AI is capable of providing the best creative options for your ad campaigns, can suggest improvements to strengthen your ads before launching them, offers various options in automated bidding to strengthen the performance of campaign objectives, and with the help of accurate data input AI can refine the targeting of customers to further improve each campaigns’ ROAS (return on ad spend).
Sheldon Singh is an experienced digital performance marketing professional and a Google Analytics specialist with a wealth of technical expertise. Sheldon is GMP certified and has a BCOM Honours in Marketing and Psychology from Wits.
View the full video transcript
NJ Webinar | AI: The next step for digital marketing analytics (2023-03-28 10:02 GMT+2) - Transcript
Speaker 1: So Sheldon's Sheldon has been in digital for about 12 years. Six of it has been specifically on Web analytics, I think is one of the most talents in Web analytics in South Africa. So used to work at Click Maven. Then when we moved over to jellyfish when Jellyfish Cooper, take momentum now he's at FNB and he's a solution architect there. So he actually works with if in these Web analytics. So he's if there's anybody that can talk about women analytics its a team. Yeah so I'm gonna end it over to you, Sheldon over to you.
Speaker 1: Oh, just just some work or…
Speaker 2: Thank you, Betty.
Speaker 1: just hygiene stuff. We we have about 120 people that'll join today. So if you have a question we love questions. We want to session to be interactive, just type them into the chat and then Angelo will Austin to Sheldon at the end of each section. And we then we have a long Q and a section at the end as well. So like, please also the questions Angela for getting anything. We are recording the session so we'll sing out to take to everybody.
Speaker 3: That's great. And people can also send Emoji cons, you know, and let us know how you,…
Speaker 1: Yes.
Speaker 3: you know what you think how you feel about? There we go. Thank you so much, the content.
Speaker 1: Yeah, so it helps as a Yeah. So as a presenter sometimes you don't know how it's going or if people are enjoying it. So like a little emoji like helps. As a formation cool. Over to you Sheldon.
Speaker 2: If you send a sad face, I will try and keep my energy up, I'm Jackie. All right. Hi everybody Jennifer today? is we're going to talk about AI and media, buying the state of analytics, Google Analytics for chatgpt and your marketing data, and then some Q&A, like what he said, Mouse. Cool. So in the past couple years we've been talking about moving to a mobile first world. So everybody knows I feel like this is old news but
Speaker 2: World has changed. We are now looking at an AI first world. And what does that mean? So AI is a very broad concept. It focuses around enabling machine or systems to sense reason, I'm just reading of the Slide Act adapt and act like a human. So basically, just taking information from the world and giving it some sort of logic that a machine can understand. Machine learning, is where it's an application of AI that allows machines to extract information and learn from it. That's what it's called machine learning. So AI is the broader term and that's what we're going to be using for the duration of this webinar is just using that term. So anything, Information is being used to give better results. And we as specifically focusing on media buying because obviously that's a big part of what you do online is buying media.
Speaker 2: But the objective media buying is actually to bring down your cost per sale or to get sales. And what we want to do is use AI to reduce the cost per sale. And the media cost is obviously the biggest part that's going to be impacted by that. So why do we need AI Previously? If you've ever bought media, you would know that you have to use some keywords, choose your audiences, write some ads, choose where you want to show it, Choose when you want to show it and on, what do you want to show it? But these were the ones that these are the variables that you would choose by default. These are the ones you had majority of your control but there's other factors things like the time of the week Cross device conversions operating system. The browser the person was using whether it's on an apple with it in a mobile browser, if there was a similar audience that you could talk, there's a lot of other factors and It's going to be near impossible for a human to make all these decisions. So what we want to do, And this is just an example of
Speaker 2: A Google study where the difference between a user in Berlin versus a user in Hamburg. There was almost a hundred percent difference in their conversion rates a user on a Saturday. Doing the search was the user on Sunday. As such, there was a huge difference, desktop was a smartphone, and then you can go quite granular in that uses in Berlin on a Saturday. And then just the device difference can make a huge difference to your conversion rates. So you obviously want to maximize your conversions while reducing your costs and all of that does rely on making sure all the factors are accounted for and this is way ai, comes it. Um, you take all the factors that take part in your auction and your search media.
Speaker 2: Combining it with your leads and sales data. And as per you put in, this is your job as the human. We do still need humans, we all need jobs. You need to import that lead to sales data into the system, combine it with all these factors and that's where you get your lower cost per sale.
Speaker 2: Any questions so far any comments? If I schedule it, the word ai, you probably in the wrong webinar.
Speaker 1: Let me let me just jump in. So so like so this is it's this is AI and that's with machine learning. But the whole idea is to remember, like like AI's is is not valuable. Unless like there's a specific business application and in this instance it is to lower the cost per leader of cost per so so it's like that's what we want to achieve, but the way we do it is slightly more technical. So sorry back to you Sheldon.
Speaker 2: Now I love, I love your voice in my effort. It's great. So looking at Google ads and how to use AI to achieve your goals is There's two parts do to to such, there's the creative. So, what you actually put into the system and then how you target. So what you're going to talk about first, is the creator, there is something called responsive such as in Google Ads where you supply 15 headlines and four descriptions into a single ad. So, previously you would write one ad with a single description in a single headline, and then write a VaR. In all different combination of it. Google Ads does this for you. So it's taking a lot of the manual work out, but then by providing these variations Google is able to test And show the best ad that it believes to the most appropriate user. So what that means is, if a person's such is very specific
Speaker 2: To one of your headlines, it will then show that headline. Or if it's according to all of Google's data, if on a Saturday at three o'clock, this specific combination performed better than the other one, then it will then show that one. So that's the using responsive search ads, which is just making the AI power that's in Google ads, more available to you, while trying to reduce your costs and improving your conversion rates or you, click to rate, we then also have A report that comes in that shows these are your different assets. So whether it's a headline and then it shows you what the words were and then compares them to each other. So AI driven marketing. So these are some of the ads we use for this webinar. And you can see here that the best two headlines. Headlines, were ROI from Data, Driven Marketing and Marketing Data Science,
Speaker 2: A AI driven marketing not as good. So what we can do is this is an indication that we should replace the specific. Like, with something else.
Speaker 2: Then also there's ad strength which is based on Google's experience. So as you make changes to your ads it will. Then tell you add more headlines or add different keywords. Make your headlines more unique or make your descriptions more unique and what that does is it gives you this little circle. Then tells you how good your ads will perform or it has how well it's predicted to perform before you even go live. So this is a really nice thing. If you are getting into updating your ads and want to be more relevant, this ad strength indicators, very useful, We then have automated ad extensions. So what this does is
Speaker 2: Is there's a bunch of different things that you can add to and add. It's not just the headline and description, you can add things like your phone number, your callouts your location. If it's a physical store and what Google ads will do, it will take that information and have it available. If the user's search query is relevant to a location. So easy example, if you are searching for A, a shoe shop and you want to shoe shop near you. It will then show you the location extension. If you want shoe shop contact details insure you the phone number. It then adjusts the query. It adjusts the result to the query to make it more real.
Speaker 2: The big one that we want to focus on is automated bidding. So you have a couple options, there's six on screen, you can select Google Ads to change your bud for your auction to maximize clicks. What that does is we'll then increase your bird to be more competitive on a user that it believes is more likely to click on the ad based on the create. You can also then use Target impression, share where you can tell the system. I want to reach a certain number, a certain percentage of the impressions online, It'll then try and maximize your bud or Adjust your bit to get you to do that target. You can then also maximize your conversions. So based on the experience that Google Ads has from your results, will then adjust your bed accordingly. Then enhanced. CPC in the same way to get you. Maximize conversions. We'll try and get you the best cost per kit that you can get the two that are really, really important. And super valuable is target CPA and target row ads. So CPA is cost back acquisition.
Speaker 2: Roetz's return on admin. so, if you have lead data or sales data, and you import it into the system, you can tell the system to adjust the bud to be more To be more aggressive to find you the best prospect acquisition or to reach a specific return on admin. Those two are really cool. They do require Send the sales data to be present in the system. Does anybody have any questions or comments?
Speaker 1: and I just want to make a mike comment, so, I'm going to use Roland that sells wine. So, Rolander idea of this is you let the machine decide how much to pay to to get that MIA list. Customer like, he's going to buy a case and you want you want the system to go and identify that great customer. A To do this is to give the data to Google ads and it'll then go and find that that customer for you.
Speaker 1: Charlene is a question for you. From Stephan the Kock you want to know, is our difficult, is it to get the sales this data into the system?
Speaker 2: So it's technically not that difficult, it is just a file import. The more complicated about is getting it probably out of your system in the right format. So what you need to do is if a sale comes through and which I try not to get too stupid technical here. But what you need to do is you need to have the identifier of the admin user clicked on that's a GCL ID or DCL ID, depending on what they click on and that just needs to be linked to your sale. You can also upload raw sales data and it will try and map it based on time that it happened, but that one's a little bit more. So it's possible. It's not that difficult. Yeah, I think we can include that when we send out the email.
Speaker 1: Yes, the phone is Olga. Get some links for you.
Speaker 2: Cool. Any other questions?
Speaker 4: Yes. Yes, I have one question. I'm running,…
Speaker 2: Back.
Speaker 4: I'm running a search campaign in the Middle East region. The search volume of some keywords are not very high. So, you know, I'm running this The campaign using maximize clicks on this building. So, using the maximized clicks, you know, I'm getting the good number of clicks and impressions on my on my target, clicks on my target keywords and on and and ads. But overall, the main goal is to generate the number of leads. So, you know, our last two weeks ago, I switch in my campaign from maximize Spectrum, maxine's conversion, but when I switch to maximize conversion, the overall impression system, probably reduce And the conversion is also reduced using, you know, using the maximized conversion as a wedding salary. So what you suggest I mean, should I also I'm in a separation bank move to maximized leak only, because using the maximize conversion, I am not getting good number of conversion.
Speaker 2: so, I think so,…
Speaker 4: Even not beginning.
Speaker 2: one of the caveats of this of using these budding systems, is that Need to provide it with a significant sample size. So what might be happening is if you're not getting enough, conversions in the 28 day period, I think it's it's 28 to 30 days. If you don't reach the threshold, it it's like trying to guess using two examples. So if if you want getting enough conversions into the system, Google Ads isn't able to give you an accurate. Prediction on whether it will convert or not in that situation I would recommend going back to maximizing your clicks try and get more information on your On your conversions to Google ads and then trying again. You can also have two campaigns with two different. Betting, strategies and try and compete them against each other. If you, if you do have the budget too but I would actually just move back to maximizing clicks.
Speaker 4: Okay. Okay, sure.
Speaker 2: Cool.
Speaker 2: Yeah. Thank you very so yeah. David you It'll it will only attribute the leads coming from Google it. To match.
Speaker 5: I get cool guys. I was just thinking we do a lot of lead campaigns on different platforms and maybe trying to get that data into sheets and trying to import that into Google. So, I was wondering if that might might work.
Speaker 2: Yeah, it does require the GCL ID. Which is a specific goal.
Speaker 5: Okay, thank you.
Speaker 2: Cool, cool mouse. Cool. So photography pa kind of jumped a little bit ahead, but what you need to do is To create an event that captures your lead conversion. So, as a user converts on your site, you want to send that specific data point to analytics as a conversion. So that the system knows this ad led to a conversion. The next part is to import the lead conversion data into Google ads, and that's where the GCL ID as well as depending on how. You want to share with Google, you can share the name, the value of the sale as well as they uses email address. And then once you have enough information, you can then change the book strategy to target CPA target or as all the other three depending on what you need. So,
Speaker 2: Facebook also does a very similar AI implementation. So in the same way, they have an advertise a bit, they have an ad quality. So this is obviously how much you're willing to spend on it on a click as well as the Predicted predicted performance often add based on how much content is in it, what the image looks like. But then Facebook uses something called the estimated action rate. And what this is this is the AI. They look at the user. Individually to say Sheldon is interested in these specific topics. He's clicked on so many ads before he is also engaged with so many posts. He likes he clicks on the Like button or he comments on specific posts and all of those interaction points. A combined into something, they call the estimated action rate.
Speaker 2: And then that will determine how much value, you get out of your ads and how much you should be paying for the click for that user. It's very similar to the the Google Way. It's data is call it, estimated action, so, the data details, so First things first, is whatever data you're collecting, please make sure it's clean data. You don't want to be putting any misfire tags or something. That's next slide. You don't want to be putting any dirty data into the system because what you put in is what you're going to get out. So whatever you're collecting, please make sure it's clean. Internet. Then you can report on it and then analyzing it. So correct. Label and formatting. So something like sending the value of a sale needs to include the currency. It's a simple mistake. I've seen it multiple times but it's something that just needs to be done. Once you have that analysis, you can activate your your data in
Speaker 2: Also Google ads in this case, where you want to add these conversion points in conversion values and then start using them in your your body. And then once you go around, you get, then start collecting more data and keeping them, keeping it clean.
Speaker 2: So just on clean data three things to look out for is a very common things that I've seen happen before is double checking. So what sometimes happen is a user will or person who set up the campaign we'll put a page tracking as well as a specific event tag on the. Thank You page. What then happens is the system then receives two indications that are two conversions happen and that that's just going to lead to inflating. Your results, must firing tags Sometimes. You could put a specific trigger on a tag that doesn't file on your Thank you page or fires on multiple pages, not just the last Thank you page or missing conversion points. So something that also happens is if you are creating campaign pages you need to make sure that every campaign Page has a specific conversion point and if that conversion point doesn't exist, Google Ads doesn't know that whatever traffic was sent to that campaign page actually was valuable. So, you do need to insure that that's done as well.
Speaker 2: Any questions comments? Anybody want to share anything?
Speaker 1: And a sheldoner. I think I just want to know. I like to focus everybody thinks AI's magic and I also have your issues and and the short answer to that is it is Absolutely amazing. But it works on data and if your data is not in place, it doesn't work. It's not magically going to fix your tracking is not magically going to give you sales unleash, you give it the right information. You can't do CPA campbelling. If the data is not correct. It's, it's like I've seen this many times where people go, Ai, Oh my God, this is amazing. It's like all we'll do do is Will let Google go and find the customers. Like, if Google doesn't know what a customer looks like, it's not gonna find your customers. It's very important. I'm sorry. Somebody asked somebody raised and Question.
Speaker 2: I see in the chat braylene. Thanks for the question. Can you apply the same principles for Facebook ads for Instagram? Yes, again. 100% The same logic applies on both sides? And there was a hand.
Speaker 1: And yes, a link to to learn more about that.
Speaker 2: Cool. Did someone have their hand up?
Speaker 3: I think they were two, people that had their hands up and he's got ahead.
Speaker 2: This we forgive you. Okay. So advanced, so if you don't like technical stuff, you can look away or just grab your cup of coffee for the next two minutes. I just want to share Stuff that I think is very valuable. If you don't want to hear this, you more than welcome to to look away now. I was like, Public Service, Announcement, Cool. So this is out of Google Analytics. Ga4 guidelines. It is their recommended events.
Speaker 2: For setting up on different websites platforms and different engagement points. So, what you want to do is this left side is the recommended events for all businesses if it applies to you. So if you don't have virtual currency, don't have a virtual currency event. But the big ones I would highly recommend is if you have a login on your website, if you have a purchase event that is super, super important, if you only set up one event that is the event you need to set up and if you have something we're using engages with your content to reach any sort of step you, you can then map them out as different steps.
Speaker 2: This side on the right hand side, is your ecommerce events, Google recommends, these naming conventions and you'll see purchase exists here as well. This naming convention is super important. If you use this naming convention for your events, you then are able to access the machine learning and the predictive models. That Google Analytics 4 has. If you had to use purchased my item, it is not going to work. The system doesn't know what the word is. It's just looking for purchase. so that's something really important to Bear in mind. So like I said, if you have these specific events, you can then start using things called predictive metrics in Google Analytics 4. And what this is is if you have a purchase event
Speaker 2: Google Analytics will if you have a thousand returning users in 28 days that have purchased something. So this is where it is just unfortunate prerequisite for the sample size is a thousand purchases in 28 days will allow the system to predict whether someone else will purchase based on their behavior. So what that means is if you have a thousand sales Google will be like Okay cool. This use is predict is doing the same sort of actions on the website. They more likely to purchase. The opposite is also true. If you have a thousand people who've not purchased on your website in 28 days, Google's able to predict Okay, cool. This users doing the exact same thing that all these other thousand people did and we'll and we'll give you a predictive metric of this user more likely to jump. The one that gets really, really cool is if you have predictive revenue. So if you have a purchase event and you send back the value to Google Analytics Group, Google Analytics is able to then predict how much a user based on their actions is more likely to
Speaker 2: Spend in 28 days, which is really, really cool. You can then also start using that information to put on. Return on admin. That's essentially where it gets to So, once you have this information and you have this predictive metrics activated in your system, you'll then get these predictive audiences, which you can then target in Google ads. So you get likely seven day purchases predicted, 28 day, top spenders. So these are your high value people. Sorry, you're likely seven day purchases, these are the users, you'll probably upsell to or give them a sweetened offer. So if you know those users about to buy but is a Hesitant, you could then.
Speaker 2: Give them a specific campaign offering to say, Come through and buy your top spend is you obviously want to upsell to the users who are more likely to churn in the next seven days. This is where you want to Re-engagement campaign. This is very useful. If you know a sale is coming up and it looks like you're trying to purchases are increasing. You can then re-engage with them users not visiting your site that's where you just do standardly marketing just to remind them that you exist and then likely first time in seven days but this is where you can. Also use a sweetener. You'll see that the screenshot has this not eligible to use. This is because There hasn't been enough. Events for this account. Cool. All right, those are looked away. You're welcome to come back, please come back.
Speaker 2: Any questions so far?
Speaker 2: Would you say it's better to stick with the Google recommended events and conversion problems in that cons. So I would highly recommend sticking to the recommended events if you can. But at the same time, if you have a KPI that's not any of those ones, then you do need to do a custom event for that specific Touch point or whatever it is. So if you have newsletter sign up there is Is Sign up. So the Sign Up event just don't call it newsletter, for example. So you can adjust these to be more in line with what you use on your personal side. David Vagina.
Speaker 5: Yeah. Now, I just thought it might be easier to just ask instead of typing.
Speaker 2: Girlfriend.
Speaker 5: So, for instance, I've got a client that Scott, four different branches, different Med areas across South Africa and then each has their own landing page on the site. So if you go sorry for your job, okay? So, and now we're tracking, for example, the individual contact form. So those are we set it up as contact form submission, Cape Town, contact form. Submission choberg, How does that that effect or How does the AI, the algorithms and South Africa? Those type of Context, phones, are those taller relevant? You can say, you've been for using the AI or is it Kind of a problem.
Speaker 2: so in that situation, In this in that situation, I'm going to directly contradict myself. So if you have a signup, let's just call it a sign of pole. Just have one sign up event that fires that you're going to use for your audience building in and…
Speaker 5: Yeah.
Speaker 2: AI, and then have another event that you can fire that specifies. A signup for Cape Town. This is a sign up for Job. Then you can have your own reporting version and…
Speaker 5: Okay.
Speaker 2: then still have leverage the AI in the system. What you can also do is these events, do not include the parameters so you can send a purchase event or sign up event with a parameter, then specifies. It's Cape Town job or whatever and then you can then use the AI on the sign of the event but then do your reporting on the parameters.
Speaker 5: Okay, cool mixing.
Speaker 2: But yeah, that's what I recommend. Hi.
Speaker 2: so, Leilah, I'm a little confused by your question. Do you want to voiceover?
Speaker 10: It wasn't a question. It was just from the baseline events. You need to use the recommended naming conventions.
Speaker 2: Yeah.
Speaker 10: Google says they recommended but they're not recommended. They're compulsory basically Just of the word recommended is giving some confusion.
Speaker 2: It's recorded it. Yeah. Okay.
Speaker 10: Another country.
Speaker 2: the air quotes, make a It's the big difference.
Speaker 10: Yeah.
Speaker 2: No 100%. So it is recommended just done in that you cannot use them. You can use whatever you want. It is a hundred percent your version, but yeah, it is compulsory It's recommended.
Speaker 2: Cool. Okay, I hope that other people came back. The truth about Google Analytics. So Ga4 is a special beast. It is an amazing data model. I will be honest. It is possibly the best thing they could have done. It just wasn't released in the best way. I don't work for Google so I'll say that freely. But what has happened is in Universal and analytics or ga3, you used to have a page view and event and then a bunch of different directions. Um, now all of those are just called events. So you get a page view event. You get an event for events that you specifically, Ai the things like your sign up your purchase. They all come through as event. Your e-commerce and transaction also comes through as an event that is the purchase event. You then also have events that come through from your app which also Team screen views It is a really, really good model. Just because now everything's on the same level, and you can then see user journeys correctly.
Speaker 2: But the problem that came with it is Google just released and said Do it. Um, they're forcing people to move over. So, on the first of July this year, Your current Universal analytics will not collect any more data, but Google will be forcing ga4 on some people if they haven't done it. And what that's basically going to do is just collect a Page view event for them. The. Intricacies or the customization that's being done. Will not be carried across. There's no expectation that it will be Google. Actually does say in the What's it in the Help center that it is recommended that you do a customer implementation. So we are being forced to move across but it is also good opportunity to clean up and make And check the things you actually want to track.
Speaker 2: There is a really big asterisk. Next to the data at the moment. Is that in GA 4? There is a 14 month, retention period, by default. It's sent, it's set to two months. So you do have to go into your settings into data, retention and change this. What this means is after 14 months, your user level data will not be available. You'll be able to see that uses visited your site and what pages they view, but the associated cookies and news identifies will not be available. I've heard something.
Speaker 2: Cool. So what you need to do is just go in and adjust those to 14 months by default. That's it, too. And that's kind of what the reports will start to look like after 14 months. There will be a flat line on certain data points, you're aggregated data, so you're page views and events will still exist. But your user account will start dropping like that, which is a little concerning. One of the. Options is adjusting how you report. So there's three different versions of reporting that you can do. You can use the Explore tab, which is built into ga4. The upside is that it's free. The downside, it's not that easy to use. The.
Speaker 2: Recommended option is local studio. It's free, it's easy to connect, but you do get quota errors and limit the data retention. So this is obviously what I just mentioned. Now, The quota area is Locust Studio, has a limit to how many times you can ping, the Google Analytics for servers in, in an hour. And if you ping it more than the recommended amount, you will then be For that out. So, if you have a really complicated dashboard like the ones that people saw clients, yeah, it's not that fun. The last option, which is what we recommend overall is Lucas Studio combined with Google Bigquery. It's a Bigquery, is Google's Data Warehouse. It is a cloud service that allows you to
Speaker 2: To on their servers, it does come with the cost so that's one of the downsides. It's not an exorbitant cost, depending on how much data you send. Most people won't be spending that much, but it is something to just be aware of. It is a little bit more complicated to set up, but the best part is that you own the data. Once it's stored there, it is, yours, it won't be deleted after 14. there's no quota errors, you can ping the system as much as you want because you are paying So let's be honest, Google did this, we have to pay for it. It is part of the game that they're playing at the moment.
Speaker 2: Cool. Any questions comments so far?
Speaker 2: Then anything.
Speaker 1: Just just on.
Speaker 3: There's nothing.
Speaker 1: You know, good Angela.
Speaker 3: Hang on as stephanus. What is the estimated cost to use Bigquery?
Speaker 1: I can answer that, so we bolted. Dashboard for a very big separati in the country, very complicated with user journeys and stuff. And I think we estimated that it would cost about $30 a month. So it's like 500 and for like, quite big complicated dashboard. So we we think that most most advertisers, Because like it's in the hundreds of rands, it's not. But if you if you big and and you and you put lots of reports.
Speaker 1: Like you, let's say, like, Take a lot like that's gonna that's gonna escalate quite quickly. That I think the nice thing is you you have a choice of it, is the quota eras, you can subvert, you can get around that by buying Google Analytics 360, which is about one point two million and a year or you can you can pay 500 and a month using bigquery and all your data will be there forever. So we looked at it from a commercial and technical point of view and we just think that Bigquery is much better.
Speaker 2: And the nice spot is once the date is in bigquery, you can use it for other things because it is, stored in SQL. So if you have any data scientists, they will probably love you even more. Cool.
Speaker 2: so, experience sharing So like we said, we built these dashboards user-friendly dashboards in local studio, permanent storage of data. So we own the data now, all the client owns the data and It will be there until they decided to delete it and once they have enough data, they can start doing their own. Like, machine learning models on that data because they own it. And once it's there in the system, it's it's there for you to use and it's AI ready because it is in a very fundamental sequel. Yes, in a SQL format.
Speaker 2: And then this is my personal experience. This is what Ga4 looks like. When you open the hood it is super super complicated. And what I actually think you should do is just choose the KPI. You need. The, the KPI, you need to just be that one dial, telling you, whether you've passing of anything and if you don't have that KPI. So if it is more than Page Yusuf, it's more than conversions. Then go into the customer implementation Your dashboard. Shouldn't look like this, it should just be one or two things to know whether you're going the right direction. That's why they use the compass back in the day. Right? Mr. Fisherman
Speaker 2: Okay.
Speaker 3: So we we've got two questions. I just don't want to lose track of it,…
Speaker 2: Okay. Profit booklet.
Speaker 3: because it's part of this conversation right now. So Stephanie's had a follow-on question. Can you then roll up different domains. If you have multiple domains across different regions across the world,
Speaker 2: So, the way it works will actually determine on how your ga4 is set up. And the reason for that is there is a one-to-one bigquery project to analytics property. So depending on how you set it up and whether you have data streams versus analytics for properties will determine whether you can do roll-up across different domains. It is, it is just a an architectural thing to decide before you go and make them.
Speaker 3: Aquila has her hand up a killer. Do you want to just hop on and ask your question?
Speaker 6: a long time. Okay, good. Good to be part of the session with your Sheldon. I know we've had a lot of discussions in the past about moving to Jay for and whatnot. Sam, you're just speaking about, look at studio, obviously that's like the a new data studio for GE further…
Speaker 2: Yeah. Yeah.
Speaker 6: if I'm not mistaken. So I just want to understand Walt Data Studio, sort of completely fall away. Also first of July, and you have to sort of migrate to local studio as well.
Speaker 2: So you Data Studio just got rebranded, under local studio. So, if you had a Data Studio report about,…
Speaker 6: Okay.
Speaker 2: Four months ago and you go to that same URL, it actually just switches to Lucas Studio. There's no yeah.
Speaker 6: Oh, okay. Okay. No.
Speaker 2: So The only thing that's happening on the 1st of July, is the old. Google Analytics, Universal Analytics will not collect any data. So if you Studio report, Look over data is based on Universal Analytics. After the first of July, it'll just beer.
Speaker 6: Okay, good to know. Thanks Sheldon.
Speaker 2: Cool.
Speaker 3: Very, very also had a question just quickly would bigquery also be a good option to implement for storing historical, Google Universal Analytics data. So you don't lose it when Google completely moves across
Speaker 2: It's a tricky question because Universal Analytics to export Bigquery required a Google. Analytics 360 License to have the connector available so if you are interested in moving to GA 360 or the paid version, that's it's a good option to do it. I would be a little hesitant just depending on how much data you have. So the query the bigquery export is and I stand on the correction but I think it is 10 billion hits. On a single export or 13 months. Worth of data, whichever is smaller. So it will look, I just depends on how much data you have sending it back into Bigquery is nice to have Google has said that you should be able to access your UA data for about a year, still after the first of July.
Speaker 3: Just to date.
Speaker 2: Cool things happen.
Speaker 3: Everyone know. I have posted both foodie and my email addresses. So you know if there's anything that isn't addressed during this webinar, he's
Speaker 2: I'm going to take a leap Lerato. Yes, it is just a rebrand. Oh, I see what he did. Respond. Cool buddy.
Speaker 1: Okay, I'm gonna like over quickly.
Speaker 1: This is the second. Okay. So so the next the next thing to just talk about is is a team of been struggling. In terms of artery, we make Google Analytics for usable. And we've been experimenting with something and we wanted to share this with everyone in the webinar. This is Experimental. But we encourage people. So we wanted to just share this with you. So, chat GTP and and marketing data. So We were like, Can we use strategp in terms of interpreting and reading, and using your marketing data? So the first Thing is a big disclaimer. This is on the on the Open AI website. And basically just going to explain it. So chat GDP runs.
Speaker 1: You know, is the Is the is the it's a language model and these two ways to interface it, Most people have been using interfacing it through the the web interface. So this is where you go into Chat GDP and you start a new chat, you choose the model, you type it in and you've seen it. Just the thing with with using this interface is it can be used to develop and improve their services. So what you basically doing is if you sending information via this interface open, AI can use it. It's in the public domain. So so what I want to just say is it's like if you have since the information don't post it posted in there, safer ways to use the API. So you can say, sign up as a as an API user.
Speaker 1: By like a very small fever, but for engagement it is more complicated though. You need to get somebody in that understands Python to actually talk to the API. But in terms of playing around with it, you can obviously anonymize your date. You can you can change the names etc but I wanted to just give you an example. So
Speaker 1: So this webinar. So a big business question is waited, we receive webinar sign ups from and this is a question that can apply to to anybody in the call. So we did my wine cells come from wait at my banking sign ups come from. We did my So, this is like a A normal business question that, you know, as a marketers get asked. So, so we we did an experiment. So we said, Okay, so the first thing you did was Can we use chat GDP for this and to answer that question. So, the first thing we did was we do some context training. So, We basically going to chat GDP, I was just using the interface here because the data is really old. So so you are now the chat GTP. But please answer questions around digital marketing, Do you understand? It's like, Yes. I understand. I'm the 9. John GDP bot program to answers. What I
Speaker 1: so we are basically we we are we teach If we are, we're doing prompt engineering. So, so to just give it some background, it's very important. When engaging with chat GTP is to ask, Do you understand? So it's almost like building a LEGO tower. We give it information. We say, Do you understand? It says, Yes. Let me give it the next block. So what we named it was I went into Google Analytics 4, which is an almost unusable and I pulled all the channel data for where we received workshop sign of from. So This is all all our channel, so it's like our email direct Google ads, Google SEO.
Speaker 1: Paid links Webinar, links, LinkedIn, etc. I think most people in the course should understand what what this means and then I've got some data columns you have got new users and then the most important year, one year in is He's workshop registration events. So these are the way we got most of Administrations from. So this is the data.
Speaker 1: To answer this question we did we receive webinar sign-ups from that's, that's where it says, but for normal business people to understand is sometimes complicated. Then it's not, it's not exactly friendly. Like, What does Google forward slash? CPC means, Oh, it's Google ads. So for example said, This is just a, just a, You know, we feel that the titles are friendly. So what we then did was we say, Okay cool. So let's take export this data into just a CSV or the limited formats. It's like it takes format and we basically gave it Entered it into chat GTP. So here is our chat. GDP interface. So this is where I trained it. I said, Here we go. So this is a little bit more information about nitro so just has got some context.
Speaker 1: And then we said I'm going to share some marketing data from Google Analytics. 4 Do you understand again and you say yes. And then what I did was I cut and paste it the data in the limited format into the interface. and then what was very interesting is like Yeah so I understand and it then Said and there Now you can ask it questions about the data. So, you Say so here picked up it say it's looking at the data to Beer's, not just email campaign was the most successful in terms of engaging users with engagement rate of 34%. So, it Clearly. Just from that statement, it can read the data which is very interesting. And now I can also questions I can say out at Bing Do so if I look at the data and said You're being Produced one visitor, That's it. And three veins. So I can then I mean it says
Speaker 1: Okay, based on the data provided being generated one user so it actually was able to pull the data up. So now that I've loaded this thing I can do a little live demo and live them as always break. But I could basically, for example, ask it How did?
Speaker 1: How did Google CBC? See do. Let's see what it answers. Here we go. So it says Your Google CPC generate, 54 new users, 26 and gauge, visitors engagement rated 28%. So I think What's interesting and powerful about this is is you can ask it questions in English and unstructured way and it'll, it can look at the data and give you an answer back. Now, in terms of the application of this, this is quite interesting. So I wanted to give some examples so There's a company Odd Moore that we have with it within the in the past. They do these amazing ceramics.
Speaker 1: so one of the things that we can do is we can give a chat GDP that the data on who has viewed, these very high end expensive ceramics and it can then tell us which users actually looked at it if they've logged into the website, So obviously, all the data needs to be anonymized, it needs to be like a code, so it'll be like customer EBC 77. But the point is that you can then give this to the artist or to anybody at odd more than they can ask, Who's look that? Well, right and can tell you, that's one example, Another example, time bank is once somebody's logged in and the business user is busy browsing around on the website.
Speaker 1: You can then give the, the client service, people and interface and sell. Like what? That fatty based, actually look at in terms of other banking, services, and can also in English. a last one will you can give it the data on what people have visited on the Woolworths website and you give the interface To the buyers and you can sell like, which code was the most popular in? April and it would then give you an answer in English. So the interesting thing that we we are playing with is like, how do we empower? Marketing people to use marketing data using this very easy interface. So, that's just something we wanted to share with everybody.
Speaker 1: If you want to learn more about how to use charity, before for marketing. And for marketing success, we have a web In May that we're going to talk about this Angela, if you just share that link in the chat at the amazing. Cool. So I'm gonna give it back to you so you can just you can present again and just put it on a qi screen and then we can open it up for for questions.
Speaker 1: Any questions around this church gtp thing.
Speaker 1: This element.
Speaker 7: Hi fatty, I wanted to ask in Sheldon the image, the images that are generated by AI. I was playing around on various websites like Journey and we're not quite at the point where we can run an entire campaign just by feeling inputs into into the AI. How far away we away from that essentially?
Speaker 1: I'll jump on that one. So so Roland. We we were, we were looking for images on or for our website, using Dali and the issue that we found was You give it an input and then it would, it would then predict what you would want. So it's like this predictive model. So we, so it would then give you four like four, you know, four different. So images, so you give it one instruction give you four different images. But what was challenging for us was the treatment in the style of each one was different. So you get like four different ones and and I think I think the challenge that
Speaker 1: Found was We want a specific style and treatment and then work it backwards almost like This is the style that we want and so give us different versions of that and in different interpretations. So. So I think at the momentum we've played, I've played extensively with my journey. I think it's amazing but I I don't I like we struggle with With that specific issue. I did find another image generation tool where you can start and you get the treatment and then you can work it backwards. So, I've signed up to the waiting list, but I haven't been a been granted access yet, so Roland. I think we, I think somebody's gonna jump on this in the next month. so like I think it's like six months to a year and then
Speaker 1: Then we'll have a have something in that space. So yeah, I definitely think it'll be solved in months or not years. That's what I think.
Speaker 7: And for me,…
Speaker 1: What if what?
Speaker 7: that's the most scary. Scary thing. That's on the horizon, it's going to change. It's gonna change the game. You just won't know what's real, any, any more and what's not. And you'll be able to set a professional campaigns with almost no budget and…
Speaker 1: He. Yeah,…
Speaker 7: compete against the big guys.
Speaker 1: I know it's it's it's
Speaker 1: Some it's an amazing opportunity that levels the playing field and I think this is especially playing out for me between Google and Bing. We Google is Big Microsoft and…
Speaker 7: Up.
Speaker 1: being was like Yeah, let's go because like they don't have anything to lose, they're 5% market share of the search engines. And Google has got 95%, so they've got everything to lose. So that's why they're going very slowly. So I think this is an interesting you know, technology because as you say, the small people can start competing with You know, established businesses again. So, I think this is a huge Opportunity for people to, to become way more effective. Is wonderful saying that I heard was.
Speaker 1: AI is not going to replace marketers market is using AI is going to replace marketers. Not using AI. Great.
Speaker 3: So we have three questions. We've got one from Georgie Darby, put his hand up, and then We've got one from Naruto. So I'll start with Georgie. It looks like, if you correctly structure your queries to chat GPT, you can actually add your brands info to be picked up by future users. Am I right in this? If so how to go about this?
Speaker 1: Oh, that's a good question.
Speaker 1: So in terms of queries, I just want to split the question. So in terms of queries and talk and and using jatt dtp, I think prompt engineering is is going to become a thing like being able to effectively work and into data into jgp like very beer from you, like a skill and and we are slowly learning on how to do this this Checking, for understanding, is it's critical to that, okay? In terms of making sure that chatted people reference your information. I don't know how. So, for example, I've shared NITRO'S information now. So, anything that's on the open web, I think. And when was it 2021 chat GDP already is.
Speaker 1: So, anything new I give it is helping them train them model. But I don't know, I think it's helping them trying to model in terms of how people use it not necessarily to build up the knowledge base. So that's that's a bit black box like Georgie I don't actually know but what is going to become a thing is, How do you, how do you make sure that your company's information? Do you have beyond chat gtp? Right? All being spot and I think there's something very interesting because you've got the Google Bot. That's spiders, your site for issue purposes on Google and he's a Bing Bot. That's quite issue your website for that. I think the Bing Bot is becoming is becoming way more way more, valuable and important. So, I think there's going to be a really emphasis in terms of Eco and how your website is indexed online by the Bots because that's the same data source that they're going to use in the Bing Bot. So, That's a thing.
Speaker 3: All…
Speaker 1: Angela.
Speaker 3: then Lerato's question. It's pretty tackled. That's a named Darby you can hop on. Let me just raise Lerato's question. Can AI generated contents rank on Google Search or Google, not penalized websites that use such content.
Speaker 1: Sheldon, you want to give that a go. You want me to answer that one?
Speaker 2: So it it's a depends on. So so if you create the content and you correctly, using it on your site, it technically will ring whether you have ownership. It does depend on which platform you're using to create it because some platforms, if you created in their system, they are the owners. I'm not sure which ones And I know they are ones that you have to pay and then you have complete right? But just use it with caution because if you created in things, like my journey, very That uses the discord.
Speaker 1: Hmm. Yeah. Yeah.
Speaker 2: So with Mid journey. If you created in in the mid journey platform, everybody can see it. So, it's not like a private creation, like you're not creating it just for you. You creating another people can view it. But if you do put it on your site, it is technically the industry that, like, normal content for images with content written by AI, they are apparently already, not penalizations, but it does flag that it's not written by human. Coming. So,
Speaker 1: Oh, I'm just too latch onto that. So one of the key things that Google look at, in terms of content, when it looks at content, it spiders is the author. so, Authors become very important to state. That who is the author and when it was written and, and the author could link to the buyer page of the author. So there's a authority, that comes with an article that was written, so if you don't specify an author, It's like you seeing yourself up that it may be in the that it was written by Jgdp. So I think like the way that I I view churches, be at the moment I meet with a, with a really good friend of mine. She's it of communication, one of the big mining hours, and what was interesting. She writ
Speaker 1: Of content. And now she started using Jatt DP to write it, but the essence of the communication is still controlled by her. And the words and the mechanical aspect of it is done by Jgdp. And I think what's interesting is, is more time, can be spent on the creative and communicating that our concepts and the research and less time on the actual mechanical writing board. So I think if you take the same approach to online content, in terms of ranking for SEO, If you use Jgdp.
Speaker 1: But you give it amazing like, bullet points. And you structure the thing accurately, you are the author, your, you know, you you citations. It's just this brilliant piece of content. Frankly. It doesn't matter if it was co-written by Jackie GDP, right? Because that was the mechanical aspect. So I think the the quality aspect will will become more and more more important because there's so much there's going to be an explosion. There has already been an explosion of content online. And what's happened? People still gravitate back to business day, people still gravitate back to New York Times in Washington Post. So this and with strategy to be, there's going to be even more content online. And again, it goes back to like, what's the authority?
Speaker 1: I hope that answers the question.
Speaker 3: Dobby, do you want to? Did you have a question? Because you did leave a comment around prompt engineers. I don't know if there was something else to ask
Speaker 5: Yeah, yeah. So getting up going back to what Rhoda asked earlier about? Doing fully. Full campaigns using AI. And what I've seen people do with maturity is actually tasted it where you prompted for but it it's it all depends on what you say, but what we prompted was specifically very specific details into what we want, the logo to look like and then you tell it in the style of this famous design. And rattled in and it was actually amazing. What was put out? What we got from that. So it what I've seen is people actually using it on the stock footage websites with some. Some of those profiles are completely AI generated content.
Speaker 5: And what we've seen is you can pick it up by some of the hands, They were a lot of issues with the generating on hands and that's how a lot of people immediately. Pick it up and with church, GPT also read up, Don't know very few, guys are awesome mod upgraded. They was a blog. I think it was on search engine journal. One of those where they say that Google say that they all maybe penalize people for using AI, just using AI like prompting. Like I said earlier using AI, the Chat tpt or other complete blog without anything. So it's true with what you said they is if you engineer it you educated and what you want and give it a bit more detail. In you might not be penalized but they all software as well that picks up if it's AI generated or user generated apparently, I've heard of that as well.
Speaker 5: So that's just one of the things that I think people should be aware of. Just look out for it's training those the AI systems to understand you more.
Speaker 1: And cool.
Speaker 2: So, 100% David. They I've also been hearing things that AI is now being used to pick up AI content. And now, we're going down a weird rabbit hole of ethics. Just because…
Speaker 5: Yeah.
Speaker 2: if if I wrote it and you wrote it in the same thing, if I wrote it an A and AI wrote it, we're actually on it. But yeah. It's gonna be a very interesting time.
Speaker 5: Cool.
Speaker 2: Any other questions?
Speaker 1: I think that's it. Angela's anything.
Speaker 2: Think so.
Speaker 1: Any other questions?
Speaker 3: No, it doesn't look like it. People are just saying they need to jump, and thanks so much and a lovely feedback. Thank you to everybody for your feedback. It really helps to hear, you know, how, how you you've taken. The all the insights in and yeah, and we're looking forward to seeing you at the next one. So if there's no more questions, we will be sending out an email with the dick and the video recording, and thanks for your engagements. So I would say That's a rep.
Speaker 1: Yeah. Thanks.
Speaker 2: Thank you, everyone.
Speaker 8: Thank you so much everyone. Bye.
Speaker 9: Thank you. Bye everyone.
Speaker 1: Just Georgie by.
Speaker 5: Well, thanks. Just, you got the next one.
Speaker 1: Cool. Pleasure Lora.
Speaker 1: Cool James. All good.
Meeting ended after 02:34:31 👋
Mar 28, 2023
You have been subscribed. Now add it to your calendar.