YT YouTube2Text

YouTube2Text → Transcripts → Google Ads ad testing masterclass with Brad Geddes and Bob Meijer

Google Ads ad testing masterclass with Brad Geddes and Bob Meijer — Transcript

by Adalysis · 10,384 words · 1,694 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:02All right, there we go.
0:05Should be running.
0:06Okay.
0:07Good morning, good afternoon, good
0:08evening everybody all over the world.
0:10We've got nobody else
0:13than Brad Geddes today for a great
0:15workshop about ad testing.
0:18Brad will properly introduce himself in
0:19a bit, but I'm going to try to do a
0:21little bit better because this is a
0:23little special moment for me because
0:26Brad, I don't know if you know, but
0:28you've been one of my heroes for the for
0:31the past couple of years, especially
0:32when I started out in Google Ads.
0:35I just showed you
0:36this book. It was actually the first
0:38book that I bought when it comes to
0:41Google Ads. It just
0:43says Google AdWords on the on the on the
0:46cover. That has changed, like you
0:48mentioned in a bit, but the the theory
0:51around well, Google Ads as a as a
0:53platform still stands, but obviously a
0:56little thing here and there changed
0:59since you wrote this book. This is the
1:01third edition, I think. From what year
1:03is this, Brad? Do you still recall?
1:05I've no idea. Last decade. It's not new.
1:11It's not new at all.
1:12It's still very nice to have.
1:14So so yeah,
1:15that goes a long way back. I think I
1:17bought it somewhere 10 years ago, maybe
1:20even longer ago.
1:22So, more about Brad.
1:24Brad, you've been in this industry for
1:26quite some time. I think since the
1:28internet started somewhere in 1998 and
1:32when paid search became a thing, well, a
1:34lot of us didn't even know what paid
1:36search was back then.
1:38I saw that you spoken around on around
1:41to 125 conferences worldwide
1:44since you started out,
1:46which is crazy to me. So, that's an
1:48incredible amount of conferences you
1:51well, you spoke at and you went to.
1:53You're also the founding board member of
1:56the paid search association, a big
1:59association in the USA,
2:02if I'm correct, and well, you're on the
2:06most most influential PPC expert list
2:09around the world, which is also really
2:11cool.
2:12So, today you're the founder of Adalysis
2:16and you've been running Adalysis, one of
2:17the leading Google Ads monitoring,
2:19optimization and testing platforms, a
2:21tool that I use myself as well next to
2:23other tools, but I think you really
2:25complement other tools like TrueClicks
2:28as well in certain areas. So, I like
2:30combining tools like those together.
2:34And when we wanted to host a session
2:37about ad testing, I couldn't think about
2:38anybody else than you because ad testing
2:41is what you what you do and what you
2:43live. So, that's why you're here and I'm
2:46really glad that you found some time in
2:48your agenda, your busy agenda to to do
2:51this for us. So, welcome.
2:54Uh I yeah, I'm honored to have you and
2:57yeah, please please take it away. All
2:59right, thank you, Figgy. Yeah, so I'm
3:00Brad Geddes and I've been doing paid
3:02search for
3:0527 years now, so sort of seen it all,
3:08right? And you know, when you think of
3:12ad testing,
3:14it has been around for decades. This is
3:18not new, but the implementation changes
3:21over time.
3:22It is something that's often
3:25underutilized
3:26in many accounts. Especially the RSA
3:28story, which I'll I'll get into in a
3:30second here.
3:31And when often you think of account
3:34management, you have things which are
3:37transient. They are good until you
3:39change it again. You know what a bid is
3:42is good until you change your bid again
3:44and all the time you spent setting bids
3:46was essentially wasted, right? Cuz you
3:48changed it. Which is why we love
3:49automated bidding now and we're not
3:51manually calculating bids over time.
3:53Right? Now, when you find improvements
3:55to like ads, they often can last months
3:58or years of time as far as how things
4:00go. So, they often have a much longer
4:03term increase in in what you're doing.
4:06Now, the problem with ad testing is it's
4:10only as good as your ad group
4:12organization. So, I'm not actually going
4:14to start with the ad testing.
4:16I'm going to start real quick with
4:17organization because if you have a bad
4:19organization, your testing is useless.
4:22Right? And and this is the problem with
4:23the RSA story. So, if we look at
4:26accounts a decade ago, they often had
4:28really good organization. And then
4:30Google came out with this whole story,
4:32"Hey, give us your keywords, give us
4:34your assets, throw it all together and
4:36we'll run your ads for you."
4:38And we saw a lot of accounts actually
4:40get much worse organization. So, like
4:43this is a a company, they're a hotel in
4:45New York City and what they're known for
4:48is being pet friendly. Hey, want to
4:50travel with your dog or your cat and you
4:51want to be in New York City, this is a
4:53great place to stay.
4:54So, they have keywords, you know, NYC
4:56dog friendly hotel, Times Square hotel,
4:58etc. And of course, they have ads like
5:01dog friendly hotel. Hey, they are by
5:03Times Square, they're near Javits
5:04Center, it's a big event place. So, here
5:07you go, Google, take our stuff.
5:09Now,
5:11the problem is Google says we look at
5:13this query, we look at your assets, we
5:15build an ad. But if we look closely, we
5:18see that the keyword, right, dog
5:20friendly hotel, has been shown 19,000
5:23times.
5:25The asset match in that is 4,000. In
5:27fact, most people searching dog friendly
5:30actually see downtown New York hotel.
5:34And you just lost that one critical
5:36piece that makes them stand out and
5:38what's different. Right? And so, you
5:40know, I said we talked to them and said,
5:41"Hey, go fix your organization,
5:43basically." And they took that one ad
5:45group, they made it a lot of different
5:47ad groups. Right? And and this simple
5:50statement, right, change increased their
5:53overall conversion rates and
5:55click-through rates almost, you know, in
5:56some cases their their conversion rates
5:58went up four times. Right? It's a huge
6:01difference cuz you're now making sure
6:03your ads, landing pages and keywords
6:05match each other. Now, I know this is
6:07basic, right? This is the essential part
6:08of any setup, but if you don't have this
6:11right, then your ad testing often falls
6:14down and it's often why we see weird RSA
6:17assets get displayed, too.
6:19Right? So, simple flowchart of, you
6:21know, how to sort of organize ad groups.
6:23Now,
6:25if you're ever struggling with ad group
6:27organization, actually ad relevance is
6:30an amazing tool to use
6:32cuz Google actually tells you, right,
6:34this
6:36ad and keyword can't work well together.
6:39And if you're ever thinking, "Hey,
6:40should we move stuff around?" just
6:42looking at your quality score ad
6:45relevance factor can tell you a lot
6:47about how you're organizing things and
6:49and what can be done.
6:52Now, beyond that, right? Once we have a
6:53good organization in place, then we
6:55think about our our ad components we're
6:58going to get into before testing.
7:00So, I'm going to use the word relevancy
7:01headline a lot, right? Relevancy
7:03headline just means someone did a search
7:05and the ad at least matches what they
7:08searched for.
7:10And unfortunately, you'll see a lot of
7:12ads, this is dishwasher replacements.
7:15And the second ad from a very large
7:17company says, "Same day, next day."
7:20Doesn't mention what the keyword is.
7:22Right? This is a $10 to $20 click.
7:25Jeep Wrangler car insurance and only one
7:28ad mentions Jeep and these are all giant
7:31companies. Right? And this is number
7:33one, do your ads even talk about what
7:36someone wants?
7:38Number two, features and benefits.
7:42Features are bullet point list of true
7:44facts about a product.
7:46A benefit is why it makes it useful.
7:49So, something's got a 9-hour battery.
7:51What does that mean? Right? Hey, you can
7:54carry around your computer around the
7:55office all day without hunting for a
7:57plug. Right? That's actually something
8:00useful as opposed to 9-hour battery.
8:03Now, when we think of something as
8:05simple as feature versus benefits,
8:07right, we have to before we decide what
8:09to use in an ad, we actually have to
8:10think about where someone is in the
8:12consideration stage of the funnel.
8:15So, to me, Google Ads is not a channel.
8:18Google Ads is really about 20 channels
8:21in one.
8:22Right? Because a brand keyword versus a
8:24buy keyword versus a product ad versus a
8:28display ad are different funnel reaches.
8:31So, if we think of where this ad meeting
8:34someone in the funnel,
8:36we can go back and say, "Should we focus
8:38on features or benefits?"
8:40If they don't know who you are,
8:42then you have to say why this is useful.
8:45We comparison shop on features.
8:48Right? You're buying a new computer,
8:50you're going to say, "How much RAM do
8:51you want? What's the screen size? What's
8:53the battery life?" Those are just
8:54features, they're true facts.
8:57We buy cuz of what something does for
8:59us.
9:00Right? Now, our next component is our
9:01CTA.
9:04Unfortunately, the vast majority of
9:06calls to action on the entire web are
9:09terrible.
9:10In fact, many are pre-written by Google,
9:12right? Subscribe now, buy now, shop now.
9:16You know what?
9:17You don't even read the words anymore,
9:19right? You basically just ignore them.
9:21We're we're blind to a lot of these
9:22words these days.
9:24So, with calls to action, it's not about
9:27saying what to do.
9:29Right? It's about saying
9:31"What do you get out of us saying do
9:35this item?"
9:37Right? So, I mean, one of my favorites
9:39was this this company, they're a a
9:41nonprofit volunteer company.
9:43And they had this ad, "Donate your time
9:44today." And it did terrible. And they
9:47changed the line to be, "Help humanity,
9:50volunteer today." And this was a huge
9:52impact by like, "Hey, just help your
9:55fellow person." And suddenly, they got a
9:58lot more volunteer hours from something
9:59simple.
10:01So, it's not about just "Hey, buy now,
10:03shop now." Shop now for Why? What do you
10:07get out of going to this place's, you
10:09know, site and spending money?
10:13But, with calls to action, we have to
10:16also think back to the buying cycle.
10:19Because if someone doesn't know who you
10:21are, you can't say buy now.
10:24But, if they're actually shopping and
10:25they say, "I'm looking to buy this
10:27product," then you want to match where
10:28they are in the funnel.
10:29So, if they're, you know, they don't
10:31know much, right? That's that soft CTA,
10:33discover, learn, find out how. Then you
10:37move to some actions, right? Where
10:39you're often maybe content gating some
10:42some, you know, PDFs or or ebooks. It's
10:45like, "Well, now you can download or,
10:47you know, find this." And then when it's
10:49ready to they're at the end of the
10:50funnel, right? Schedule a demo, buy,
10:53call us, etc. So,
10:55if someone's further down the funnel,
10:57right? And you put in discover and
10:58they're ready to buy,
11:00you've left money on the table cuz
11:01you're not matching where they are cuz
11:03they're actually ready to take action.
11:05So, often want to think, you know, where
11:06do they sit?
11:08Now, there's a whole lot more we could
11:09get into and I'm not going to spend
11:11forever on this cuz we could talk about
11:12pain points. What are you solving?
11:15Authority statements can be huge. If
11:18there's 20 options on a page,
11:20why are you the one they should pick?
11:23Right? We haven't mentioned FOMO. Fear
11:25of missing out or savings, emotions. So,
11:28there's a lot of other pieces here. And
11:30these are things to think about when
11:32you're just creating your assets,
11:34thinking, "How do we want to speak to
11:37users? Right? What do they need to know?
11:40What's our brand message? Where are they
11:42in that funnel to match our brand
11:43message to to put this out there?"
11:47So, now when we think about assembling
11:48our ads,
11:50right? Number one is
11:53when you're adding assets, are you using
11:56roughly equal numbers of various types?
11:59So, if you have
12:0110 assets related to the the ad group
12:04and you have, you know, three calls to
12:06action,
12:07then most of your ads are just going to
12:08be duplicates that are related to the ad
12:10group.
12:11If you have eight calls to action and
12:13two relevancy lines, you'll see ads that
12:15are just called to action, called to
12:16action, which don't make any sense to
12:18users.
12:19So,
12:21one of the issues and and Google
12:23actually made a change to how they
12:25rendered ads because of this was a lot
12:27of people had very similar assets and
12:30too many assets of one type.
12:32And you know what happens when an ad
12:34shows three calls to action and three
12:36headlines? No one knows what to do
12:38because that just is overly aggressive.
12:41So, that's where Google last year, the
12:43year before, made it so an ad can only
12:45show one headline.
12:46And this helped many people because
12:49instead of showing feature feature, they
12:51could just say one feature in headline
12:53and not duplicate lines. So, when Google
12:56went to only showing one headline,
12:59it was in reaction to how poorly many
13:03ads were actually made. Now, some people
13:04have fantastic ads, right? But, when you
13:06look across the board, there's just a
13:08lot of bad combinations.
13:11So, we have our assets and we say,
13:13"Let's use a mixture," which could be
13:14authority statements, calls to action,
13:16relevancy lines, so forth.
13:19Then next, we have to make this
13:20decision, do we pin?
13:24And with pinning,
13:26I just want to make sure this with
13:28pinning, you're telling Google, "I want
13:30this asset in this particular position."
13:33But, you could put multiple assets in a
13:36in a same pin position. See, in this
13:38second example,
13:40we have two lines pinned to position
13:42one.
13:43So, Google can switch back and forth.
13:44You could have five lines pinned to
13:46position one.
13:47So, you're telling Google, "You know
13:49what? We know from years of data,
13:52headline one works best when it's
13:54related to the keywords in the ad group.
13:57Headline two works best when it's a a
13:59feature, an emotional appeal, a call to
14:02action."
14:03So, we don't need you to go through all
14:06this math to do that combination we know
14:08that's actually how we want the ads
14:09rendered.
14:10So, pinning speeds up this process, but
14:13you can pin more than one asset to a
14:15line.
14:16Now, when you don't pin and you don't
14:19have a lot of data,
14:21what you end up with are ads that just
14:23repeat themselves cuz Google defaults to
14:25relevancy when they don't know better,
14:27meaning they're going to pick lines that
14:30match your keywords.
14:32And this company doesn't have a lot of
14:33data. Like, these are often 20
14:35impressions, 15 impressions for a dry
14:37cleaner.
14:38And ads saying
14:40"Wedding gown preservation, wedding
14:42dress preservation" over and over again
14:44doesn't do anything.
14:46It's not because they have bad assets.
14:49In fact, they have some very good
14:51assets.
14:52Express it, they're over 100 years old.
14:55Right? They they pick up and deliver for
14:57free, which is a dry cleaner that comes
14:58to your business, grabs your dry
15:00cleaning, takes it with them, and
15:01delivers it back to you. That's awesome.
15:04Google had shown those lines a total of
15:06once.
15:08When those lines were in their ETAs,
15:10their conversion rates were almost 3 and
15:121/2 times higher.
15:14So, this is where pinning is very useful
15:16cuz you're telling Google, "This is what
15:19really matters to our users."
15:22Now, the other reason we pin, it's a
15:24math problem.
15:26So, if you have 15 unpinned headlines,
15:31that's almost 3,000 combinations.
15:34If we also have four unpinned
15:36descriptions, that makes 50,000
15:39combinations.
15:40Now, if we pin three lines to headline
15:42one and we have 12 more unpinned assets,
15:45that makes 435 headline combinations.
15:48We pin three headlines to headline one
15:50and two, that's 93.
15:52So, what we're doing is we're increasing
15:54data density,
15:55right? Of how much data Google really
15:57has to understand what works within our
16:00framework to render ads better.
16:03Now, it's hard to visualize 50,000
16:06combinations.
16:07So, instead, let's talk about time.
16:10If you have 15 unpinned headlines
16:13and you have four unpinned descriptions,
16:16and you have an ad group with 10,000
16:19impressions a month, which is a decent
16:20healthy number of impressions per month
16:22for an ad group.
16:23And Google likes 100 100 is really their
16:27magic number for making decisions.
16:29And you want 100 impressions for each of
16:32those potential renderings.
16:34That would take 41 years to accomplish.
16:37Right? So, I mean, you're obviously not
16:39going to ad test. If you add a second
16:40ad, you're talking 82 years. Clearly,
16:43testing having an ad test done in the
16:45next century is not going to be useful
16:47for you.
16:48Right? So, then we say, "Well, how much
16:50time do we have based on our
16:52impressions? So, if we had 100,000
16:56impressions a month,
16:58then, you know what? If we pinned five
17:00lines to headline one and five lines to
17:02headline two and we had 10 times this
17:04data,
17:05we would actually have a nice
17:06combination in the in about a month and
17:08a half.
17:09Right? So, when we're thinking about how
17:11much do we want to pin,
17:13it's a math equation in the end,
17:15right? Saying, "How much data do we need
17:17to to really get this information?"
17:21And this is why, you know, in most
17:24places, pinning three to five lines to
17:27headline one and two
17:29has enough data per ad group to get
17:33results, but you're still giving Google
17:34some freedom, right? Cuz you Hey, they
17:36have five headlines and five
17:37descriptions. They have some freedom to
17:39serve and find combinations,
17:43but you know, "Hey, we want, you know,
17:45relevancy in headline one.
17:47Maybe we'll add some authority
17:49statements and
17:51calls to action to headline two and then
17:52we'll kind of let Google play with
17:53three." Right? And and that's where you
17:56can have some control,
17:58but still use machine learning.
18:01Now,
18:02the other aspect I'm going to get Katie
18:05to ad strength in a moment because it is
18:07a useless metric, but we'll cover ad
18:09strength in a second. I saw the comment
18:10pop up. So, the second part of this is
18:14sitelinks. This is something that I
18:16talked about.
18:17So, Google can take two of your assets
18:21and turn them into sitelinks.
18:24When Google takes an asset from an ad
18:28and makes it a sitelink,
18:30the clicks on that asset go to the same
18:32landing page as your ad.
18:35So, now we have to think about this a
18:37little bit, right? So, some companies
18:39are like,
18:40"We have a very small website and we're
18:42often just randomly making up stuff for
18:44sitelinks, so they have to go to
18:45different pages."
18:47Well, suddenly having three links that
18:49go to one page ads great, right? Now, we
18:52don't have to think about as much of
18:53this. Others
18:54are like, "Hey, I actually have a very
18:57strong sitelink strategy and we need to
19:01use our sitelinks to send people to the
19:03right place because our website's
19:04massive. We don't want Google to turn
19:08our assets into sitelinks cuz we want
19:10that variety. So, this is much more of
19:12your strategic level of sitelinks."
19:15Now, in this ad, we've got, you know,
19:17six lines made, just a random number.
19:19Two are pinned. One has to be a headline
19:21one. So, of the other three headlines
19:25left, two of them can turn into
19:27sitelinks.
19:29Now, if
19:30we we want, "Hey, you know what? We like
19:33this idea of more assets being turned
19:36into sitelinks." We could take an ad
19:38like this and we've pinned two lines to
19:40one, two lines to two, and one line to
19:41three.
19:43But, we have four more assets.
19:45Those assets can only be sitelinks.
19:50So, there are interesting ways to use
19:53extra unpinned assets as sitelinks. Or
19:57you might say, "You know what? We we
19:59actually don't want Google to use our
20:01assets as sitelinks cuz we need this."
20:03So, if we just say pin in any position
20:06to these lines, then Google honors
20:08pinning. They cannot become a sitelink.
20:11So, anything pinned can't turn into a
20:14sitelink. So, often pinning is as much
20:16about sitelinks as it is also about ad
20:18control.
20:20Regarding ad strength, so pinning word
20:23does affect strength, you're absolutely
20:24right.
20:25Ad strength does not
20:28affect quality score. It does not affect
20:31ad rank. It does not affect your
20:34eligibility for the auction.
20:36In fact, we've done a lot of research on
20:38this and we're actually refreshing our
20:40research right now.
20:42Lower ad strength ads traditionally have
20:44better click-through rates and
20:46conversion rates than high ad strength
20:47ads.
20:49Now, if you think about it, it's
20:50actually the math goes back to the math
20:52problem. If you have 15 headlines and
20:54four descriptions, that's 50,000
20:56combinations. So, it takes forever for
20:59Google to figure this out. If we pin a
21:02few things, yes, we have a lower ad
21:04strength,
21:05but then suddenly there's better data
21:07density and Google serves it better.
21:09So, often lower ad strength is better
21:12metrics.
21:13Now, there's a couple things to note.
21:15One,
21:17if you have two ads in an ad group with
21:19different ad strengths, the higher ad
21:22strength usually, but not always, gets
21:25more impressions. Cuz Google thinks it's
21:27better, right? Cuz it but it doesn't
21:28affect your ability to be in the
21:29auction.
21:31Second,
21:33if the reason you see you have a low ad
21:36strength is because you have the include
21:39popular keywords in your headlines.
21:42This is often an indication
21:45you have too many keywords for your
21:47current assets and they're just not
21:49relevant enough.
21:50Or you really haven't made some lines
21:52that really encompass the keywords. So,
21:55and it could be one of those weird
21:56situations where,
21:58you know, lawyer, attorney, barista,
22:01etc. are very similar terms, but they're
22:03technically different, right? And Google
22:05sometimes does
22:06weird things with small word variations.
22:09If you see make your headlines more
22:12unique and you are not pinning, that
22:15usually says you have just too many
22:17similar ads and you want to to change
22:19things up a bit more.
22:20But if you see,
22:22you know, include popular keywords in
22:24your headlines is checked
22:26and you're pinning a lot and you have a
22:28strategy, the rest does not matter. So,
22:31they are good suggestions, especially
22:33the include popular keywords cuz that
22:35usually says you don't have good
22:37relevancy lines or you have too many
22:39keywords in your ad group for this
22:40particular asset type.
22:43All right, so
22:44the hardest thing for marketers in many
22:46cases is to get good ad ideas. And it
22:48it's not because we're not creative,
22:50it's because in many cases you are the
22:52only person in your company managing
22:53search and you don't have someone to
22:55bounce ideas off of. Or you're an agency
22:58and you're in the B2B side and everyone
23:00else is business to consumer, right? And
23:02so it's hard to bounce ideas off of
23:03them. So, this is where, you know, LLMs,
23:06I'm not a fan of them managing accounts,
23:08but they're great for ideas. Right?
23:11Something simple like, here's my
23:13keywords, here's my landing page, here's
23:15the language, right? Write the ad in the
23:17same language as the keywords. Create
23:19this prompt. This is a very simple
23:21prompt. This is actually one of my
23:22favorites, which it's much longer, so
23:25there's a link to download this cuz it's
23:26it's you can't fit it on a on a screen.
23:29But this will make about 45 headlines,
23:31right? It'll make emotional headlines,
23:33it'll make CCAs, etc.
23:35These are not copy and paste for most
23:37people. These are get ideas, wordsmith,
23:41use. But if you're looking for a large
23:43variety of ideas,
23:47then LLMs are great for this.
23:50Now,
23:52where you can have some fun, right? Is
23:54add additional rules to a script like I
23:56put in there. Like, "Hey, make headlines
23:58that rhyme or use business language cuz
24:00it's for C-level execs." Or make the ads
24:03more playful if you're doing, you know,
24:05kids toys and so forth. So, you know,
24:08there's a lot of ways you can use just a
24:11few words and get very different lines
24:14out.
24:15Now,
24:16if you write some lines to and you
24:21really don't like how the ads are
24:23generated, they didn't fit your
24:26business, how you want it be be
24:28portrayed,
24:30save them.
24:31Google is slowly rolling out
24:35their
24:37this is how auto-created assets, the
24:39rules they should follow, right? Which
24:41will be boxes you can write in some text
24:43and tell Gemini, "Don't go this
24:44direction."
24:46Those lines you wrote that you did not
24:47like the results are actually going to
24:49be great lines to paste into the
24:51auto-generation guidelines of saying,
24:53"Don't write ads this way for us." So,
24:55there's actually some good stuff to save
24:57to use when those get fully rolled out
24:58for for asset creation.
25:01All right, so let's move to ad testing.
25:03So, there's two types of ad testing.
25:05One is single ad group testing. We put
25:07some ads in an ad group and we say,
25:09"What's best?"
25:11A lot of times people ask, "Well, if
25:13we're giving Google assets and we're
25:14giving Google
25:17data, they're making the ads, do we
25:19still need to ad
25:21This ad group, single ad group, spends
25:24about a million dollars a month. This is
25:2560 days of data.
25:27And if we look at impressions, Google
25:30has served the ad with the lowest
25:32click-through rate, the lowest
25:33conversion rate, and the lowest
25:34conversion per impression the most.
25:37The pause button got this company about
25:40little more than 300 more conversions
25:42per month by just pausing the middle ad.
25:45Right? And just because Google has the
25:47data, it does not mean they always use
25:49it properly and it's not uncommon to see
25:54Google show the worst ad the most. And
25:57this is also a really unusual one
25:59because
26:01our highest ad strength ad, the bottom
26:03one, has by far the least amount of
26:05impressions.
26:06Right? So, Google's not even serving the
26:08highest ad strength the most.
26:10These types of examples are super easy
26:13for us to find. It's not that this is an
26:14exception, they're incredibly common.
26:17Right? This is just from ad analysis,
26:19this is from our software.
26:20So, we definitely want to test. The
26:22question is, do we test within an ad
26:23group or do we do ad group aggregation?
26:28Where you have those bigger question
26:30words. Like, every time a big brand
26:32comes on and you look in their ads,
26:34they've got their brand name in every
26:36ad, even for non-branded products. Like,
26:39you know what? You should run a simple
26:40test
26:41of looking at your non-brand keywords,
26:46making one ad branded and one ad
26:48non-branded and see which does best.
26:51Now, when you think of a test like that,
26:53they don't want to do this in an ad
26:54group, right? They may have tens,
26:56hundreds, thousands of ad groups, they
26:58don't want to know this. So, often with
26:59multi-ad group testing, we're thinking
27:02about an idea and rendering it across
27:05large numbers of ad groups, then
27:07re-aggregating that data.
27:10So, when we think of testing, right? We
27:12can do unpinned versus unpinned. Unless
27:15you have a crazy amount of data, you
27:17will never get a good result here.
27:19Right? Cuz if you have 10,000
27:21impressions a month, we're talking again
27:23decades to actually get Cisco
27:25significance.
27:26Now, you'll get Cisco significance, but
27:28half your assets will barely get
27:29impressions.
27:31We can do unpinned versus pinned. Now,
27:35this is a and often a verification type
27:38test where you give Google some assets
27:40unpinned or you take what you've had
27:43running and you look at at the top few
27:46renderings, you know, that Google has of
27:49what ads they served the most.
27:51And you take those and you make them
27:52pinned ads. You say, "Okay, Google,
27:54you're serving this one the most,
27:57why? Is it actually better than the
27:59others? Is it worse? Did you find some
28:01lines we didn't realize? Or was this
28:03randomness?" And that kind of is a
28:05verify what Google has done.
28:08The most common test by far
28:11is what I often call theme testing.
28:13Right? Ad one is discounts, ad two is
28:16prices, ad three is shipping. Or ad one
28:19is call to action, ad two is authority
28:21statements. Should we focus on CTAs or
28:24authority? Right? By far the most common
28:27test run is just here's a few
28:30higher-level ideas. Should we care about
28:32our ratings and reviews or downloading a
28:34white paper? What really matters to to
28:37our metrics?
28:38Then if we have this idea, right? Which
28:40we're often pinning that in a headline.
28:43We might pin other headlines, too,
28:44right? Which is which is fine.
28:47And if we have that idea, how do we want
28:48to test it? Right? We could say,
28:51"I want to do this in one ad group."
28:52Which is single ad group testing, right?
28:54Good for insights for those keywords.
28:56You have really high traffic ad groups.
28:59You have ad groups your boss watches
29:00every day and they say, "Why is our ad
29:01not running for this keyword?"
29:03You often want to really control what
29:05those ads look like.
29:06Right? And then we have our multi, which
29:09is saying, "We want to know for all of
29:12our non-brand, should we put geographies
29:15or not geographies in our ads? Should we
29:17localize them or not?" Right? How does
29:19that matter? And that's often your your
29:21sort of bigger market research
29:22questions.
29:24So, if you have the big questions,
29:27right? Does quality matter? Does
29:28shipping matter? We often see in B2B
29:32e-commerce
29:33that rush shipping is more important
29:35than free. They'll pay for it, they want
29:37it tomorrow.
29:39It where in B2C e-commerce, it's the
29:41opposite. Right? So, that's often a a
29:44good test in business lines.
29:46If you're doing multi ad group testing,
29:49then you need a key to look at your
29:51data, which right now is labels.
29:54It's going to shortly change to being
29:56assets, and I'll get to that in in a
29:58second.
29:59So, like in this case,
30:01they are are a installation company, and
30:03they wanted to know
30:05if we match price to the person's
30:08experience level, does that make a
30:10difference, right? So, if we say, "Hey,
30:12get an apprentice, which is an official
30:14title,
30:15then it's cheap versus master craftsman
30:19will come out, but they're really
30:20expensive."
30:21And it turns out people are willing to
30:23pay for the best, right? And that's what
30:24this company learned out as far as how
30:26their ads go.
30:29So, you have your ideas.
30:31You know, do you want to test this in
30:32one ad group or multiples?
30:34And then we need to set up our tests.
30:37Now,
30:39the most common
30:41testing metric people use
30:43is click-through rate, which I need to
30:46test, and I'll get to why in a second.
30:47It's by far the most common metric.
30:50Number two
30:51is conversion rate.
30:53Now, the problem with conversion rate is
30:56it ignores how many clicks you get in
30:59the first place.
31:00Right? So, let's say I have an ad, ad
31:03one
31:05it Well, I'm going to use simple
31:06numbers. Ad one is a 10% click-through
31:08rate
31:09and a 10% conversion rate.
31:11Okay?
31:12Ad two
31:14is a 3% click-through rate, but a 23%
31:18conversion rate.
31:20Which one's actually better?
31:22The one that's a good click-through rate
31:24and a good conversion rate or really bad
31:25click-through rate, but a really good
31:27conversion rate?
31:28We have mixed metrics, right?
31:31And And so, when I think of testing,
31:33I think of the impression.
31:36When your ad is displayed, you have a
31:39chance of a conversion. Right? You're
31:41Someone You have a chosen a keyword,
31:42someone searched for a keyword, your ad
31:44has a chance of a conversion.
31:46Now,
31:47in B2B, we might be doing things like
31:49pre-qualifying users and saying, "Don't
31:51click my ad if you're not this or you
31:52luxury travel, you know, our package is
31:54starting at 10,000." So, there are times
31:56that everyone's actually, you know,
31:57potential conversion.
31:59But, we still want to say, "Here's our
32:02opportunity
32:04is an impression."
32:06So, then for lead gen, as opposed to
32:09conversion rate,
32:11we're often thinking of conversion per
32:13impression.
32:14Hey, I have this many impressions.
32:17I got this many conversions.
32:19Now, which combination of click-through
32:22and conversion rate gets me the most
32:24potential customers?
32:26Right? So, this is simple math. I mean,
32:27we're just this conversion divided by
32:28impressions. It's not a difficult
32:30formula to do,
32:31but it takes click-through rate
32:34and conversion rate into an account.
32:37Now, this is great on lead gen, but as
32:40soon as you have variable checkout
32:41amounts or you're tracking ROAS, this
32:43this fails.
32:45So, on the e-commerce side, we can do
32:47the same thing, right? We could say,
32:49"What is our either revenue? If we have
32:51profit, we can use profit." This sort of
32:52depends on how you pass data data back
32:54to Google.
32:55We could say, "Revenue divided by
32:57impressions or revenue minus cost
32:59divided by impressions."
33:00When ad one is served, I make 33 cents.
33:04Oh, sorry, when ad one is served, I make
33:0546 cents. When ad two is served, I make
33:0733. I'd rather make 46 cents an
33:09impression than 33. That's a winner.
33:12Right? So, again, we're not doing
33:13complicated math. We're just instead
33:16saying,
33:17"Let's look at the impression
33:19opportunity as opposed to the click
33:22opportunity in how we measure success."
33:27So, then all we're thinking about is
33:29what should we use?
33:31Well, if you want the most conversions,
33:32it's conversion per impression.
33:34Now, there are companies who are just
33:35saying, "I want the most people
33:36possible, right? Just click on my ad."
33:39Then that that would be CTR.
33:41If you're really struggling with quality
33:43scores, right? Then you then CTR is a
33:46fantastic tiebreaker.
33:48If two ads really close in conversion
33:50per impression, one has a much better
33:52click-through rate.
33:54The better click-through will give you
33:55more audience people, give you more
33:56people for remarketing, right? So, that
33:58one is it's often a fantastic tiebreaker
34:01metric.
34:02Right?
34:03Now, we if when some cases, we may have
34:05filtering metrics, where, you know,
34:07there's an a needing from the company
34:09that says,
34:10"You must You can't go over a $30 CPA."
34:14Well, if ad one is 35 and ad two is 25,
34:17ad one can't ever actually be a winner.
34:19It's above our metrics. So, it can't be
34:21a winner. We have to choose maybe the
34:23lower CPA cuz it fits the goals.
34:27Now, next, we often want to define
34:30minimum data.
34:32Right? And And this is where
34:35it's an essential process to say, "We
34:38don't want to look at results until we
34:40have at least X amount of information."
34:43Cuz Google will start to make decisions
34:46within 100 impressions. But, if you ever
34:49run statistical significance,
34:52and you say, "What is this ad test at
34:54100 impressions versus 1,000?" It's
34:57often significantly different.
34:59Cuz statistical significance has this
35:01assumption that data that came before
35:04will be data like the data that comes
35:06later.
35:07Except when we look at paid search,
35:10your keyword being searched on a Monday
35:12morning versus a Wednesday night versus
35:14a Friday afternoon have different sort
35:17of intent from users to them. So, we
35:18often want to look at larger time
35:20frames.
35:21Now, there is no number that says this
35:25is what you should use for minimums.
35:27Cuz you know what? If you're Expedia,
35:29and you probably want 100,000
35:31impressions for your brands. If you're,
35:33you know, my small regional travel
35:35agency,
35:37then you might not get 100,000
35:38impressions a year. So, you might need
35:40to use a much lower number. So, there is
35:43no one number. What you should keep in
35:45mind is
35:47if one person makes a significant
35:49difference on your data, it's not
35:50enough.
35:51So, if I have one conversion on 100
35:54clicks, I have a 1%. If the next person
35:56converts, I went from, you know, I went
35:58from 1% conversion rate to 1.98%. I
36:01doubled my numbers in one person. That's
36:03not enough.
36:04Right? So, we often want to define those
36:06minimums. A lot of companies do monthly
36:09ad testing. That's great. You know what?
36:10It's four weeks. It's often a couple
36:12sales cycles. That's a great way of
36:14doing it.
36:15Doing a test on Monday and finishing it
36:17Wednesday usually is a very not enough
36:19data regardless of, you know, how much
36:21impressions you have.
36:23Next, we have our confidence levels, our
36:25statistical significance, which is just
36:27a percentage chance something is not due
36:30to randomness. That's all it really
36:31means. But, if you have an 80%
36:33confidence, just means
36:3520% of the results are random.
36:37So, you know, some companies who are
36:40really small will go to 80% confidence,
36:43but you're saying
36:4420% of our ad tests are due to
36:47randomness, not actual data. So, usually
36:4990% the lowest most people want to go.
36:53And then you've got, you know,
36:54third-party brands you sell. You might
36:56go to 95. For your branded terms, often
36:5895 to 99 depending on how much is
37:01served. Right? But, essentially, what's
37:04your tolerance? What's your percentage
37:06chance you will pick a winner because of
37:09randomness versus actual data. That's
37:12what you're deciding when you pick out
37:13your confidence levels.
37:16So, then we think about actions, right?
37:18It's real simple. You make an ad test.
37:20You've got two ads in an ad group or,
37:22you know, two ads in 50 ad groups.
37:23You're aggregating data.
37:25You say, "If we're below minimum data,
37:26we wait.
37:28If we're above minimum data, we
37:29calculate confidence factors. If we're
37:31below, we wait. If we're above, we take
37:34action."
37:35Right? And then if we're above max data,
37:37we we didn't we actually set a cap to
37:39say, "If we don't have a result in three
37:40months or this much data, we also take
37:42action." Action is just
37:44pausing a loser or quitting a new ad.
37:46Now, do not do this by hands. Use a
37:50script, use a tool, do something.
37:53This is what happens when most people
37:54sit down to to do ad testing when
37:57they're running the results in like
37:59Excel.
38:01Week one, they sit down, and they make
38:03100 ad tests, and they're great.
38:05And, you know, they wait a couple weeks.
38:07And so, after a couple weeks, they they
38:09go, and they download data once
38:10statistical significance, and they have
38:12one winner. Like,
38:14I just sat there and downloaded data and
38:16uploaded here and ran all these
38:17formulas, and I have one thing to do.
38:20So, they do the same thing the next
38:21week. And oh, there's three things to
38:22do.
38:23And after a couple months, they've made
38:25nine decisions, and they spent all this
38:28time setting up and like,
38:30"You know what? This one looks better
38:31than that one. Just pause. This one
38:33looks better." And as humans, we will
38:36find a pattern in anything. We're
38:37amazing at finding patterns in absolute
38:40randomness.
38:41Right? So, as soon as you say, "Well,
38:43that one looks better,"
38:45you've lost the way. You will make bad
38:47decisions. So, whether it's a script
38:50added to ChatGPT or software does this,
38:52too.
38:53Don't do it by hand cuz you will find it
38:56gets so boring and tedious, you will
38:58make bad decisions.
39:01Now,
39:02the other way to get some really cool
39:04insights into data, right? Is by looking
39:08at Google has has now put out there,
39:11"This is your actual data by assets."
39:15So, you can now go and see, "Hey, this
39:17asset is used in 11 ads, and here's its
39:20aggregate data."
39:21Right? Here's its click-through rate,
39:22its conversion rate. This ad This line's
39:24in 57 ads.
39:27So, we have this.
39:29Now,
39:31if you have a large account,
39:33staring at 3,000 lines of this
39:37will not get you any insights because
39:39you're just staring at so much data.
39:44This is useful. So, what I've done here
39:48is I took my RSA data,
39:51I downloaded it.
39:53I added it to in this case chat GTP with
39:56a script that said,
39:58"Turn this into n-grams." And if you
40:00don't know n-grams, you're looking at 1
40:022 3 4 and 5-word patterns across your
40:05data, right? Across all your ads.
40:09And then I can download it in that
40:10conditional formatting.
40:12So, instead of looking at this huge
40:14thing, I can say, "Hey, this row this
40:17this line this pattern is in 131 ads.
40:21And when it shows up, I have a really
40:23low CPA.
40:24You know what I should I should take it
40:26note of that." Or "This line's used in
40:2891 ads. When I use this, my conversion
40:31per impression is amazing compared to
40:33other places."
40:34Now, like this was a senior living
40:36center.
40:37So, they have, you know, independent
40:39living, assisted living, Alzheimer's
40:42care, etc.
40:44And senior living centers can be really
40:46expensive.
40:47So, they had a lot of assets focused on
40:50price, insurance, government funding,
40:52and so forth.
40:54Right? And they had, you know, some ads
40:56that were based on calls to action like
40:59schedule a weekend visit, you know, come
41:01see us, talk to other residents, etc.
41:05And they had some assets that had
41:07reviews, ratings, and even quotes from
41:09like residents or family members.
41:11And they're like this is a pretty good
41:13mixture.
41:14Now, when we looked at the n-gram data
41:17for this account,
41:18it turned out that
41:21when they were talking about prices,
41:23their click-through rates, conversion
41:25rates, and and ROAS were significantly
41:28below the norm. They're actually really
41:29bad.
41:30When they looked at their calls to
41:32action of come visit us, it was about
41:34average for the campaign.
41:36When they showed ratings, reviews, and
41:39quotes,
41:41their ad their their data was amazing.
41:44Right? And they realized it
41:46the lot of people searching with them
41:48weren't as price sensitive as they
41:49thought.
41:50What they'd heard is a lot of horror
41:52stories about senior living, and they
41:53really cared about the reviews and
41:55ratings.
41:56Right? So, they said, "Well, this is
41:57just it's our RSA data. This was not a
41:59true ad test. We took, you know, our
42:01non-branded account and ran this.
42:03So, now let's make an ad test.
42:06Let's do one where we take our headlines
42:09and we pin it to headline one related
42:10to, you know, the keyword someone
42:12searched for Alzheimer's versus
42:13independent, etc.
42:15Let's take headline two, and let's make,
42:18you know, one set of assets that is, you
42:20know, five you know, 5,000
42:234.9 plus star ratings. We've been in
42:25business for, you know, 50-something
42:27years.
42:29One set of authority statements and
42:30ratings, and then another set on the
42:33calls to action. Come spend a weekend
42:35for free with us, right? See what it's
42:36all about.
42:37And then they ran this in a couple
42:40hundred ad groups,
42:41right? To see which one really does make
42:44a difference overall.
42:45And it it turned out that for, you know,
42:4885% of their ad groups or something, the
42:51ratings really really mattered. And
42:53there were a few groups like the
42:54Alzheimer's group that the come visit
42:56and actually spend time really mattered
42:58in in some very small subsections. But
43:01now they have this information, so you
43:02can completely redesign how they're
43:05speaking to people. And they realized
43:07that things like pricing and insurance
43:10were good site links cuz people cared,
43:13but they weren't the main focus of the
43:14ads. Right? And that's the kind of
43:16insights you can pull from this. Again,
43:18this isn't hard to do, and this is also
43:20going to be in PVC hub where you can
43:22download your RSA data.
43:24Um there's a prompt you can use, and
43:26then just add conditional formatting,
43:27the color scales across the lines.
43:30You have some interesting data to go
43:31look at, right? So, this is a way to
43:33pull insights, especially if you haven't
43:35done a lot of ad testing, but you have a
43:37lot of assets, and you can sort of learn
43:40how things have gone so far to give you
43:42ideas on on what to do.
43:46All right. So, you know, when we think
43:47about ad testing as a whole,
43:49we saw the brainstorm. What do we want
43:51to know? What are our assumptions about
43:53people? What do they think they want? Or
43:56what does our boss say we have to use?
43:58And let's see if that's actually good or
43:59not, right? Whatever it happens to be.
44:01And and then, you know, we say,
44:03"How do we pick winners and losers?
44:05What's our our winning metric, our
44:06confidence levels we want, our data,
44:08etc."
44:10We set up our tests. Now, we may use
44:12some of the LLM scripts to get some
44:14ideas for the test, which is great.
44:16Right?
44:16And then, all right, we use we look at
44:19our test results, and then we can make
44:21um analysis to say, "How does this
44:24really work? What is best?" And in some
44:27companies, depending on what you're
44:28testing, you might pass this to other
44:30teams.
44:31So, for instance,
44:33we often see, you know, we'll have SEOs
44:36test ads for click-through rate because
44:39your ad title is similar to your
44:41headline title that gets your headline
44:44that gets shown in the search results.
44:46We'll see others that are testing short
44:47calls to action cuz they want to know
44:49what button to use in the email they
44:51send out because an email button, right,
44:53is the same or a button in your email is
44:55the same whether it's in a display ad
44:57versus it's on an email.
44:59And so, there's a lot of other ways you
45:02can also use this data. I mean, when
45:04people went to bookstores a whole lot,
45:06which they still exist, it's not as used
45:08as much,
45:09we we had publishers who would test ads
45:13before they named books
45:15because a click-through rate of an ad,
45:18it turned out, was was relevant to the
45:21pull rate off a shelf of someone looking
45:23at a book.
45:24Right? So, when you're looking at sort
45:25of market research, there's a lot of
45:27stuff you can pull from your ad metrics
45:29cuz this is
45:31what people care about. What's say what
45:33they want, right? I mean, and it's even
45:35called an ad group.
45:38Because the ad's first, that's what
45:40someone sees. The keyword just say show
45:42me an ad or don't show an ad. Right?
45:44It's Google Ads. Ads first. Ads are your
45:47marketing, your connection point to
45:49users. Everything else just says,
45:52"Should I show an ad or not, and how
45:53much should I pay for it?" Right? That's
45:55all every other thing says.
45:57The ad is the part that actually talks
45:59to someone and says,
46:00"Hi, we're so-and-so, right? We can help
46:03you with this. Come be our next
46:04customer." Right? That's the ad part.
46:06Everything else is just supporting
46:09metrics for how the ad's displayed.
46:12Um if you are looking we we've been
46:13running software for years. We love ads.
46:16We love all kinds of making life
46:17efficient. We offer a 30-day free trial
46:20ad analysis. So, if you are looking at
46:21being more efficient in how you manage
46:24things or insights, come, you know, give
46:26us a try, too.
46:28And with that, let's move to some some
46:30questions and thoughts from everybody.
46:35Thank you so much, Brett. That was
46:36really insightful.
46:39Do you have time for a couple of
46:40questions until 4:00, maybe?
46:42>> Oh, yeah. Let's go for it.
46:43Amazing.
46:44Amazing.
46:46So, I saw one question in the chat
46:47already from Katie. Katie, do you want
46:50to unmute yourself and ask a question
46:52yourself, or should I
46:54repeat it?
46:55I'm not sure if you can unmute yourself.
46:58Okay. No worries.
47:00So, Katie basically asked, "Pinning
47:02reduces ad strength, and my managers
47:04thinks ad strength affects quality score
47:07{slash} ad rank. How can we challenge
47:09these types of beliefs?"
47:11How can we challenge these beliefs?
47:13That's education. That's why you're
47:15here, right? And even so, Google even
47:18put out a So, this is such a common
47:21misconception.
47:23Google, I believe on their ad strength
47:25page, it may be on their quality score
47:26page, in their help files even made a
47:29notation that ad strength does not
47:32affect quality score. And and Ginny
47:34Marvin, who is the Google Ads liaison,
47:36has even put up posts on like LinkedIn
47:39telling people these are not connected
47:41metrics. But it is
47:44I mean,
47:45users thinking that ad strength affects
47:47ad rank or quality score, it's so common
47:50of a misconception that I mean, there's
47:52even official words from Google on this
47:53one.
47:56Yeah, so just sending over that
47:58particular support article would
48:00would end the discussion, hopefully.
48:02Hopefully.
48:05And otherwise, you would need to bring
48:06some data.
48:07So, a question that I have, Brett. So,
48:10you mentioned some data property
48:12problems earlier on in your
48:14presentation, where you showed that 15
48:17unpinned headlines and 10,000
48:19impressions per month, it would take up
48:21to 2.5 years for each headline
48:24combination to get 100 impressions,
48:26which is insane.
48:28You can obviously pin or partially pin,
48:31uh but you can also limit the amount of
48:33assets, like limit the amount of
48:35headlines and description lines
48:37purposely
48:38to increase data density of of Yeah.
48:40What do you recommend, and why? Do you
48:42do you often choose between one of the
48:44two tactics, or do you sometimes even
48:46combine them?
48:47Yeah, I mean, that's a great question,
48:49right? So, in fact, let's we let's go
48:52put that slide up because
48:54I think
48:56right? This is so
48:58my preferences,
49:00right? If it is
49:03a let's call it a mid-sized account. You
49:05have some decent amount of data, but you
49:08you don't have brand ad groups that are
49:10getting 20, 30,000 impressions a month.
49:12Your your ad groups are getting 1 to
49:145,000 impressions a month.
49:16So, then
49:18I'm I'm going to probably pin three
49:20lines to headline one and three lines to
49:22headline two.
49:23Headline three is more about that site
49:25link strategy, right? Do I But I'll
49:27probably just add three more assets. So,
49:29I've got nine assets.
49:31Google's got, you know, 27 headline
49:33combinations. It's actually a little bit
49:34more to to play with,
49:36and that's a good starting place.
49:39Now, if I am a larger brand and I just I
49:45get a lot of data,
49:46then I I might increase
49:50the number of headlines I'm pinning. I
49:52might, but in many cases
49:55trying to think of five good calls to
49:57action, right? When they all do the same
49:59thing, download this paper, is sometimes
50:01really hard to come up with some ideas,
50:02right? So, sometimes I might still be at
50:04three or four because the others are
50:05just kind of like repeating themselves.
50:07They don't add anything new. Authority
50:09statements, I think are the easiest ones
50:12to have five or six good ideas to to do
50:15things. So, you know, with larger
50:18brands,
50:19you're often going to be in that, you
50:21know, three to five range. Smaller
50:23brands, two to three.
50:25And there are times you might only do
50:28one.
50:29That's often when it is a specific brand
50:32message, you're matching a TV commercial
50:34you did, and that really needs to be
50:36matching, you know, offline assets or
50:37something.
50:38But, I'm usually going to be in the
50:39smaller number for that.
50:42Mhm.
50:43Makes sense. And then rarely do I use
50:45all 15. I'm In fact, it's actually
50:47pretty rare I use all 15.
50:49Yeah, same over here. Same over here.
50:51And I think more and more people are
50:53starting to reduce their amounts their
50:55amount of assets within the RSAs.
50:57Especially now you can also aggregate
51:00the the data on the well, asset level,
51:04which is extremely helpful. I'm so so
51:06happy Google pulled that one off.
51:09So,
51:10when it comes down to
51:11the pinning strategy and well, well,
51:14testing with partially pinning, fully
51:16pinning, etc.,
51:18where do you start? Because you can you
51:20can test multiple things, like you
51:22showed in your presentation. Unpinned
51:24versus fully pinned, unpinned versus
51:26partially pinned, partially pinned
51:28versus partially pinned. I think for a
51:30lot of advertisers, it's hard to know
51:33where to start.
51:34Do you start quite aggressively?
51:37Uh for example, from a fully unpinned
51:40situation to a fully pinned situation?
51:42Or do you start testing waters a little
51:44bit and do things slowly? So, I I have
51:48gotten to the point that
51:50even
51:52crazy sized accounts, so this
51:58I'll pull this up here a second. So,
52:00this account is [snorts]
52:02using
52:0312 unpinned 10 headlines per asset.
52:07Now, they're getting over 100,000
52:10impressions, right? For This is two
52:12months of data.
52:13And this is a big number, right? That's
52:15a giant number of impressions, and
52:17Google still not getting it right. I've
52:18gotten to the point
52:20I rarely start an ad that doesn't have
52:23some pinning involved. Now, I wouldn't
52:25have said this a couple years ago, but
52:26I've just like after seeing so much, my
52:29default is three lines pinned to
52:31headline one, three lines pinned to
52:33headline two, headline three pinning
52:35versus not being a headline strategy.
52:38And I'm usually
52:41So, headline one being related to ad
52:44group is just it's worked for 20 years,
52:47right? Like there's that's not magic.
52:49So, I might even use the same ones in
52:52both ads I start. What I really want to
52:55know initially is what makes the
52:58difference to consumers. So, I'm usually
53:01going to start with that more theme
53:02testing of
53:04is this price sensitivity? Do I need to
53:06care about, you know, shipping? Or if
53:09it's a headline one, it may be like
53:10geography, so like an insurance. Do you
53:12need to add the location to the ads for
53:15insurance or something like that?
53:17If it's, you know, travel,
53:20do we are we talking it's cheap travel,
53:23therefore price sensitivity matters? Is
53:26it a large brand, therefore loyalty
53:28matters? Is it a place that consumer
53:31service matters more? So, I'm more
53:34likely to sort of sit back and say, you
53:36know, what's worked for you before, what
53:38hasn't worked for you before, but not
53:40believe those completely, right? That's
53:41just kind of ideas,
53:43and start with that theme testing of
53:46let's just figure out this high-level
53:48stuff first of how things go.
53:50Which is generally a headline two more
53:52of a headline two test in most cases.
53:56And let's say you're completely new to
53:58ad testing and you've you haven't done
53:59any of this before. Do you see like a
54:02hierarchy in which component types of
54:05what types of angles or even sub angles
54:08tend to win most often?
54:10>> so so Yes, so Yeah, so generally
54:15benefit statements and calls to action
54:17usually win more than anything else.
54:20Because benefits like that, what do you
54:22What's the consumer actually get out of
54:24it? And when you can combine benefits
54:28with, you know, emotional appeals, with
54:31pre-qualification or something, that
54:33really does well, right? If you can sort
54:34of combine couple ideas together.
54:37That being said, like authority
54:39statements sometimes are all the winners
54:41and so forth. So, I really want to know
54:42initially
54:44what's the bigger picture focus? What
54:47should it be? Right? On on those types
54:50of ads. And so, I'm going to start often
54:52much more with a multi-ad group sort of
54:54testing approach
54:55where my headline ones are probably
54:57going to be the same in two, potentially
54:58even three ads, depending on how much
55:00data the company gets.
55:02And then headline our ad one has, you
55:05know, three discount statements, and
55:07headline two has three or ad two has
55:10three calls to action of different
55:11types, and ad three is three authority
55:13statements, whatever it is.
55:15And get that initial idea of which one
55:18of these is, you know, best, middle,
55:20worst, or maybe there's two best and one
55:22worst, right? But, what does that look
55:24like initially before I refine more?
55:29Yeah, so basically big picture first,
55:31and then
55:32>> Yeah.
55:32>> And when you think big picture,
55:34it's not account-wide, right? It is So,
55:37I I love labels so much, campaign labels
55:40of like this is brand, this is
55:42non-brand, this is third-party brands,
55:44this is competitors, right? This is just
55:46plain stuff. And I'm not going to look
55:49at RSAs account-wide. I'm going to look
55:51at them by those segments only.
55:54Mhm. Yeah, makes sense. Makes sense.
55:57I've got one last question. I've got
56:00more, but I want to respect your time.
56:01And it it revolves around the future of
56:05ad testing.
56:06Because we've seen Google messing up a
56:10lot of our otherwise clean tests with
56:13introducing stuff like AI Max.
56:16So, obviously, clean ad testing
56:19requires, well, a lot of controlled
56:21variables and clean data, right? To to
56:24properly test with. How do you see a
56:27future of systematic ad testing in a
56:29world where Google keeps taking away
56:31controls and keep, well, messing up,
56:34well, where certain keywords are being
56:36triggered, stuff like that. Yeah, AI Max
56:39is interesting in that. So, I think as a
56:40as a conceptual level, ad testing is
56:42going nowhere. There are too many brands
56:44who care about their message, and you
56:47know what? You start messing with BMW's
56:50classic statements and so forth, BMW
56:53gets mad, and they spend a lot of money,
56:54right? So, like you have brands that
56:57have spent, you know,
56:59decades and sometimes over a century,
57:01right? Like refining their brand
57:02message. So, I it's going to stay there.
57:05Now, things like AI Max are are
57:08interesting cuz you've got, you know,
57:09really three components, right? One
57:11is the keyword expansion, which is the
57:13default, right? And that's where ads can
57:15suddenly show from different ad groups
57:17and so forth.
57:18I think there's potential world there,
57:20but and I'll cover that in a second,
57:21right? If we go like three minutes over
57:23here.
57:24Two is do I let AI Max write assets?
57:29Right? Or do I let PMax write assets?
57:33And
57:34right now, I believe their assets are
57:37terrible. And in fact, you know, ad
57:39strength No, not sorry, not ad strength,
57:41quality, right? Your asset quality.
57:44Where it'll say, "Hey, this quality is
57:45low, above average, whatever."
57:47Google will not rate their own. Have you
57:48ever noticed that, right? Like you can
57:50get a low, but Google can't get a low.
57:52They won't rate their own created ones.
57:54So, the fact I've heard rumors recently
57:56that Google's going to deprecate and
57:58remove that metric. Now they're showing
58:00stats by asset, which is which is great.
58:03Yeah. My my real hope for things like
58:06that auto creation on AI Max is if I've
58:09got a thousand ad groups,
58:12writing callouts and sitelinks for a
58:14thousand ad groups is It's terrible
58:16work. But, it really can help. If AI Max
58:19can go through and get it 80% right, and
58:21I can just clean it up, boy, that is so
58:24useful. So, my hope actually on on some
58:26of those things, I wish I could say AI
58:28Max
58:29write my extension/assets,
58:31but don't write my ads, right? Cuz I
58:33think that's that's incredibly useful
58:35for some of these things.
58:36The ad testing can get messy at times,
58:39and it it will.
58:40Authority statements within campaign cuz
58:43AI Max is only is ad group or campaign
58:45level.
58:46Authority statements, calls to action
58:48are often similar across ad groups. That
58:49part's okay. It's them messing up the
58:52headline ones,
58:54which definitely cause some issues
58:56because suddenly you're saying A, the
58:57consumer searches for B, and it's
58:59problematic.
59:01That And right now, it's not a big
59:02number. If if that number increases,
59:05then we may have to make some
59:06adjustments to how we read the data or
59:09try to look at the AI Max impressions
59:11and toss them out of ad testing. We're
59:13not there yet. It doesn't do this
59:14enough, but, you know, that's that's a
59:16potential change, too.
59:20But, I think as a whole, ad testing is
59:21going nowhere. This is
59:24been evergreen in I mean, the first ad
59:26was written in ancient Egypt like 5,000
59:29years ago, right? Or maybe 3,000 years
59:31ago. I actually looked this up once,
59:32right? Like when was the first ad? It's
59:34ancient Egypt and it was I forget
59:36someone who did something at their
59:37house, right? Wrote it on a cave. And
59:40and so
59:40>> they did ad testing
59:425,000 years ago.
59:43>> because I don't think so because I don't
59:44think when you're writing rocks you want
59:45to go and try to erase your your
59:47engraving, right? But I mean this is
59:48we're talking thousands of years, right?
59:50So yeah, ad testing didn't become bigger
59:52till a few centuries ago, but no, I
59:54don't think it's going anywhere. Just
59:56the way we do it may change some.
59:58Yeah, yeah, yeah, yeah.
59:59Great stuff. So Brett, once again, I
1:00:02want to respect your time. Thank you so
1:00:03much for doing this for all of us. If
1:00:06people want to learn more or maybe
1:00:08perhaps want to start a trial for your
1:00:10software, where do they need to go?
1:00:12So if you go to whoops, I just went way
1:00:14I scrolled way too far. If you go to
1:00:16adalysis.com,
1:00:18right? You'll see it and then you can
1:00:19click a free trial button.
1:00:21If you want to see I post a lot LinkedIn
1:00:24is the only social platform I actually
1:00:25spend any time on these days. So if
1:00:27you're looking for me, you search my
1:00:28name at LinkedIn and you'll find me and
1:00:31and you know, I don't post all the time,
1:00:33but when I do I try to be
1:00:36more interesting and insightful than
1:00:37random news.
1:00:38Yeah, you are. Especially with the large
1:00:41case studies you you did lately, which
1:00:43is amazing, by the way.
1:00:45Excellent, excellent. Glad you like
1:00:46them.
1:00:48Thank you so much. Enjoy the rest of
1:00:49your week and I hope to talk to you
1:00:51soon, Brett. Excellent, excellent. Thank
1:00:53you and thanks everyone for being here.
1:00:55Enjoy.
1:00:55>> Yep. Have a nice day, everybody.

About this transcript

This page contains the full transcript of Google Ads ad testing masterclass with Brad Geddes and Bob Meijer by Adalysis, generated from the public captions YouTube serves with the video. The transcript has 10,384 words across 1,694 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.