Analyzing Data Points for SEM Optimization
Hello fellow advertisers! Welcome to another episode of my video blog series. Today I will be going in-depth on how to analyze different data points for SEM optimization. We’ll find insights into the value of a few different variables within your AdWords Account. This should help you properly determine how much weight to allocating towards any given column of information. Enjoy!
Hello everybody and welcome to another episode of my video blog series. I am your host, Nick Rennard, and today we are going to talk about analyzing data points for SEM optimization so let’s get started. We’re going to be talking about … Looks a little silly here but we’re going to be talking about something we call Squirrel vs Ninja. This is something that I took from a blog years ago. A guy made a reference of separating data points as labeling one as squirrel-types of data points and one as ninja-types of data points. We’re going to go over how to determine which is which, how they’re useful, and I’ll do breakdowns in some of the other slides here in a sec.
Let’s go over them really quickly. A squirrel data point is not necessarily telltale of overall performance, they’re generally only relevant in a vacuum, when all the other variables are not in consideration. Some examples of these would be something like clicks or average position or click-through rate or even things like impressions. If you think about a click by itself, let’s say you’re analyzing the performance of a keyword, if one keyword has 10 clicks and one keyword has 5 clicks it’s hard to determine, actually it’s impossible to determine, whether or not one of those is actually better than the other. That’s what we’re wanting to determine when we’re analyzing our data is what keywords are performing, what campaigns are performing, and the end results are important, what’s not performing. Only looking at something like these 3 statistics you have here, clicks, average position, click-through rate, you can’t actually determine the performance of any given keyword or campaign just based off of these stats which is why we call them squirrel stats.
The follow-up here would be ninja stats. Ninja stats are always asking the question, “So what?” If you were to look at a campaign and it says it has 250 clicks, a ninja would be asking, “So what?” Like, “So what if it has 250 clicks? What does that mean? What’s the follow-up there?” A ninja stat is going to be opposite of a squirrel stat in that it’s going to be much more telltale of the overall performance. We use ninja statistics as our KPIs, our key performance indicators, to determine whether or not these campaigns are doing well or not doing well. Generally speaking, a ninja stat will have a denominator in it. The reason is that when you look at any given variable by itself, like clicks or average position, the examples above, they don’t do anything by themselves but when you divide it by some other variable then there usually is some kind of a sweet spot in there. If you’re looking at something like cost per conversion, the higher the cost per conversion goes, the worse the performance is because you would much rather generate a lead for $10 than $20 because it’s cheaper to generate it for $10.
That’s why the denominators are useful, because there’s usually that break point where we can start determining that more than a certain amount is not profitable and anything less than that amount is going to be a good lead for us. Like if let’s say a phone call is worth $25, any keyword that converts for $25 or less we’re going to consider that good because the ninja statistic cost per conversion is at that threshold that we want it to be at. If it went any higher then it wouldn’t be profitable for us any more. Some examples here. It’s mostly going to be based on conversion data and the reason for that is some other people will value some of the squirrel statistics more than our company does in our management, we are very CPA focused so when we’re analyzing performance we’re going to take a very ninja approach to it and ask, “So what?” To pretty much anything. The so what is almost always, if you ever talk to someone who runs a business, they’re always going to be asking where are the leads? Where are the form fill outs? Where are the phone calls? Where are the quote submissions? All that kind of stuff is what they’re going to be asking which is why the conversion data is the primary thing that we look at for determining the answer to the question, “So what?”
Let’s go ahead and move forward and break down the squirrel stats a little more. In my explanation there I made it sound like squirrel stats were kind of stupid and pointless and then ninja stats were where you should be focusing all of your attention. That’s not true. Squirrel stats are not useless but I will say that they are way too often over-valued, especially by people who don’t know a lot about AdWords. Someone who doesn’t know a lot about an AdWords account will come into an account or they’ll see a report from any given agency that has a lot of squirrel stats in it like … Some of the answers you hear: how much you spent, how many clicks you got, what the click-through rate was, et cetera, et cetera.
The thing is, let’s take click-through rate for example, if you don’t know a lot about AdWords and you look at a report and it says your click-through rate was 2%, that variable might seem high or low to you. Let’s say you think it’s low that only 2% of the time that your ad shows up that it gets clicked on, that’s 1 in 50 times that it gets clicked on, you might think that that’s bad. The thing is, you could have a keyword that has a click-through rate of 2% but then it has a cost per conversion of like $10 and then you have another keyword that has a click-through rate of 10%, much higher, 5 times higher, but it hasn’t converted at all. You look at an example like that, where you have a drastic difference in the click-through rate, but the conversion data is really what’s telling us is more important. Again, it doesn’t mean that these variables aren’t relevant. What it means is that we need to be analyzing them from a standpoint where we understand that AdWords is a beast with all of these moving variables orbiting around the account.
Clicks, costs, conversions, click-through rate, cost per click, average position, all these things are going to be variables within this whole equation but we need to make sure that we properly value each one of them. That you’re not saying, “Oh, that this data point is useless.” But at the same time you’re not shining all the light on that data point to the point where you’re saying that, “My click-through rate needs to be 5% otherwise this campaign is unsuccessful.” A statement like that, it’s just not true. Click-through is not something that you should be using as a key performance indicator but click-through rate is still something that we want to keep in mind when analyzing our accounts performance.
Let’s go through each one of the examples here. Cost. The reason cost is so important is because you’re always going to have a budget. Even if that budget is unlimited, there’s still going to be a maximum amount of money that you can spend. Some people will say that they want to spend $3,000 a month on advertisement. Some people will say they want to spend 10,000, 20,000, 1,000. The number could be anywhere. Cost is something that you always want to be keeping an eye on. Again, cost isn’t … If you spend $10,000 that doesn’t mean that that account is better than an account that spent $3,000 or vice versa but if there is a target, let’s say $4,000 is the target, you need to make sure that you’re keeping an eye on your cost so that you hit that target. That’s why it’s relevant.
Clicks and, I have in parenthesis here, also impression because impressions are similar to clicks. An impression is how many times your ad shows and a click is how many times it actually got clicked on. When I look at clicks, what I’m looking for is statistical significance. For example, if I’m A/B split testing 2 ads and I see that the click-through rate on one of those ads is 2% and the click-through rate on another one of the ads is 10%, if you’re just looking at that variable, you’re just looking at click-through rate there, you’re going to obviously go with the one that says 10%. The problem is that if you don’t have statistically significant data, meaning that you don’t have enough examples of … You don’t have enough people who have clicked through on your ad, let’s say each of those ads only has 20 clicks on them, that’s not enough to determine whether or not that 2% and 10% are going to stay consistent over the course of 6 months or a year. That’s why we look at the number of clicks. The general rule of thumb, and you’ll see this in some of the other blogs that I have about A/B split testing, general rule of thumb is you want to see at least 30 clicks when determining statistical significance.
When you’re looking at the number of clicks, just make sure you keep in mind that that just gives you an idea of how much data you have. It kind of tells you how seriously you can take that data. If there’s 20,000 clicks on an ad, which is a lot, probably should have been rotated by then if you have that many clicks on the ad, if you have 20,000 clicks on the ad, you can be 100% confident that that ad has statistically significant data. However, if the opposite is true, where you only have 10 clicks on the ad or 5 clicks on the ad, you can’t even, no matter what the other data says, you can’t do anything. You can’t determine whether that ad is performing well or not performing well because you don’t have enough data in that account to tell you that. That’s why clicks are useful. Again, this is why they’re called squirrel stats is because when you look at them by themselves they don’t tell you anything but when you apply them in the appropriate places or when you compare them to other campaigns or other ads and how those squirrel stats are measuring up to each other, so how they’re relevant to each other, then these squirrel stats can actually be very useful. Let’s move on.
The next one I have here is click-through rate. The thing that I use click-through rate the most for is for boosting quality scores in a campaign and I’ll tell you how I do that. When we A/B split test ads, the number 1 thing that we are looking for is click-through rate. Second to that, it’s kind of …. I wouldn’t say second to that but just as high up would be conversion data. Again, when we talk about the ninja stats being conversion-based, that shows you how important the click-through rate is with A/B split testing because it’s just as important as the conversion data itself. The reason that we value click-through rates so high when we’re A/B split testing ads … Let’s say you’re running one ad with a certain ad copy and a second ad with another ad copy and you run those ads for 90 days and you find out one has a click-through rate of 2% and one has a click-through ad of 10%, there’s hundreds of clicks in the campaign, it’s definitely statistically significant data, why do we value click-through rates so high? The reason is because click-through rate is how Google makes money and because of that it’s Google’s most dominant variable within their algorithm for determining the quality scores of your keywords and your ads.
What Google will do is if you are keeping the ads that have higher click-through rates than other ones then Google is going to reward you with higher quality scores. By favoring the ad that has the higher click-through rate, you’re actually going to be saving money in the long run because if you’re paying for an ad that has a quality score of 9 versus paying for an ad that has a quality score of 4, you’re going to be paying significantly less for the one that has higher quality score. Your ad rank is going to be much higher, you’ll show more, your impression share’s going to be higher. Quality scores are one of those things where when it’s higher then everything is better.
That’s a little insight into click-through rate and why we value that but, again, click-through rate by itself doesn’t tell you anything but when we compare the click-through rate from ad to ad it actually tells us a lot. Again, this is why we call them squirrel stats, because by themselves they don’t mean anything but when we look at them from a different perspective or we compare them to something, then they actually become very relevant pieces of data for us to analyze. We have to make sure that we’re analyzing them properly and not just vomiting up a bunch of data like, “You’re click-through rate was 8%,” or, “You had 29 clicks.” No one cares about that statement when it’s just by itself, you have to take those statements and use them in a fashion that’s actually applicable to the account that you’re working with.
All right, next one here. Cost per click. The reason that I look at cost per click is generally budget related. If someone says that, “Oh, I only have $3,000 a month to work with,” which really isn’t that much money, it seems like a lot of money but from a PBC standpoint it’s really not that much. Let’s say we have an account, they’re spending $3,000 a month, but let’s say they’re capable of spending $30,000 a month so if I wanted to or if they had the budget for it we could actually increase their budget by 10 times what they’re actually spending. That’s a lot. What happens is we can’t just tell the client or tell whoever we’re working with that, “Oh, you should just be spending 10 times more,” because they probably don’t have the money to be spending that much. If you have a situation like this, this is where cost per click would come in because if you have … Let’s say you have a set of keywords and the cost per click on those keywords range anywhere from $2 to $20, if you’re budget is limited, you’re generally going to want to be bidding on the keywords that are $2 and not $20.
In your head you might be thinking, “Well I always want to bid on the keyword that’s $2 versus one that’s $20.” That’s not true. This is why we call cost per click a squirrel stat, because by itself it doesn’t actually tell you anything. A click that costs $2 is not necessarily better than a click that costs $20 because if that click that cost $2 never converts but the $20 click has a 15% rate or it does convert, then we’re going to be choosing the $20 keyword because that one actually generates leads for the business that you’re working with. That’s why we look at cost per click is because when we’re limited by budget or when budget is a consideration on whether or not these keywords are affordable, we may have to go with the lower cost keywords because the only way we’re going to be generating statistically significant data within the limited budget that we have is by going with the keywords that have lower cost per clicks.
All right. Last one I have here that I look at every once in a while is average position. Average position is a little tricky because your ad can have an average position of like 3, let’s say, meaning you’re showing up in the third ad slot, and even though that’s not number 1 you could still have a 100% impression share, you just have 100% impression share in the third ad slot. Meaning that you’re always hitting that third ad slot. Average position is one of those things that we kind of have to test. If we see that a keyword is performing well in the third position, that my not necessarily be true in the first position. If we increase bids on something … Let’s say it’s doing well in the third position, let’s bump it up to the first position. The problem is the first position’s going to cost you more. We have to keep that in mind.
That’s why when we’re looking at average position, it’s one of those things where we want to compare it to other variable within the campaign. If a keyword is under performing and it’s in the number 1 position, maybe it’s because we’re paying too much. Being in that number 1 position is expensive and we’re having to pay to beat out all our competitors so let’s try bidding less, going to a lower ad position, like the second or third ad position, and see if that’s actually profitable for us. That’s why average position is good is because it gives you an idea of where on the [search 00:17:16] page that you’re ads are landing.
All right, let’s move on. Ninja stats. Ninja stats are the primary variables that you should make bid adjustments on. That’s a true statement. Conversion data is definitely … We’re going to actually go over something besides just conversion data. Ninja stats are definitely what we want to be using as our KPIs. If one keyword has a cost per conversion of $10 and one keyword has a cost per conversion of $20, 100% of the time, I don’t care about any other variables in the account, 100% of the time I will choose the one that has a cost per conversion of $10. That’s the difference between a ninja stat and a squirrel stat. A ninja stat I can look at that variable and I can tell you, “So what?” just based off of that variable’s data. The squirrel stat is the opposite of that, where if we just look at that variable it’s still like, “Okay, you still have 30 clicks, that doesn’t necessarily tell me anything.” The examples I’m using here are conversion data, which we’ve already gone over quite a bit here so I don’t need to keep harping on why conversion are good, but the next one I want to go over here and the one that probably flies under the radar too much is impression share.
If you go into your AdWords account, there’s actually 3 columns for impression share. One of them is going to be your total impression share, the other one is going to be your impression share lost due to rank, and the third one is going to be your impression share lost due to budget. Impression share is shown in the form of a percentage and the way that they derive that percentage is they divide the amount of time that your ad showed on Google by the amount of times that your ad could have shown on Google. Let’s say your bidding on the keyword “red shoes.” Let’s say there is 1,000 search queries for red shoes in any given day and your ad shows up 750 of those 1,000 times. That means that your impression share is going to be 75%, means that your ad showed up 75% of the time that someone triggered that keyword “red shoes.” That extra 25% is going to be lost to either rank or budget.
Again, impression share … Remember when I said that ninja stats generally have a denominator? Impression share is a percentage with a denominator. This gives you an idea of how, by itself, telling you how many times you showed up on Google might not be relevant but when we divide that by the total amount of times that you could have shown, this variable actually becomes much more relevant. That’s kind of a tangent there but shows you why that denominator’s important. Losing impression share to rank, what that means is that you’re either getting outbid by competitors or your competitors have higher quality scores than you or both. What that means is if you’re losing impression share to rank, it means that you need to boost your quality scores, that could be one thing, and there’s a lot of ways to do that. Often times it has to do with your landing page, where the person lands on the site. If they land on a really crappy landing page, it doesn’t matter if you set up your AdWords campaigns perfectly, you’re still going to get low quality scores. Keep that in mind. You can also lose quality score to not having ad extensions, little things within the AdWord account. As long as you’re doing all the best practices you don’t have to worry about that too much.
The other one can be bid. If you’re getting outbid by competitors and you’re worried about that or you want to be … Let’s say you have 75% impression share but you want to push it up to 90% or you have the budget to go up to 100%, then you would want to be bidding more on those keywords to mitigate losing your impression share to rank. The last one here would be your impression share lost to budget. Impression share lost due to budget is when you … Let’s say you have a max budget of you’re spending $100 a day on Adwords, if you hit that $100 mark by 5PM, your campaigns are going to shut off. Once those shut off, any other keyword that gets triggered within your campaign past the time that they got shut off will be considered loss due to budget because you didn’t have the budget to be bidding on those keywords.
The reason that these variables are relevant. Impression share lost due to budget is the most important one. If your impression share lost due to budget is anywhere above about 10 to 15% then you’re bidding too much on your keywords. The reason is because you want to be spreading your budget out throughout the whole entire day so you’re not turning your campaigns on at 12AM in the morning and running them until 9AM and then you max out your budget and then your campaigns are shut off for the rest of the day. We want to stretch that out longer. By bidding less, we can make it so that our … By bidding less we’re also going to be losing more impression share due to rank but you want to be making sure that keeping your impression share due to budget … I always say the sweet spot on impression share lost due to budget is anywhere between 5 to 15%. It’s going to tell you how well your stretching out the budget that you have on any given day. If you look back at the past 7 days worth of data and you see that your impression share lost due to data is like 45%, you know that you’re bidding too much because you could be bidding less, stretching that out for longer, be getting cheaper clicks, be getting more clicks. Overall, you would just be getting more bang for your buck out of that.
That’s kind of an overview of both of them. The truth is you need to be good at implementing both squirrel and ninja data variables. It’s not one or the other. It’s not black and white. You just have to know with the squirrel stats that they’re only applicable when you compare them to something or when you put them up against something else whereas ninja stats are a little easier where it’s like, “Okay, this is how it’s doing. End of story.” Hopefully that gives you a little more insight into some of these variables. For some of the squirrel variables that I went over today, try not to overvalue or look into those too much but rather stay focused on things like conversion. Stay focused on things like impression share. I think you’ll have more success within your AdWords campaigns by doing so. Thanks for watching and I’ll see you guys next week.