer in that case that's
38:22that human view, right? And then that's
38:25that's your recommendations and
38:26automation section. And then you have
38:28your task list which like, okay, let me
38:31make sure these pieces are working
38:33correctly. And and that's so important
38:36today.
38:37Now, this is hosted by PSA. Not all of
38:39you are members. Um, if you want a copy
38:42of the presentation, and I'm going to
38:43show a bonus section in a second, join.
38:45It's free. There's no reason not to. Um,
38:48we're at Alysis. We're a recommendation
38:50system, workflow system, paid search
38:53management tool, one of the larger ones
38:54out there. And we stand behind what we
38:57do. We give you a 30-day free trial. So,
38:58if you're trying to make your life
39:01easier with paid search, uh, look us up
39:05and go to PSA, join it, and I'll put
39:08this on later, but this really walks
39:10through all the tasks and reasons why in
39:13much more detail free for you to get.
39:17All right. And with that, um, Sarah, do
39:20we have any questions? If you have
39:21questions, everyone, please chime in.
39:24Let's have a fun discussion on where
39:27where are we going with automation and
39:29management.
39:30>> Awesome. Thank you so much. So, we
39:32actually do have some questions queued
39:34up. I'm going to show them on the screen
39:36at the same time here. Um, so we have
39:39this one coming in from LinkedIn and
39:40it's already suggesting that we should
39:43not be adding relevant search terms as
39:45exact match. Oh gosh. No, no, no. The
39:48opposite. The opposite. If you see a
39:51search term doing well, make it an exact
39:55match keyword. Well, it's important if
39:58you only run search campaigns. If you
40:00run PMAX, it's essential. So, no, I'm
40:04the the opposite.
40:07>> Okay, perfect. That was a great answer.
40:09And so, we have Higman, long time
40:12friend. Yeah, I was like Anthony
40:14Hickman. Um, on search terms, how much
40:16do you think hidden search terms affect
40:18campaigns? And do you trust that Google
40:20is really limiting them the way that
40:23they're saying? He laughed.
40:26>> Yeah, Anthony, I know your feelings
40:27about all this. Um,
40:30I I yes, I believe hidden terms probably
40:33affect campaigns more than we think. Um,
40:36where it's kind of messy is that
40:39Google spell corrects then shows terms
40:42for search, but they don't seem to spell
40:44correct them for the search terms in
40:47Pmac. I don't know why one would be
40:49spell correct and when wouldn't, but
40:50there's they're differences. Um, do I
40:53really trust Google's limiting the way
40:54they say they are
40:56from your perspective? No. And it's
40:58because Google is so good at legal ease
41:02that whatever they say, read between the
41:05lines like four different times and
41:07maybe you're close to an actual answer.
41:10They're so good at positioning legal
41:12ease.
41:13>> I love that. That's so correct. Um, so
41:16while we're waiting on more questions, I
41:18have written down a ton here. So here's
41:20one. We already got one. I'm going to
41:22put the users ahead of myself, so I'm
41:25not going to be selfish here. All right.
41:27So, this one is,
41:30should we completely ignore ad strength
41:33and utilize three pinned headlines and
41:35two pinned descriptions based on RSA
41:37classic ETA? I'm so curious about this.
41:39>> All right. So, no. Right. But that
41:42that's so ad strength has
41:47two recommendations which are useful.
41:50Um, so when you're going in edit your
41:52ad, the make your headlines more unique
41:55shows that you probably don't have a
41:57combination of USPs, CTAs, and relevant
42:00lines. So that uniqueness is actually
42:02useful. Um,
42:05make your headlines relevant to the
42:07keywords in the ad group. If you see
42:09that, that means your your headlines
42:11really aren't describing your keywords
42:12correctly. So while yes, I ignore ad
42:14strength, those two recommendations I
42:16think are actually really useful to use.
42:19Now, should we pin everything is more
42:22about how much data do you get? So, my
42:26my favorite these days for not really a
42:29great math reason, just one of my
42:30favorites these days is to pin three
42:33lines to headline one and three lines to
42:36headline two. And that's, you know, nine
42:39combinations for Google. If you have,
42:41you know, headline 3 is displayed, it
42:43could be 36 combinations. And the real
42:46question for me more these days is do I
42:49want to pin headline three or not? And
42:51and the reason I say that is do I want
42:56the assets turned into site links? I
42:58think this is sort of a mi a mis thing.
43:00Like I know some people they've spent
43:02hours on their site links by ad group
43:04and they do not want this. So in that
43:07case you really do want to pin all your
43:08assets so Google doesn't overwrite your
43:10site links. Right now you could pin to
43:12any position. Um, if you're like, "Hey,
43:15I hate that we have to make landing
43:16pages and spend so send stuff to about
43:19pages and stuff," then these are
43:21absolutely fantastic, right? Because now
43:23you have more site links going on your
43:24landing page. So that's kind of the
43:26headline three is really more about site
43:28links than it is about pinning.
43:31>> Uh, and along the same lines, although
43:34changing it to expected clickthrough
43:36rate and quality score, do you find the
43:38pinning headlines affect those metrics?
43:40So no, it does not affect them. Expected
43:44CTR right now to me is broken.
43:46>> Um I look at at expected CTR and there
43:51are times I don't get it. And before
43:54like I understand how it's done. I
43:56understand the math behind this.
43:59There's times I I think Google just
44:01can't figure out with so many
44:03combinations what quality score should
44:06be.
44:08>> Yeah. No, I These are so interesting
44:10because you've really dug deep here in a
44:12lot of these questions. Very similar. So
44:14now, how much does quality score affect
44:16the ad strength? And
44:17>> no, unrelated. They're unrelated things.
44:20You can have an ad strength of low and a
44:2310 quality score. You could have a three
44:25quality score and a whatever the top ad
44:27strength is, high or whatever it's
44:28called, I don't remember. Um, they're
44:31unrelated, have no bearing on each other
44:33at all.
44:35>> Okay. So interesting. So this sort of I
44:38want to go back to something that you
44:40said before more questions come into the
44:42chat. So uh one thing you said and I
44:45happened to have this conversation with
44:47another PPC practitioner is you made a
44:50comment about how um I wrote it down
44:55account level data versus campaign level
44:57data. And I find that one really
44:58interesting because I had a conversation
45:00yesterday about this. Um, I've always
45:03believed that to be true with the
45:05account level data. And this actually
45:06had to do with the um AI Max, which I
45:10know we didn't get into, but I said that
45:12if I were implementing AI Max, I would
45:15start it on a new campaign where I could
45:17control the budget because I felt like
45:20that campaign would learn from the
45:22account as a whole and I wanted to
45:23constrain budget before I went into AI
45:26Max versus this practitioner felt like
45:29you should apply on an existing campaign
45:31to take into account the campaigns
45:33campaign learning. So dig in there.
45:36>> Yeah. Okay. So historically,
45:39>> Google has used account level data for
45:42things like quality score. Um they've
45:45used it for various things. It wasn't
45:47used from bid purposes. now and and
45:51right now I believe they still say that
45:53but I've seen too many examples of
45:58companies with 5 to 10 conversions um
46:01per campaign using target CPA bidding
46:04which historically the that combination
46:06doesn't work well
46:07>> but they've got 10 or 20 campaigns and
46:10they're all hitting their goals or close
46:12to all because CPA is never quite right
46:16and I've seen more of these in the past
46:19three or four months than probably the
46:21past, you know, two or three years,
46:22right? To where
46:25I think Google is using more of its own
46:27data to learn, which I think is
46:28fantastic, right? I actually I'm a big
46:30fan of this. I think this is great.
46:32>> Um I I can't prove it. I don't know how
46:34you could possibly prove this, but I I
46:37feel that it's definitely happening. So
46:39to your AIAX question
46:44I the question becomes are you testing
46:46AIAX to understand
46:50if you use it will you expand and get
46:53more
46:54>> or to manage existing because if it's to
46:57expand and get more you'd want a new
47:00campaign so that way based on Google's
47:03hierarchy rules it wasn't serving over
47:05your existing campaign it would truly be
47:07net new
47:08>> correct Right? If you're just testing,
47:10hey, can we turn this on and it does
47:12stuff for us? Then you'd probably be
47:14existing. So, it's different goals of
47:15what you want out of the testing.
47:17>> That absolutely makes sense. Um, and
47:20apparently Higman liked that question
47:22too. So, here's another question from
47:25our audience. Does it make sense to do
47:27search campaign combined with audience
47:29targeting only option? For example,
47:31targeting real estate in Utah with
47:34related keywords to people only using
47:35apps that are common for medical
47:37doctors. I personally do this all the
47:39time to constrain.
47:40>> I love it. Right. Yeah. Yeah. I I if you
47:42have enough data, right? Now, obviously
47:44there are times you can't there's not
47:45especially like if you want to do true
47:47rsa which is your audience list and
47:49>> yes
47:49>> you get a thousand visitors a month you
47:51can't do it. It it will won't get any
47:52impressions. Yeah. If you absolutely
47:55absolutely
47:56>> thank you and this gets into spending
47:58your budget properly right that's going
48:00to be better more budget there less
48:02budget non audience right great yeah go
48:04for it. Uh yes, I would agree there. So
48:08another one from Higman here is do you
48:11have workaround? Of course for hidden
48:13search terms.
48:15So there used to be two workarounds.
48:17They're both gone. Um
48:20I
48:22you might I mean the only thing that I
48:24haven't tested recently to see if it'll
48:26work is value track settings. There may
48:29be something in value track that that
48:30triggers that. outside of value track.
48:33No. Um, universal used to show more than
48:36Google ads did. Universal analytics,
48:37that's gone now. Your MCC query reports,
48:41again, Google sunset of those like five
48:42years ago, gone. So, I don't know any
48:45good workarounds except maybe value
48:47track. I don't know if that'll work or
48:49not.
48:50>> Um, and then another question for the
48:52audience and people are thinking Anthony
48:54for the hard questions. Um, does account
48:57level learning um, only apply to search
48:59network or display campaigns as well?
49:02>> It's a great question. I have no clue.
49:04Um, display is so inconsistent
49:08that
49:09my guess is it's not display because you
49:11rarely have very similar display
49:13campaigns are usually very different.
49:15But I I have no idea the answer to that
49:17question. I just know display,
49:19especially video campaigns seem to be
49:21like fantastic one week, poor the next
49:24week, great the next week. They seem to
49:25eb and flow so much. Um, I'm not sure.
49:29Yeah, and I don't I doubt it, but I
49:31don't really know.
49:33>> Okay, perfect. No, that's a really I
49:35wouldn't have known that answer either.
49:36So, I guess I want to go back to also
49:38sort of a a topic with keywords. So, one
49:42of the things you were talking about is
49:44if you find the keyword adding it into
49:47search campaigns. So, when I was looking
49:51at KMAX or Kmax, what am I saying? PMAX
49:54campaigns in one of the accounts I was
49:56auditing. Um, one of the things I
49:58noticed, um, and it has to do with I'm
50:01going to call it for easy sake the DSA
50:05component, but I also know from looking
50:07at this client that they did not have
50:10pages on this topic. So, I found it very
50:12interesting in the PMX search campaigns
50:15and it was also a restaurant. I'm not
50:16going to give away the client. So, the
50:19people are looking for like grilled
50:21cheese restaurant, tomato soup
50:23restaurant. I mean, it was really
50:24specific, clearly having some type of
50:26craving, but if I had really gone into
50:29that rabbit hole and started creating
50:31campaigns around every possible menu
50:32item, you can imagine where I'd be going
50:35with that.
50:36>> Okay, so odds are, right? Let's let's
50:40Maria Corkran, who no one ever hears
50:42from anymore. She wrote an amazing
50:44article 15 years ago, right, or
50:47something on
50:49how search partners
50:52are sometimes not search sites and
50:57you'll have a an account a company who
51:01has a massive website
51:03>> and Google looks at the URLs or their H1
51:06tags to serve ads, but they're big
51:08enough to be search partners, not
51:09display, but the ads are triggered off
51:12of, you know, title tags or URLs or so
51:14forth and that is your search term in
51:18Google. So that's why when you see like
51:21did a thousand people really search for
51:22O748 Samsung 36 inpi
51:27TV? No, they went to a URL or a search
51:31partner and that's what triggered the
51:33ad. So to your example there, I would I
51:37would take a look at the search
51:38partners. Okay.
51:39>> Because I doubt it's search
51:42>> genius. Okay. Okay. That's a really good
51:44and in a lot of ways that ties back to
51:47what I know about old school SEO, right?
51:50Which makes sense because Google is
51:52crawling.
51:54>> All right. I mean, they're crawling to
51:55for DSA. Okay. So, um, audience is is
51:59hot with some questions. I'm going to
52:01put another one up here. Um, and this
52:03comes from Aaron, and she wants to know,
52:06"I'm getting more near me. How are we
52:08dealing with search terms uh that end in
52:12this?"
52:13>> Eric, I I hate this query so much.
52:16Google invented this query. And you know
52:19what? All of you have terrible quality
52:20scores for it because you can't write
52:22ads near me. You put in a geography, but
52:24the geography isn't near me. Everyone
52:26has the same problem. Your quality
52:28scores are terrible. There is no
52:30solution to this problem right now. I w
52:32if you find one, please post it on
52:35LinkedIn. We will all share it, every
52:37one of us, because it is an awful
52:39problem near me converts well for most
52:42people, but they can't write good ads.
52:44The quality scores are terrible. It
52:47looks bad, but it works and it's all
52:50Google's fault for advertising with this
52:51term.
52:53>> That's a really great point. Um, so this
52:56is a slightly different question. Um,
52:58are you gonna publish?
53:00>> Not anytime. It would be like 1500 pages
53:03now, right? The last one was what, nine,
53:05800 pages? Like, it would be insane.
53:08>> And things change too fast.
53:10>> True.
53:11>> They just they change too fast right
53:12now.
53:13>> Wow. Questions are coming in so fast
53:15now. Um, do these concepts apply to
53:17Microsoft Bane?
53:19>> 99% of these do. Nava, if you're here,
53:22I'm sure you'll put out some article of
53:24like which ones don't apply. So, I'll
53:27leave that to Nava to answer, but for
53:28the mo conceptually, absolutely yes.
53:31>> Yeah, I love that. I love that. Um,
53:33here's another one. Have you had success
53:35using PMX for Legion? I seem to get a
53:37ton of junk even though I So, he's doing
53:40the right thing with uploading offline.
53:43>> Yes, but it's inconsistent, right? So,
53:47some people are doing well with PMX for
53:49lead genen. Um,
53:53now what I would caution are a few
53:55things. one, if you're a local business,
53:58they're often driving directions off of
54:00maps. And so, it's not that PMAX is
54:02doing well. It just it's stealing your
54:04map impressions.
54:05>> I hate that one.
54:06>> Um, there's no pattern to why some
54:09people do well with it and some don't.
54:11And we'll see people who only are using,
54:14you know, very specific ways. And it's
54:18so right now I think this is why there's
54:20a big pullback like my third slide or
54:23something on PMX lead genen. Um, it's
54:26worth trying, but if it doesn't work,
54:28don't bang your head trying to make it
54:30work. Just get rid of it.
54:31>> True. True. I I heard a long time ago,
54:34it's like being in the parking lot of
54:36Walmart getting directions to Walmart.
54:38You're already you're already there with
54:40the maps.
54:41>> Um, so this one, um, any insights from
54:44portfolio bid strategies, uh, versus
54:46campaign?
54:47>> Uh, no. I mean, they're both great.
54:49There's not I mean campaign level let
54:52you especially if you've got a bunch of
54:53campaigns using slightly different
54:54metrics um larger accounts portfolio is
54:57super useful.
54:59>> I use whatever one makes sense to you. I
55:02wouldn't say there's inside differences
55:03but I mean I use portfolio a lot when it
55:06makes sense.
55:06>> So I used it uh most recently because I
55:10had to segment campaigns for reporting
55:12purposes but I wanted them combined for
55:14budget sharing purposes.
55:17So I mean it is an operational reason.
55:20>> I mean I think my only problem with some
55:23of the you know shared budget portfolio
55:25things is that Google might sometimes
55:29overs serve one and
55:32instead of saying this campaign does the
55:34best give it the best budget versus this
55:36is the second best second most budget
55:38etc like a waterfall effect. Um that
55:40doesn't always happen. So sometimes you
55:41get slightly better results going
55:43campaign level budgets and bids, but for
55:45the most part, portfolio works great for
55:47most people.
55:48>> Yeah. So another question, do you have
55:51any thoughts or opinions of large
55:52language models to help manage accounts,
55:56uh, theme grouping, ad rank or
55:58impression share?
56:00>> I have many thoughts. Um,
56:04I find them very I find them from an
56:06idea standpoint amazing. Like I love
56:09going and saying like this is what I'm
56:11advertising for. Give me ad group ideas.
56:14Um and then even segmenting it down into
56:19different parts of the buying funnel and
56:22then you know putting it having it do
56:24the ad suggestions even by buying funnel
56:26stages if we're going to do buying
56:28funnel segmentation.
56:30But half the suggestions are awful,
56:32right? Half of them are well 25% are
56:34great. 25% you would have thought of
56:36yourself. 50% you're like that's what a
56:39terrible suggestion. I would not use an
56:41LLM to manage my account right now by
56:44any means. Um from a data analysis
56:46standpoint from a here's a bunch of
56:49unrelated keywords group them for me.
56:52You still need to proof it. It's a great
56:54first step though. It's it's a great
56:55first step to just quickly do but I
56:57wouldn't copy and paste from it.
56:59>> Yeah, I would definitely agree there.
57:02Um, so this one, uh, going back to your
57:05location, um, still getting people
57:08outside, what can I do besides excluding
57:10locations?
57:12>> Okay. So, what what happens there is
57:16like my wife works
57:19five miles from our house. Doesn't
57:22matter if you target her home location
57:24or her work location. She's in both
57:26regularly and they're right next to each
57:28other. So, you know what? There's no way
57:30because it's in or regularly in you're
57:32gonna you're gonna see that difference.
57:34Um there's not a good way to do it right
57:36now. The only thing you can do is like
57:39you know add geographic keywords to a
57:42geographically targeted campaign but
57:44your impressions drop so much it's not
57:47worth it. Um there's no good solution
57:49right out of this.
57:51>> Yeah, that makes sense. Um so I don't
57:55see more questions on chat. So I get a
57:56question which is always
57:57>> actually there. I take that back. You
57:59could do exclusions in the areas around
58:01you. I mean, that's I think that's what
58:03you already know.
58:04>> So, actually, I'll kind of go back to
58:07that one. So, like if I've ever been
58:08managing account in New York City, I
58:10feel like this has been super important.
58:12It's just really like getting super
58:14granular in those exclusions. It does
58:16feel like whack-a-ole, but you know,
58:18sometimes when you're trying to save
58:20budget anywhere you can, um, taking the
58:23time to exclude on top of inclusions
58:26>> um can can be super important. Um, so I
58:30loved when you went back to old school
58:32stuff. I bet you have the best
58:33screenshots ever from from your years.
58:37So I wanted to talk about taskbased
58:41management um because I totally agree
58:43with you um with regard to that. So
58:47pretend somebody comes brand new and
58:49they're like I don't even know what task
58:50to think about. First of all, does your
58:53calendar still exist? Um and then if not
58:57um where what would be a good first step
58:59for them to start to sort of train their
59:02brain to think of like their weekly
59:04tasks?
59:05>> Yep. So, if the calendar exists, I don't
59:08know, it would have content bids on it
59:10still, right? Like we haven't had the
59:11content network in years.
59:13>> Um, so the presentation,
59:16the bonus section, which I dropped
59:18through, right, you can download it, has
59:20every task on it with a bit more detail
59:23of why you do it.
59:24>> Correct?
59:25>> So, that's really going to be your
59:27starting place of this is daily and why
59:29I do this daily, right? This is weekly
59:31and why. Um, those of you more
59:34experienced are going to adjust those,
59:35right? Some of you should be doing some
59:36of these daily, some may be quarterly,
59:38right? But I would say that's your
59:40starting place. Um, and then
59:44pay attention to the repairs category in
59:47Google Ads. So, I'm not a huge fan of
59:49Google recommendations. I don't think
59:50most people are, but the repairs
59:52category is super useful. But again, use
59:55your brain because like one of them is
59:58remove conflicting keywords.
1:00:00>> Sometimes you actually want the keyword
1:00:02removed, not the negative keyword
1:00:04removed. Right. Google only lets you
1:00:06easily move the negative keyword, not
1:00:08the keyword itself. So, make sure even
1:00:10with those you use your brain um and how
1:00:13things work.
1:00:14>> Yeah. And I actually wrote a search
1:00:16engine land post the other day that I
1:00:18said even that like the application of
1:00:20it should be done with caution or by the
1:00:23human but the suggestion in itself isn't
1:00:26flawed. Um but I had an example of one
1:00:29of my colleagues had uh duplicate
1:00:32keywords broad and exact and of course
1:00:34it took away the exact and left the
1:00:37broad when she applied it through the
1:00:38platform. So just again
1:00:40>> and that's how I mean that's I mean I'm
1:00:42we're almost done here. So that's how
1:00:43adalysis works is we I mean I've been
1:00:46writing accounts for 27 years. So we
1:00:49suggest by default for the human to use
1:00:52their brain
1:00:53>> and then you can automate right or you
1:00:55can overwrite our automation and say no
1:00:57I want this criteria instead
1:00:59>> and I think that's a piece that Google
1:01:01misses is let me redefine the criteria
1:01:04for getting a suggestion.
1:01:06>> Exactly. Okay. So we're one minute over.
1:01:08I'm going to be nice and let one last
1:01:10question slide because you know we get
1:01:12you so not as often as we would like. Um
1:01:15I manage certain accounts that don't
1:01:16have enough budget. Not all is being
1:01:19spent and I still don't get traffic
1:01:21coming in. Any tips?
1:01:22>> Hold on. So, I managed to certain
1:01:24accounts that Oh, have enough budget.
1:01:26>> Oh, yeah. She's giving them the money
1:01:30and I don't get enough traffic coming
1:01:31in. So,
1:01:34in some cases,
1:01:36>> you can't do anything. I'm just going to
1:01:38say this now, right? You want to spend a
1:01:40million dollars a month for a lawyer in
1:01:41Fargo, you can't do it. Um, right? So,
1:01:44there are sometimes it's just not
1:01:46possible. You're you're going to look at
1:01:49can the company expand, right? Sometimes
1:01:52companies can expand beyond what they
1:01:53are. Then maybe that's putting the onus
1:01:55back on the business.
1:01:57How is the account currently restricted?
1:02:00So, is the account restricted due to um
1:02:03its time of day settings? Do other
1:02:05things like that, right? Can you um
1:02:09change that reach some? Um number three,
1:02:11what's your quality score? Because if
1:02:13your quality scores are really bad, you
1:02:15might have a lot of, you know,
1:02:17restricted ad serving or not serving.
1:02:20So, you know, making sure that you're
1:02:21not restricting yourself to either
1:02:23quality score, you have limited ad
1:02:26approvals for whatever reason that that
1:02:28don't need to be limited um so forth.
1:02:31Your next step is can you add more
1:02:33keywords, right? Are there more
1:02:35targeting to be had? This is where, you
1:02:38know, Google suggestions, chatgp,
1:02:41and if you have enough conversion data,
1:02:44even broad match, um, can be useful. And
1:02:47we're going to do another one of these
1:02:48in a little bit because broad match with
1:02:51some bid methods is surprisingly better
1:02:53than I thought it was. I'm changing my
1:02:55mind on broad match, Anthony. We can
1:02:57have that discussion later. Um, right.
1:03:00So,
1:03:02um, next then is are there other
1:03:04channels you want to use? So maybe you
1:03:07jump to Meta right instead of Google and
1:03:09say can we spend it on Meta or or or
1:03:10Microsoft. Um then do you want to test
1:03:15demand genen or pmax or something again
1:03:17getting more into display do you need to
1:03:20just make more demand awareness right so
1:03:22that's YouTube video campaigns so if if
1:03:26you can't spend it because they have too
1:03:28big of a budget for what exists then
1:03:30there's nothing you can do but in some
1:03:32cases it might be how do we make our own
1:03:34demand right and that's where video or
1:03:36display can be very useful so hopefully
1:03:38that gives you some ideas to kind of
1:03:40think about Sarah do you have any to
1:03:41Um, no, I agree. I mean, I'm always a
1:03:44fan also adding a DSA and keyword
1:03:47exploration and then I always say like
1:03:49broaden somewhere, right? Match type
1:03:52targeting. I mean, unfortunately, you
1:03:53got to go wide to come back in and be
1:03:55efficient. It's like the old school cast
1:03:58a wide net and then become more
1:03:59efficient. But your metrics aren't
1:04:01always going to be ideal. So, it's it's
1:04:04exactly what you said.
1:04:05>> You know what? Last, I have one last
1:04:07one. If you're using target CP or target
1:04:10ro bidding,
1:04:11>> can you ask the company, would you
1:04:13rather have more sales if they were more
1:04:15expensive and increase bids?
1:04:18>> Yeah.
1:04:19>> Yeah. Yeah.
1:04:20>> I'm not a Google rep.
1:04:22>> Right. Right. Exactly. Um well, thank
1:04:24you so much. I think this is fantastic.
1:04:27Cannot wait to have you back. Um and and
1:04:29to the audience and everyone who stayed
1:04:31along, this will definitely happen
1:04:32again. So, thank you all for being here.
1:04:34Is there any last thing like how to
1:04:37contact you? We'll put everything on the
1:04:38paid search association. Thank you for
1:04:40the plug there.
1:04:41>> You can find me on LinkedIn. I live on
1:04:43LinkedIn way more than any other social
1:04:44platform.
1:04:46>> So that's the easiest place to get a
1:04:47hold of me. Or you can go to our website
1:04:49atalysis.com.
1:04:51>> Um we write a blog with a ton of
1:04:52information. We don't publish much, but
1:04:54when we do, it's worth reading.
1:04:55>> Good. Yes. Awesome. Well, thanks again
1:04:58and uh thank you for everyone who came
1:05:01today. Enjoy. All right,

About this transcript

This page contains the full transcript of Managing Google Ads accounts in AI era – Adalysis & the Paid Search Association by Adalysis, generated from the public captions YouTube serves with the video. The transcript has 11,268 words across 1,654 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.