
Episode Description
In this episode, our Co-founder & CEO, Frederick Vallaeys sat down with Cory Lindholm to dive deep into the world of PPC data analysis.
Cory kicked things off by covering the essentials. Then, he walked through the process of performing an analysis to uncover valuable insights about seasonality, how your accounts are performing, and even correlations between different metrics.
He also discussed the importance of having a solid foundation in statistics and data when it comes to making informed PPC decisions.
There’s so much value packed in this episode. And we hope you find it useful.
Watch this episode to learn:
- The fundamentals of PPC metrics
- How to measure the impact of seasonality and other PPC factors
- How to integrate your PPC data with your business data
- The common data analysis mistakes to avoid
- The best tools for data analysis
Episode Takeaways
The Fundamentals of PPC Metrics
- PPC metrics like click-through rate (CTR) and cost per click (CPC) provide foundational data to assess ad relevance and budgeting effectiveness.
- Quality score remains a critical metric that influences ad cost and placement, emphasizing the importance of optimizing this metric to improve cost-efficiency.
How to Measure the Impact of Seasonality and Other PPC Factors
- Analyzing seasonal trends helps predict variations and prepare for future campaigns effectively.
- Use historical data to anticipate and adjust budget allocations and marketing strategies according to predictable seasonal performance changes.
How to Integrate Your PPC Data with Your Business Data
- Link PPC data with other business metrics to provide a comprehensive view of performance and ROI.
- Utilize data from different platforms (e.g., Google Ads, CRM systems) to improve accuracy in targeting and customer understanding.
The Common Data Analysis Mistakes to Avoid
- Ensure ample sample sizes to avoid bias and ensure data reliability before making campaign decisions.
- Be wary of confirmation bias and overfitting in data analysis to maintain objectivity and accuracy in your conclusions.
The Best Tools for Data Analysis
- Tools like Optmyzr, Tableau, and Power BI enhance PPC campaign analysis through advanced data visualization and reporting.
- Programming languages like Python and R are recommended for conducting complex statistical analyses and integrating AI to enhance predictive modeling.
Episode Transcript
FREDERICK VALLAEYS: Hello and welcome to another episode of PPC Town Hall. My name is Fred Vallaeys. I’m your host. I’m also the CEO and co founder of Optmyzr, a PPC management tool. Today, we have the great pleasure of welcoming a return guest, Cory Lidholm, and he’s going to be sharing with us everything he’s been doing with statistics to get a better sense of the world.
Of Google ads and PPC metrics. So we’re going to start with the basics like CPC and CTR, but we’re very quickly going to go into some really cool statistics and we’re going to see some really cool charts that are going to give you deeper insights into how your accounts are doing things like seasonality correlations between different things.
And really arm you right away to better analyze accounts and drive better performance for the clients that you manage. So with that, let’s welcome Cory.
CORY LINDHOLM: Okay. So. If you don’t know who I am, Cory Lindholm, adsbycory. com. I’ve been doing PPC, Google ads, Microsoft ads for nearly a decade, but really I’m a self proclaimed data nerd, data scientist aficionado.
So you will see foundational things in here, but you’re also going to be seeing some more advanced things that I really nerd out on. So I think it’s going to be a little bit of everything for everybody. In this slide. So yeah, let’s get right into it. So PPC data analysis, key insights and best practices.
So click through rate, what does it actually mean? It really indicates ad relevance and engagement cost per click. That’s your CPC. You’ll usually hear it obviously important for budgeting. It’s your average cost per click conversion rate. This is an indication of your landing page effectiveness, right?
Now, if we’re talking about sales teams, it’s going to be similar. It’s going to be The indication of your sales team’s effectiveness or your software being able to close that lead, whatever it is a cost per acquisition, the lower CPA suggests an efficient spending. If it’s obviously a higher cost per acquisition, it’s going to indicate to you that there are some inefficiencies in your ad spending return ad spent.
I’m not a huge fan, but it’s a very common metric to indicate the profitability. of your advertising spend and then impressions. This is really going to indicate your ad visibility, you know, brand awareness, things like that. Last major one that we’ll touch on is going to be quality score. Obviously, this is going to affect your cost per clicks and your ad ranking.
We used to say it affects your ad position. That’s no longer relevant. We don’t have that metric ad position. With that said, it’s still going to indicate or it’s going to indicate sort of the impression share or amount of times you’re showing up in relevant auctions. Hopefully they’re relevant. Those are more the more common metrics I would say.
FREDERICK VALLAEYS: Exactly. And so that ad position or the rank has been deprecated by Google, but obviously there’s still a decision process of which ad shows above the others, even if Google doesn’t name it necessarily that way. And quality score is a big factor in that. But like you also said, it’s a, it’s a discount factor.
So the higher your quality score, the less you have to pay to maintain that same sort of ranking on the page vis a vis the competitors. And then the rank how high up you are on the page that determines how many impressions you get CTR tends to be higher, higher up on the page. So all of these things kind of help each other. And I’m still a huge fan of quality score. I think it’s still one of the metrics that you can really use to optimize and get your costs down. Again, if you want to look at CPA, great, that should be your benchmark metric, but if you’re not happy with it, quality score is that lever to getting a CPA that you like better than before.
CORY LINDHOLM: Yep, and sometimes you, there’s nothing, it’s seemingly there’s nothing you can do with quality score. Sometimes you just hit a point of diminishing returns. You’re putting all your time into landing page optimization, add captive optimization. It just doesn’t seem to make a difference. But is it worth it?
Absolutely. I do get asked a lot. Should I even bother optimizing around quality scores? Is it really that important? And I always say, if you can get the same thing for a lower cost. Wouldn’t you do it? Well, that’s what quality score allows you to do. So definitely still. All right. So let’s get a little more nerdy.
So statistical foundations. Now guys, I apologize if the images are hard to read. I can definitely provide and will provide a slide deck so you guys can download this. You can zoom into the visuals. I’ll try to cover them as best I can, but let’s get into the foundations here. So I get asked a lot if I was starting out, what is the first, you know, in PPC starting over?
What are the, what’s one of the first things that I would learn? It wouldn’t actually be going through the entire Google Ads Academy. That might be an unpopular opinion, but actually the foundation of good PPC management is going to be a good foundation. In statistics, and I know you’re getting nightmare flashbacks of bell curves and high school and things like that.
But guys, statistics is what PPC is grounded in. When you log in and you see all these numbers, that’s statistics. It is it. These are indicators of performance for your business and your advertising. So if you don’t, have solid statistical background, how do you know whether what you’re looking at is statistically sound?
How do you know how confident you can be that the future is likely to be X? How, how, how confident can you be to say that the previous period you’re looking at is better than the previous before it and whether that’s enough data to make Sound decisions off of. So first point I’m going to mention here is the descriptive statistics.
So this is something we use all the time when we’re looking at a new data set, but as well when we’re looking at changes over time. So essentially, it’s going to summarize our data features as we like to call them. So mean, also known as the average, which tends to be the default values that you see in your Google ads account.
You’re click through rate, which is average click through rate. You’re going to see CPC, cost per click, that we’ve talked about. That is the average cost per click. Here’s the problem. Averages often lie, especially when your data is what we call skewed. There’s a large distribution. In other words, you’ve got outliers in there, right?
You’ve got one day where this particular campaign had a cost per click that was 25 and most days it’s only in the 15 range. Well, in that case, if you find that happens often, you actually don’t want to be utilizing the average when you’re, when you’re truly trying to track performance, you probably want to go to something that accounts for the outliers, which is going to be the median.
So quick little lesson. Again, this is why statistics and foundational knowledge and statistics can be really helpful for you rather than just trusting the default values that come with your PPC interfaces and software.
FREDERICK VALLAEYS: Yeah, and I mean, that also makes me think about data exclusions. So if you see any wonky data whether you know exactly why it was, maybe your landing page broke and so your conversion rate.
Was really horrendous. Or if it’s a seasonal thing that you weren’t expecting, if you see that there is data exclusion, and that’s really important because otherwise Google’s machine learning systems will think this is valid data and start basing decisions on it. And then automated bidding starts to overbid or underbid.
And it’s very easy. You can just set a trigger. That’s like, Hey, if we see something wonky, don’t look at that data. We got a problem. And then as soon as it’s fixed, you allow Google to reevaluate that data. Your bidding is going to be nice and consistent. It’s not going to be thrown off by what happened.
CORY LINDHOLM: I love that. I don’t think that’s a feature that’s talked about enough, the data exclusion. So when conversion tracking is broken as a prime example, if you know from this time to this time, this date to this date, conversion tracking was inaccurate. It’s best to go in there and tell these algorithms, I don’t want you to account for that.
in your bidding, right? I don’t want you to see that as you’ve made the wrong decision on an audience that you think is good or something like that. I want you to just get rid of that for now. It’s better to do that than leave it in. Also to add is seasonality adjustments. Again, I feel very underutilized.
I think a lot of advertisers and Freddie, let me know if, if you find this with, with with your crowd as well, that people, they assume that these algorithms are magic. They assume they just can predict and see, Oh, you have a sale going on. We’ll just account for the conversion rate and we’ll adapt accordingly.
I would rather, even if that’s true, communicate clearly to Google with the seasonality adjustment that we are having a sale from this date to this date or this time to this time. And I expect, based on historical performance or my own predictions, that conversion rates are going to go up by this amount.
That way, I can give it a clear start and end. It can formulate its predictions off of that. It can bid accordingly. But most importantly, which I feel is really underestimated, is once the sale is over and you’re telling it, I expect conversion rates to drop, It doesn’t all of a sudden, like if you hadn’t used seasonality adjustments, drop all your bids because it goes, Oh, I really messed something up.
And then all of a sudden you’ve got to work. It’s got to work its way back up in terms of bidding in order to feel confident to hit whatever goal you’ve communicated to Google. So those are really two important features. I’m glad you brought that up.
FREDERICK VALLAEYS: Yeah, totally agree. I mean, the seasonality, Google will certainly pick up on the big ones, like the holidays back to school.
It’s the same thing every year, right? But if you’re going to run that one special. Or even like Amazon prime day. I think it changes every year exactly when that happens and Google doesn’t know. And so if you’re going to do a special promo the system will pick up on it, but it might be two days until they do.
And if that was your weekend special, you know, how helpful is it that by Monday, they figured out that conversion rate is going to be higher because now they’re like bidding more aggressively, but your sale, your discounts are gone. So now it’s bidding more aggressive to get traffic. That’s not quite as valuable.
So it’s all a bit backwards. Like you said, if you can tell Google. This is going to happen from this day to that day. Then they can be ready for it and they can make the adjustments exactly as you want.
CORY LINDHOLM: Consider Google to be your advertising assistant. It is not your genie. Moving into inferential statistics.
So sounds fancy, but essentially just lets us know based on the sample data, whatever data we’re looking at. What is the confidence intervals? In essence, how confident can we feel about the statistics that we are analyzing, especially. Period over period or when you’re running a hypothesis test, you’re saying, should we go with a, should we go with the inferential statistic?
Again, I’m not going to get too into the weeds, definitely do your research, use AI to get more details on that topic. It gets really in the weeds on that. So I too crazy. Regression analysis. So understanding the relationships and predicting outcomes. I get this question all the time is, you know, if we increase spend, How likely is it that we’re going to get more conversions?
Essentially, you’re trying to say, will this thing moving in this direction correlate with this other thing moving in the same direction or a different direction? So for example, the most common, like I said, it’s going to be spend and conversions. If we increase spend. are conversions likely to go up? This is something that we usually use regression to try to figure that out.
Now it’s really tiny but on the right side you’ll see an example visual of what that looks like. Generally you’d use a scatter plot if you want to visualize this correlation. So on the bottom you’d see cost as that goes to the right, cost is increasing. On the left, on the Y axis is the conversions.
And as that goes up to the top, that is an increase in conversions. What you’re hoping to see as an advertiser, let’s say you’re looking at a a campaign that seems promising and you want to see if I increase spend on this campaign, are we likely to get more conversions? What you’d hope to see is all those dots, which could be dates could be ad groups that those are all closely aligned.
to that line in the middle, and that there is a nice linear trend. So that’s going to let you know that there is a high likelihood that as you increase cost, you increase conversions. And like I mentioned, there’s a lot of ways to use this. It doesn’t have to just be cost and conversions. There’s also things like multi regression analysis.
You can actually plug in multiple values to see how that affects other other metrics. Lots of fun things you can do with that, but it is the way that we answer that question if we do X. How is that likely to affect? Why? Great way to use it. A B testing, so we all as advertisers have heard of this before, but being able to compare two versions of something in order to determine better performance.
Now, for me, because of the way that RSAs, responsive search ads, have changed over the years and we just get impressions, which is great. Probably the most frustrating thing. I also think about ABE testing to previous period, a certain period versus a previous period. And essentially, you know, let’s say we made a core change, like a restructure.
We launched new products. We did a new promo. We changed the pricing of something from that date forward. versus the previous period. That’s also a type of A B testing. You know, was that thing that happened, that main event? What, what’s the statistical significance of that change? And can we really say, for example, with confidence intervals, that it was statistically because of That event, as opposed to just randomness.
So these are things that these tools can help us with.
FREDERICK VALLAEYS: Yeah. And then this is where generative AI is great because you can go and ask it questions about like, how do you do a proper AB test? Cause there’s a thing like one ended and two ended AB tests. Yeah. And the one ended, which basically says, what are the chances that this.
change made things better, but it’s not evaluating the chance that that change made things worse, which is the two ended. There’s a two possibility outcome, not just things getting better. And so you might actually end up in a situation where there’s a good chance that your test made it better, but likewise, there’s a pretty good chance that it made it worse.
And so then you might want to make a different decision than if you only looked at the one ended AB test. So this is very complicated. You but again, Jenny, I can sort of help you understand what are the pitfalls so that you don’t make those mistakes and end up shooting yourself in the foot.
CORY LINDHOLM: I also want to throw in the tip on when you’re using these chat GPTs or whatever it is, Gemini, you can.
Be careful about just feeding any data in there unless you’re using the enterprise level tool for example on open ai They’re going to use that data you give it within their training sets. So unless that’s okay with you Be very careful about anonymizing anything that you put in there before you do Especially if you got api keys in there, which i’ve heard of people doing this They just Think, ah, whatever.
You know, it’s just an API key. You do not want that somewhere publicly. So be very careful about just feeding any data set in there. I would really recommend anonymizing things as much as possible.
FREDERICK VALLAEYS: Yeah, exactly. I mean, so you made the key point here, which is if you’re going to use OpenAI. You are feeding the data into the model that’s going to learn.
And it may produce that API key in a response to someone random. And then now they’re using your API key. And then it’s sort of like in software tools, like Optmyzr, we don’t send any data to open AI until you enable your sidekick capabilities, your gen AI features. But then because we are using API, we are using enterprise.
That data is used to generate a response, but then it gets thrown away. It doesn’t get used to build future models. So it’s, it’s much more safe. But you always, you know, if you’re an agency or a contractor, you just want to acknowledge to your clients who, which software vendors you work with, where your data is going to go and validate that they’re not going to do something.
That you don’t want because your competitors having your CPC data would not be a good thing.
CORY LINDHOLM: Brad, I would love to put our tinfoil hats on for just a second. I’m really curious about, cause I know you’re an ex Googler, right? So you might be a little biased, but I trust you. I trust you. So with this new ability to feed Google our profit data and it being able to bid off of that, how do you feel are some ethical considerations and how advertisers should think about approaching that?
Thanks. Giving that data to Google, because there’s a lot of debate about that topic.
FREDERICK VALLAEYS: Yeah, I mean, so I still believe that Google fundamentally means to do the right thing, but they are also a for profit corporation. And so they are going to look for ways to drive more profit, right? And if they know that there’s more.
Competitive pressure they can introduce into the auction because people are making good money off of those clicks. Then, yeah, they might do things that make the auctions a bit more expensive, so you got to be careful with that. And that’s where we, as a third party tool, then come in and we say, well, listen, you can also bring that profit data to Optmyzr, construct some rules and some logic around how that should fit into your overall strategy.
And then the only thing that we sent to Google is the actual Cool. TROAS or TCPA, or in some cases CPC bid, which is informed by your profit. But Google never knows whether you’re bidding to zero profit and maximizing revenue or whether you’re bidding to like having huge profit for every sale. They don’t know.
So it does safeguard some of that information a little bit more.
CORY LINDHOLM: All right. Getting a little more nerdy guys. One, one slide at a time. So the role of data analysis and PPC strategy. So. I’m going to cover three main points here. One is going to be performance monitoring, then optimization, and budget allocation.
There’s plenty more that we could talk about. I think these are the core three. So in terms of performance monitoring and how we can use statistics and data analysis and these tools like you know, OpenAI, Gemini, these kind of AI tools to build visuals for us very quickly. Now, hint, all of the things you’re going to see on the screen are things that.
I built with my own proprietary software that I use for my clients and my clients only. So you’re not going to get these exact visuals, but you could pick them apart and probably try to get the AI to do something similar, but it is a little bit more resource intensive. So as an example of performance monitoring on the top right and the bottom left, sorry to make that confusing, on the top right, A couple of things you usually want to track.
One’s going to be that actual value, whatever you’re looking at. Let’s say in this example, it’s going to be return on ad spend. So what was the actual value of return ad spend over time? Now, this is what you would see as conversion value over cost inside the Google Ads interface. Great, but that doesn’t tell you everything, right?
You want to be able to extract more insights than just what was the actual value. So this is where the Using things like Python or software can really help you out. So there’s a couple other things I throw into this visual. One’s going to be the three month moving average. It does not need to be three months.
You’re going to have to customize this based on your data. But in this example, we’re using a three month moving average. So rather than just seeing the ups and downs on a daily basis of ROAS in this example, we’re looking at the overall trend when looking back 90 days the average return on ad spend.
Now, maybe you want to use a different metric, etc. But this is just a rough example. So it lets us know we shouldn’t freak out about short term fluctuations. Those are going to be expected. However, if it’s an outlier, you do want to be able to notice that and you want to be able to back that with statistics instead of you just saying, hey, it looks like it’s pretty high.
I think that’s an outlier. You want to use math to identify the outlier. So in this case, you can see that that red dot at the top. This is not just a high day. It’s an outlier day based on the data set that we are giving into this particular program. So you’ll be able to spot those outliers. You’re always going to want to look for that because that is going to influence your average or your mean.
And then I also like to see just a static mean over the date range. You can see right here it’s a little under three. That’s that dotted line. I also like to throw in there’s kind of overkill, but median does normally what I’ll do is I’ll put all of my major metrics in graphs like this so I can quickly scroll through them and I can see if there is a mean and a median big difference between them and actually see both of those lines to let me know.
There is a big difference in the mean and the median, which means I need to go into my analysis with that in mind for that particular metric. And then the last point I’ll mention is that gray shaded area. That is your IQR, also known as your interquartile range. This essentially lets you know what is the the central tendency of the data you’re looking at.
So between what range and what range is normal, right? Now you can customize this as needed. Generally it’s between 25 percent of your, your data points and 75 percent of your data points and anything outside of that is not necessarily a statistical outlier but is outside of what the central tendency of that data is.
So in this example, anything below that gray bar. is outside the norm and anything above that line is also not outside the norm, but in this case is a good thing because it’s a higher return on ad spend. So it lets us know roughly what’s the distribution of the return on ad spend over a given date range and lets us know whether something is falling in or out of that range.
Very, very helpful, especially when you’re looking at multiple metrics next to each other to see you know, what is normal? So when you’re looking at that moving average, you’re not freaking out over short term fluctuations, right? With
FREDERICK VALLAEYS: something like this, you probably also want to have a little bit of a breakdown between brand and non brand and different product categories.
So that because obviously brand and non branded have very different performance. And so what is normal would be nothing would be normal, right? Like, what is really high? One is really low. So everything would basically be an outlier. So you want to make sure that that’s not the case when you’re looking at your data.
CORY LINDHOLM: Absolutely. Yep. Always consider your segmentations. You don’t want to be, as you mentioned, brand non brand search with shopping, especially if you see dramatically different distributions of data or means, whatever. When you look at that descriptive when you run descriptive statistics, we were just looking at before and you’re looking at the mean, you’re looking at the max, the men of your different data points, you want to be bucketing your groups of data together, right?
So you want to be putting your, you want to see what’s the distribution, what’s the mean, what’s the median of certain metrics for your branded search. How does that compare? to your performance max campaigns, your standard shopping campaigns, your non branded search, your display campaigns. You want to put those things into different buckets accordingly because they are to Fred’s point, they’re going to have dramatically different means, medians, maxes, mins, standard deviations, really important to classify each and analyze those separately.
And then in terms of optimization. So, you know, obviously when you’re doing data analysis, you’re going to be looking at ways to refine targeting. Your ad copy, your bid strategies, lots of things you can do there. Obviously those are more foundational for PPC. But one of the things I tend to feel like it’s overlooked maybe because it seems so obvious, which is budget allocation.
But Fred, I can’t tell you how many times I’ve audited an account and they’ve gotten 80 percent of their spend to these campaigns that they really want to work, but they’re just not working. And then They’re limited by budget on these campaigns that are clearly winners from a standpoint of profitability, volume, potential, and like, have we ever tried putting more budget towards either?
Well, yeah, but we already know they’re winners. So why aren’t we scaling that? It’s surprising. So I feel I have to mention it, but. In order to utilize that within our data analysis process, this is where predictive modeling can come into play and factoring in things like external factors to PPC like seasonality and promotions, things like that.
But we can essentially figure out how should we, let’s say given an annual budget, distribute that budget dynamically throughout the year. And what would that look like in terms of number of leads we get? For example, that’s on the bottom, right? A good example of something, an analysis that I just did for a lead generation client.
They’re a SaaS company where we essentially took three years of historical data. That’s all we had available from their internal systems. And we were able to, with the focus of lead generation, predict to a very high degree of accuracy. how much budget of our annual budget we need to distribute throughout the year on a given month, and what is likely to be our lead volume for that year when factoring into seasonality.
Incredibly powerful way to utilize this if you just have the knowledge to get in there. All right, let’s go forward that just to really quickly say that bottom left. It’s kind of the same idea as the top right one, but this is a way to look at a previous period versus a current period. So that orange line current period versus the previous period and noting outliers really, really substantial in terms of insights because a lot of what we do is going in there and we look at, okay, since the last time I optimized the account, what’s the change?
What’s the comparison of values and metrics from that period versus a previous period or Compared to last year, whatever that is, really nice to be able to visualize those things. And then you can use your human brain to start picking apart the why, because that’s a really important piece here and using your domain knowledge to actually say, okay, here’s the what, here’s what’s going on, but why is it occurring?
Yes. You can use data analysis and predictive modeling to get in a rough idea. Yes. You can ask chat GPT or something to guess to what it might be. But even those technologies like the open AI will tell you Here’s what I think, but you need to use your expertise as a domain expert to really figure out what’s going.
All right, so integrating, I talked about this a lot, but integrating PPC data from other business data. I think a lot of times as PPC experts, whether you’re doing Facebook or whether you’re doing Google or Microsoft, whatever, we get so hyper focused and we start thinking in a vacuum and we just look at our data that we’re managing.
But it’s so important, especially in, in today’s. Environment of online advertising and marketing that you, you’re thinking on the channel, you’re thinking about how not just how the increase in spend, let’s say on Facebook influence Google ads. But what about the consumer confidence index, right? Is, is that possibly going to influence your sales?
These are things you want to be keeping in mind you know, is there a trend in the local economy that might be impacting sales? If you don’t know these things, you could think that it’s the changes you made that are causing it. when these other factors need to be also considered as well. So this is an easy example, again, of utilizing some market research to essentially pull in and compare and contrast the CCI, Consumer Confidence Index, in the states versus your PPC conversion.
So again, really high level, super simple. You can make it as complex as it needs to be, but the real core idea here is to integrate external factors into your PPC performance and don’t just think, Just about your Google ads data in a vacuum. You’ve got to be considering the whole picture.
FREDERICK VALLAEYS: Yeah. And I love that, Cory, because I think you really have to bring in your own business data.
And again, like I said, bring it as close as possible to what actually matters to your business. Like CPC, CTR, who cares? Like let’s focus on profits and revenues. And you can only do that if you connect your CRM. So, you know, Which lead actually led to a sale. And then in terms of like understanding, did I change something in the business that’s causing me to have better or worse performance while it helps to look at vertical benchmarks.
So how is everybody else in the industry doing, or look at tool like PPC investigator? So if you see that your conversions are way up, what’s the underlying cause of that? Is it because you have more impressions for the things that you advertise? And then you can start asking questions. Are there more impressions?
Like, is the news talking about this? Is this a trending topic? Is there some shift in consumer demand that’s causing them to look for my offering more than they were before? And so the more alerts that you can have to tell you when it is time to go and investigate, because ultimately it is a difficult problem, right?
If you think about Google and they have now, it’s been exposed, there’s 14, 000 plus ranking factors, but Think that you could also probably use 14, 000 plus things to consider for, does that impact my business, but that’s a lot of effort to go through all of these things. So at some level, even knowing when to start asking that next level of question, and then I think we’re agencies and consultants really earned their keep is because they often work with more advertisers in that space.
So, like you said, you’re just working with a SAS company. And they might say, Hey Cory, what have you seen? And you’re not going to divulge like that specific other client, but you’re going to say, yeah, we’re seeing sort of a trend where maybe the economy is a little bit better. People are more willing to spend money on software and this and that.
And so now you all learn from each other. So that’s, that’s really huge value.
CORY LINDHOLM: Absolutely. That that’s a really good point. It, I get that a lot, that question of, well, you know, should I just do it myself? And it’s like, well, you might be able to with this amount of data volume and your own background in this and your time availability, whatever you might be able to, but consider the competitive advantage that you have, if you work with an agency who has these shared insights across an industry like yours.
Like if you work with an agency that specializes in retail e commerce, you They might have insights that you’re just not going to have because they’re able to see those insights across and those patterns across different other, you know, other retail e commerce providers. So yeah, really, really important and a good reason to work with an agency or a specialist as opposed to just trying to do it yourself.
All right. So this is one of my favorite analyses to do. I do think people overthink this a little bit, so it sounds really complicated, but it’s. It uses, utilizing seasonality and external PPC factors. We touched on a little bit like with the CCI, but this is a little bit more in depth. So essentially we want to utilize seasonal trends to predict and prepare for variations.
So here’s the thing guys, when I say seasonal, It’s very easy to default to thinking, you know, autumn, fall, spring, summer, and winter. It’s not what I’m talking about here. So this could be seasonal patterns that you might have no idea about, that you really weren’t clear. You might’ve had some instinct, the longer you’ve been running your business.
But you you want to actually use data to uncover whether that is predictable, whether it’s just random chance, random noise, that’s what a seasonal decomposition analysis can do for you. So again, this is one of my favorite things to do for a company is to be able to take their historical data and be able to make predictions off of the data that we have, but also integrate seasonality into that, as well as get us an idea of how we should be, like I mentioned earlier, budgeting and forecasting and preparing maybe for promotions or inventory levels what we should be expecting given seasonal patterns.
So on the right side is what you’ll normally see with a default seasonality decomposition. So if you were to ask ChachuPT or something, you know, Cory said, do a seasonal decomposition on this dataset. I don’t know what it means, but just go ahead and do it. This will probably be the default output.
It’s going to be essentially Whatever the observed thing is, in this case, we’re using total leads, so it’s everything combined. It’s just what you would normally see if you were to graph out your leads over time. Great, not that useful, right? You could have done that yourself, no code required. The second part is where things start getting more interesting.
So this is the general trend of the data, right? So it’s essentially, it has removed the noise, if you’ve ever heard the saying, know the signal from the noise. This is the signal. This is the general trend over time. That third one is going to make a little bit more sense in a second. That third one is the seasonal patterns that we see.
And in this data set, you can see actually a pretty clear pattern, right? I think it’s around, yeah, June or so is when we start seeing a huge seasonal increase, but essentially we start seeing peaks and, and leads that are attributed to seasonality and not attributed to random noise. This is where things get really interesting, right?
Because this last part on the bottom right is the residual, just a technical term for the noise. This is things that cannot be explained by the seasonality in the data set. So for example, in the very beginning, you can see that blue dot that’s really high up there. That cannot be explained by seasonality in the data set.
Therefore, it gives you something to go, well, I wonder why, what were we doing? Were we running a promotion? Did we have a product that that we sunset that maybe, you know, we shouldn’t have. It lets us know, is it, this is the kind of the random stuff. The stuff at the top is to let us know this is the actual trends.
These are the actual seasonal patterns. Fantastic for again, predicting and projecting out into the future of how we should manage things like inventory, ad spends, et cetera, and what we should expect So I always say Fred to consulting clients as well. That you need to consider what Google calls your conversion cycles.
This is what Google can see in terms of your attribution journey. How many days from the initial click does it normally take to get a conversion? This can be helpful in deciding which of these data points are going to be more useful between your internal, between what Google sees. But also lets you know that there is going to be this expected lag.
And again, it depends on what you’re using, what tool you’re using, etc. But just to keep it simple, if you’re using Google, and it looks like your average conversion cycle is going to be four weeks, for example, you don’t want to be, you know you know, jacking up all your bids for June if actually everything starts.
Four weeks later. That’s when you’re going to be, because you’re going to want to account for those conversion cycles. It’s a little bit complex, but
FREDERICK VALLAEYS: have you written a GPT or basically the plug I want to make here is people write GPTs that are specialty built to do certain things. I’ve written a GPT to help you write scripts and I’ve fed it information about these are the five things you should ask someone who comes to you with a script request, right?
Because someone’s going to come and say, Hey, can you write me a script that helps me set better. TRO as a bit, but they might not be thinking about the look back windows. They might not be looking at seasonality. They might not consider you can do an MCC script or an individual account script. They may not have thought about the breakdown between brand and non brand campaigns.
And so there’s all these things that you and I and specialists know. And so you can actually build a GPT that forces that user to address those questions, or at least think about them. Think about them before to get the output and hopefully get a better end result because of that.
CORY LINDHOLM: Yeah. My primary usage of the GPTs of the world is to write Python scripts and to help me adapt those Python scripts and to challenge my own logic within that to check for biases, things like that.
But I do have my own little custom ones, again, proprietary stuff. But what I actually like to do, Fred, again, a little nerd out, but also humble brag is I like to have three that I build, so they all sort of, they don’t connect to each other, but I have one that gives me the prompts for the script builder.
And then I have a third one that QA is the script that’s being writing because I don’t lie, I find actually better results by having these three assistants that don’t really know completely what’s going on with the other. But it’s better because if you just have it all in one, it tends to get confused.
It loses the context. You have to repeat yourself. Even when you give it a very, very clear prompt. What I don’t want to do is have one that’s building a script. Then I ask it questions and I kind of get it off track. It tends to have an issue coming back and correcting itself. So I like to actually have three.
So I have kind of a custom thing that essentially uses three assistants to check on each other’s work and get amazing results for them. Yeah,
FREDERICK VALLAEYS: that’s brilliant. I mean, if you don’t have to pay extra for one more assistant and why not, right?
CORY LINDHOLM: Exactly. Yep.
FREDERICK VALLAEYS: Very cool.
CORY LINDHOLM: And then just to wrap up the seasonality, a couple of things that I really recommend to simplify this for stakeholders, because if you show them that stuff on the right, They’re probably going to get totally lost and not understand what you’re saying.
So put it in a language that they’re going to understand. And the bottom middle graph is the one I probably use the most. This is the average seasonal effect on, in this example, leads by month. So this would essentially tell us that on a given month we can expect versus the yearly average of leads, how many more or how many less leads in a given month do we expect?
And again, you need to be considering the quality of your data and conversion lags, et cetera, as we just touched on. But in this example, we can see in June, we actually expect 15 leads less than our yearly average in June, whereas in May. We expect 41 more leads than average. So these are again, ways that we can just quickly graphically see it.
It’s, it’s a very clear visual that even if you weren’t a hundred percent up on all this stuff and you didn’t have a strong statistical background, you can go, well, clearly that large green bar shows we need to be spending a lot more in May or maybe the, the month prior to our point about conversion lag.
In order to really capitalize on that clear seasonal effect that’s going on, and we should probably not spend as much in June or July, or maybe we just need to investigate why we’re seeing seeing that, you know, is that the time when our sales team tends to take a lot of vacations and that’s impacting some of our sales and you can run those regression analysis that we talked about earlier, if you have enough data to see me.
Is that, you know, let’s say sales team going on a vacation. Is that, you know, correlating strongly with our sales going down? It’s just a kind of a silly example, but just a rough idea of how you can utilize these things to think through this.
FREDERICK VALLAEYS: And so Google ads is very nice because they give you the data to the beginning of when you ran a campaign.
So you often have multiple years of data you can feed into GPT. But when it comes to something like Amazon ads, they have a 90 day Cap on how far back you can look. Have you ever found that you can leverage the insights from one platform to predict seasonality on a different platform?
CORY LINDHOLM: Oh yeah, absolutely.
I do this a lot. So this is what we call in the data science world feature engineering. So you grab. Whatever you can, whatever data is out there, Shopify, Amazon, Facebook, I want everything. And I want to run that through correlation metric or matrices to figure out statistically what metrics from what things need to be combined.
Does that influence the accuracy of my prediction models, et cetera. So most definitely I am constantly pulling from APIs to get as much data as possible to make really good predictions and, and, and come up with these solid foundations in my analysis. Yep.
FREDERICK VALLAEYS: Very cool.
CORY LINDHOLM: All right. And then the bottom left for all those who are curious, what’s the one you didn’t mention?
This is a really confusing topic, but I just want to throw it in there. It’s autocorrelation. So the, I’ll just keep it really short and sweet because we’re running out of time. If you find the dot is outside the shaded area, which in this case was three that tells us that every three months we have a negative correlation with the past three months.
In other words, every three months we expect negative performance in our leads, and then the three months following are, we actually expect a positive increase in leads. So these guys tend to see. that their their leads follow a very clear business cycle of quarterly. So every, every three months or so, we are, we expect to have a few down months and a few up months.
Now, again, you have to take with a grain of salt. It’s not always going to be the same up and the same down every three months. And it may not be as bad as the three months prior, and it won’t be a hundred percent. None of this is to say With a hundred percent accuracy. These are all estimates using mathematics and install a solid statistical background.
But I don’t want people to get into this and say, we’re going to be able to predict with a hundred percent certainty that may is going to be amazing because we did a seasonality decomposition. This just gives us our best inference or our best. Hypotheses or best estimate as to what’s likely to happen doesn’t guarantee it’s going to happen.
And so predictive analytics, we’ve touched on this a little bit. I just want to show a visual of what this might look like, but essentially with predictive analytics in the context of PPC, which I don’t see a lot of content about this, it tends to be You know, utilizing biology concepts or whatever, and it’s hard for people to, to understand how to apply it to PPC.
So generally speaking, we’re going to be trying to predict future performance using the historical trends, et cetera. One common example is to predict customer lifetime value. You’ll see this in a lot of data science agencies where that’s one of the first tasks they do is to take all of your products and as many features and data points as they can to try to give you a rough idea feel Of what your customer lifetime value is likely to be for a given product or a particular cohort, a cohort of an audience.
So that could be like a type of audience, people who buy premium products that start by buying this product. That would be something like a market, a basket analysis, something like that. But how we might apply this really easily to PPC in this case was lead generation. So what I utilize for this, there’s a lot of different predictive models you can utilize.
I’m a big fan as marketers of Facebook’s profit model. Fred, I’m sure you guys have utilized that to some degree because what’s awesome about this model is it takes into account. Holidays, which if you were just building your own predictive model, you’d have to kind of custom build that in and adapt it every year to the holidays and the dates that change Facebook’s profit model already integrates that into its predictions.
So this is a great way to see in this case, our blue line is the general forecast, but then the upward and lower bounds of that prediction. This is really important. It goes to our point about nothing being. a perfect hundred percent prediction. We’re, we’re hoping in this case, and we are likely to see that the actual value of leads is going to fall within that green line and the red line.
So for example, next year, we expect to see a really, really good June. It might be one of the best that we’ve had. And again, this is built on a solid foundation of clean data. And then there’s some things that won’t align with what we saw with the seasonal decomposition, which is really interesting. So when we look at historical performance, it shows that these given months are not so great, but when you utilize our predictive models, they might not align with them, which, you know, it’s a whole nother topic, but you will want to investigate why that might be and how accurate your predictive model is, et cetera.
But you can always try to optimize and you should be optimizing those models to get better accuracy over time. All right, let’s speed through these guys. So a data experimentation, hypothesis testing. So in this analysis, what we wanted to do is we wanted to get an idea of if we changed our top ad groups and the you are the URLs that are associated with those top spending and top performing ad groups What’s the risk and what do we expect to get in terms of revenue and return on ad spend if we went from blog URLs?
To product page URLs. This is actually a pretty common question for larger e commerce advertisers because, you know, we have been running for search campaigns, mostly blog URLs because we were told that was the best practice, but we never really tested it. This is a good way to get into hypothesis testing on something that’s clearly applicable for PPC.
So in this case, the first thing we did, one of the first things we did, So we looked at the general distribution of return on ad spend for both of those two things, return on ad spend for blog URLs versus product URLs. You can see the distribution is a lot more right tail skewed for the product URL. So the much higher return on ad spend for some of those, but most of the blogs tend to follow more of a normal distribution.
So tend to be hovered around the 0. 5 to the two X range and return on ad spend. So Greg gives us a general idea. Moving forward, what can we actually expect in terms of predicting return on ad spend? Now, that bottom right is where that comes into play. So essentially, this is going to show us on the x axis, the bottom there the horizontal axis is going to be the number of samples.
So this is a model that the more samples, the more times it’s testing this thing, what is the, what’s the distribution? How much return on ad spend should we expect if we use product? Versus blog return on ad spend. And the clear thing you can see here is that the blog return on ad spend as the samples go up, the amount of tests that we run in this model goes up, return on ad spend is generally much lower than the product return on ad spend.
So if your goal is to increase the efficiency of those ad groups. This might give you some cause to say, well, things look pretty good in terms of returning ad spend. But then what you need to do is follow up and say, well, what about revenue though? Because percentages don’t put money in the bank for the business, right?
It’s the profits. It’s the revenue. That’s actually going to keep us keep, keep the lights on. So you’re also going to want to run this through and say, okay. Now I know what return on ad spend looks like. Looks like product return on ad spend is going to be the winner for us in terms of efficiency, but if I do that same thing and then see revenue is just starts tanking with the more samples that I run, maybe not going to be worth the compromise, but really important to run these types of analyses.
So common data analysis mistakes to avoid. We have touched on several of these, so we’ll speed through it. But one big thing, and you’ve probably heard this since high school, but correlation does not equal causation. So we need to be really hyper aware of these things in our PPC data as well. Just because you’re seeing, you’ve run a scatter plot, and it looks like as CPCs go up, return on ad spend also goes up, which would be strange, but let’s just say for, for example’s sake that that’s the case, that’s the case.
That doesn’t necessarily mean that you increasing your bids and therefore your cost per click is going up means that it’s causing your return ad spend to go up. There’s other factors that need to go into that type of analysis. Ignoring small sample sizes. We touched on this at the beginning of the call.
Ensure reliable sample sizes. So this happens a lot with newer advertisers or business owners trying to do this stuff themselves. And they’re freaking out because they have 13 clicks in the last week on this particular ad group and the return on ad spend is awful. And we don’t know what to do.
Should we, should we go back to PMAX? Should we, you know, should we drop all of our bids? What should we do? And the first thing I’m going to look at is the sample size. Well, we’ve got 13 clicks, so there’s a lot of randomness in this. It could just be complete random chance. We probably don’t want to be making dramatic decisions with such a small sample size.
Now I can also understand as a business owner, right? You’re trying to pay the bills. This thing looks very unprofitable. But if you can take away anything from this video and from this episode, it is monitor the sample sizes and know what normal looks like for a particular segment. What does normal look like for your branded search campaigns in terms of return ad spend?
What is normal cost per click for a given ad group look like? It can be tricky to track, but we’ve already touched on how you might be able to do that stuff with Python and using software. You can try to keep track and, and benchmark your account and your campaigns, your ad groups, et cetera, your campaign types.
And that’s going to be really helpful so that when you’re considering how many clicks is enough before I make a good decision, Well, that’s going to come down to what is your normal conversion cycle look like? Is it three weeks? Well, give it three weeks first before you start making changes. How many clicks is in there?
How many clicks does it normally take to get a conversion within that similar product or whatever we’re looking at? Well, you need to factor that in as well. Generally speaking, most people will say 30 is the bare minimum. So 30 conversions before you can make a statistically sound change. Every account is going to be a little bit different.
There’s gonna be a lot of accounts that wouldn’t be able to change things every six months if they waited for 30 conversions. So you’ve obviously have to take that into account. There are statistical analysis there are ways to utilize statistics on very small sample sizes, but that is a very complex topic and you’ve got to be taking all of those insights with a grain of salt because of the smaller sample size.
That’s a really big takeaway. Confirmation bias, avoiding focusing only on data that supports. What you want to find this happens in reporting calls all the time with agencies where you hide all the stuff that’s bad and you show all the stuff. That’s good Like look at our click through rate.
It’s through the roof 35 percent month over month and you hit away the ROAS that went down by 40 percent you don’t want to be doing that But you also don’t want to be doing that yourself when you’re going into your analyses when you want to investigate and answer a question You Don’t go into it with a preconceived notion that, you know, I want to look for any evidence that supports that me starting and working on this account has increased revenue.
Fred, I know you’ve talked about this in your books, you know, the difference between descriptive and prescriptive, right? We’re all pilots here as data analysts. There’s going to be different levels to how you analyze your data. And these types of things like confirmation bias play a really big part.
FREDERICK VALLAEYS: Yeah, absolutely.
I mean, the data can tell any story you want it to. And so you can use GPT. And we actually use an interesting example of that. An Optmyzr with our psychic. So there’s one way that you can prompt it to give you. the happy story. Give me the good news based on the metrics that you can go to a client meeting with.
And that’s really helpful, right? Because it starts the conversation on a positive note. And, but then we have a corresponding query for GPT that says, okay, now tell me where I could have done better and how things could have been improved. And so now that the client is happy and understands that things are moving in the right direction, let’s focus on a few things that we could do even better.
But it’s always the same data and it’s just how you present it and where you sort of maybe filter things out or how you look at it.
CORY LINDHOLM: I love that. Okay. And then overfitting. So this is a little bit more complicated of a topic, but when you’re building models, you need to be considering whether the models you’re building are building can generalize to new data.
A really common example is when you’re doing clustering analysis, like K means clustering to figure out how your products in your catalog should be clustered together in different shopping campaigns. Really common analysis. So if you’re using a K means clustering algorithm that you’ve designed, You need to make sure that as new products get funneled into that k means clustering that the model can generalize well to those new products.
You just want to make sure that you’re not overfitting things and that it’s it feels like it’s a great prediction even though the math does not show that it’s a very good prediction. You need to be factoring those in when you’re evaluating the accuracy of your models. And the visuals on the right and the bottom are just things that you would normally look at when you’re trying to assess, you know, is something statistically significant, which metrics, this is something I built into my software that essentially only gives me the metrics.
When I look at a given period versus a previous period, what are the metrics that are statistically significant utilizing a P value? Now, again, we’re not going to get you back in high school statistics, but a P value, if we have, we want 95 percent confidence. 95 percent confidence. We want to essentially see a p value that’s 0.
5 or less. In this case, these metrics that you see, click through rate, CPC conversion by date, ROAS, those have all seen statistically significant changes in a given period since an event occurred. Then I can dig into my data and see, okay, I’m just going to be focusing on those metrics because I know that those have statistically sound data.
Those other metrics I should look at and consider, but I know that mathematically, those are not really ready yet to evaluate for that given event. So really, really important. So let’s wrap this up, Fred tools in tech for PPC data analysis. So, you know, if you’re not using Optmyzr, you’re crazy. The guys are building and innovating all the time.
Definitely be using Optmyzr and this is not just because I’m on the podcast. I really believe in the company obviously the founder is is is very intelligent guy. He knows what he’s doing So utilize Optmyzr for your pvc needs. i’m also a big fan for data visualization and reporting Because I want to be able to really customize stuff myself as you guys can tell i’m very into customization I love using tools like tableau or power bi So that I can tell the story of what’s going on in the data.
This is great for those weekly, monthly reports, whatever it is to the stakeholders that say, here’s the trends. Here’s what I’m seeing. Here’s what we’re going to do about it. Right? That is the power of data storytelling. We need to stop this thing in our industry of just showing a graph and expecting the stakeholder to know what’s going on.
They’re all going to have their pros and cons. Everyone asks me what’s the best one. There is no best one. Experiment. Do the one that works best for your particular needs. Google Analytics, we’ve talked about plenty. Spreadsheets, Excel, Google Sheets. I’m a big Excel guy because Google Sheets is too slow for me.
Obviously, if you got a script output, it’s going to go into a Google Sheet, but then I usually pull it into an Excel or Python or something like that. And the last one, Python R in terms of programming. Great for advanced statistical analysis. All the stuff that you’ve seen throughout this is great.
Things that I’ve actually done. These are recent analyses that I utilize Python for. So if they seem really cool to you and you’re really interested, I highly recommend learning R or Python. Generally, Python is a little bit more user friendly than R is. But yeah, that’s how you can get there is, is learning a little bit of programming and using chat GPT to help you along the way.
FREDERICK VALLAEYS: Nice. And thank you for that praise for Optmyzr and myself, really appreciate that and a slew of other amazing tools. One little tip when you use GPT to generate your Python code, save it, validate it, because if you’re going to rewrite that same code tomorrow, it’s going to come out differently and you never know who’s making mistakes.
And so. That is going to be the worst thing if you go to a client and you’ve used that analysis a thousand times and then that one day GPT has drifted and it gives you the wrong answer and you look like a fool in front of the client, right? So once you find it’s working, save it, and then you can start reusing it and then actually have reliable consistency in the output.
CORY LINDHOLM: Absolutely. And as I’ve mentioned, guys, if you’re going to go to that level where you’re going to have this AI write scripts for you, you need to know the difference between a t test and a ANOVA. Again, you’ve got to know these things. You’ve got to have the foundations because if it gives you a function, To utilize, and you don’t understand how that function works.
It’s it. I found Fred that it generally does not follow best practices for Python coding or for data analysis. Even when I’m using the data analysis plugin, it tends to skip things. It tends to assume that I know those things, which I do. But then when I look and I QA the code, I’m like, why would you use that test?
You’re like, that’s a terrible idea. But again, that’s where your domain expertise comes in and you need to know the foundations, or you could write this down. Great big script that you think is amazing, but it’s, it’s fundamentally flawed because you didn’t know those foundational principles of data analysis and statistical analysis.
FREDERICK VALLAEYS: Yeah. Super advice. Thank you, Cory.
CORY LINDHOLM: Of course. All right, guys, we had a slide on future trends. I know this, we’re, we’re totally out of time here. So I just wanted to really touch on it and show, you know, we do have advertising that’s happening inside of these open AI things, shopping ads, search ads related articles, et cetera.
But I did want to at least call out and mention, this is probably going to be influencing us as PPC marketers, right? The utilization AI assistance. We’re going to be seeing some changes probably in our search terms. And obviously as these ads are coming in here it’s going to play a part. It’s a very mobile first advertising.
So something we’ve been hearing about for a long time.
FREDERICK VALLAEYS: Yeah, I think it’s going to be huge changes, but I think we also have a couple of years to make that transition. So everything you’ve covered today absolutely makes sense. I just think we’re heading more to a world where your generative assistant knows so many things about you.
It basically contains your memory of where you’ve been, what you’ve seen, what you’ve looked at, what you’ve asked. And so the importance of things like what’s the keyword, what’s the audience, all of that matters much less than how the assistant understands you. And so that’s going to make it more difficult for us to really query and start seeing these correlations in terms of what we have done, because it’s not really us doing things anymore.
It’s. It’s the generative assistant figuring things out on our behalf as advertisers. And so how we manipulate that I think that’s a big question and we’ll we’ll have to solve that together but until then use all the Cory’s best practices here use a lot of these heavy duty statistics And have some amazing insights.
So, and so some of these things that you’ve showed us today, Cory, very advanced, others a little bit more easy to get started with. So what would you say in terms of people starting if they haven’t done statistics before?
CORY LINDHOLM: Yeah. It’s, it’s going to sound like we’re on a bandwagon of AI, but really I think AI, I mean, it’s a language first task task.
It is able to explain things very well for people. Honestly, you got to understand that there’s going to be some stakes in there. That’s why it says chat GPT may make some mistakes, you know, check the information, but it’s a great way to learn. It’s, it’s a huge advantage. We’ve never had in the past where you can go, okay, I want to learn more about applying data analysis to PPC and it can give you some starting points.
And then you can ask follow up questions. You know, can you elaborate more on. what this part means what is multivariate testing and then just start, you know, learning more about that and get into the deeper dive, start reading books about it. If you’re really that interested, start learning a little bit of code.
And I highly recommend when you’re, when you get the answers from chat, TPT, when it’s running an analysis, For you you can expand the code that it’s using. It’s using Python code. So again, it’s not magic. It’s just running the same code that I’ve showed you that I do myself that essentially you can start breaking it down and saying, okay, what code does it use the most?
And I wonder why it’s using that, you know, you can, you can take some of those and start applying those to writing your own Python scripts for data analysis. So highly recommend using those tools for learning plenty of good books out there. If you want to learn the basics of statistics, Naked Statistics, hilarious title, but it is a fantastic book on the basics of statistics.
He makes it really humorous and an enjoyable ride along the way. And of course, if all of this sounds way too complicated, you have no interest at all, you can go ahead and hire me. You can find me on adsbycory. com. Reach out if you have any interest in consulting account management. or training for your agency and staff.
You can also find me on LinkedIn. I post regularly there. I’m also on YouTube under Cory Lindholm, and I host, co host a podcast, PBC Unfiltered, where it is two specialists, and I mean specialists that are super nerds at this. We’ve been both doing it collectively for about 20 years, and we are completely unfiltered for better or for worse.
And we are guys that actually Manage the account. We’re not just content gurus who talk about the latest trend. We talk about everything and we’re at the front lines of this stuff. So if you want to tune in for a hilarious, but very educational podcast we post a video about every single one I’d say every week or so, give or take.
FREDERICK VALLAEYS: Nice. Hey, well, Cory, thank you so much for sharing all that amazing expertise and everyone for watching. If you want to see more of these episodes, hit the buttons at the bottom, subscribe and also put in the comments what you want to see more of, or if you have any questions for Cory, and I’m sure he’s going to be looking at those and providing answers.
We’re putting the show notes in the description. So if you want to look at the slides, you can find them right there. But thanks again, Cory, and I hope to have you back on for another one very soon.
CORY LINDHOLM: My pleasure. Thank you, Fred.