I had a number of requests to share the slides from Brighton SEO talk. The slides do not make much sense by themselves, so I decided to make this quick video so that you get the slides but you also get my commentary. It is going to allow me to do some interesting stuff as well as we go through. I have added a few extra little bits and pieces as well, because I did not have a lot of time during Brighton SEO - so this is the full, unabridged version.
This is basically about testing Facebook advertising, and it is from me spending a lot of time, a lot of money to solve a problem with Facebook advertising for a start-up. So the start-up environment provides a great opportunity because there is a lot of money and there is a solution that must be found, which is a bit different to how things work agency-side when you get a little bit of money and if you do not find a solution you kind of move on to something else. Everything you are going to see is from spending a lot of money and a lot of time figuring the stuff out the hard way.
So the key to using Facebook, well the key to using social for pretty much anything, is to get away from the one-to-one and to get into one-to-many. So Google pay-per-click ads are a great example of one-to-one advertising. You pay for ad to be there, you pay for each click. So you pay a dollar, you get one click, the person buys or they don't buy, but you never get more than the one person. Facebook is different. Facebook, where people went on Facebook, is where they can manage to get one-to-many to work, where you can spend one dollar but actually get many people to click because the content or the story or however you're saying it is what people want to share but you also, technically, put it in the way that people will actually share it.
So here's an example of that, really. On the left you've got SMX giving me just a standard advert, work in marketing, yadda, yadda, yadda. I click that ad, maybe I sign up, maybe I don't. It never goes beyond one-to-one. On the right you've got Coca-Cola who have spent some time making some content, and because they've made content that content can be shared. You can see it's got 171 thumbs up, it's got 31 comments, and it's got 13 reshares, where people have taken this and shared it onto their own Facebook pages. So for the money that Coca-Cola have put in, they now have a whole lot of people who have seen this.
This is probably the thing I ran into first, and I ran into this in the first start-up I worked at, where we just threw some money at Facebook, it didn't work, and we went, 'It doesn't work.' That's because we were in the false Facebook learning curve. So the false Facebook learning curve is where you learn the mechanics of how to put an ad live on Facebook. And you get there and you think, 'Wow! I can put an ad live. This is quite simple. It's quite easy. I must know everything there is to know.' You put some ads live. Maybe you give it a couple of months. You spend a couple of grand. You go, 'Wow, no, it's not really for us,' in most cases because you're stuck in that kind of one-to-one because that's what you know already. It's what you know from advertising on Bing. It's what you know from advertising on Google. This one-to-one relationship. And Facebook has many ways you can advertise on the one-to-one relationship but it also has many ways which allow you to open the door to the one-to-many relationship.
So here's the real learning curve. This is where I got into actually learning what was going on with Facebook. It's when I had a method of testing the messages to specific Facebook segments, and then everything took off. So being competent with putting an ad live means absolutely nothing. Having this method, having this commitment, testing over time, and having the tools to do it, that's when you really start to learn.
Difference between Facebook and Google adverts
There are a number of fundamental differences between Google ads and Facebook ads. I'm just going to talk over probably the biggest one which is Google visitors. I think of them as being in a stream. They're like fish in a stream. On day one, somebody's looking to buy shoes. It's pretty much of the same kind of conversion intent as somebody who comes in day 100 and is looking to buy shoes. So the variable that you have there is your ad. Over time you can improve your ad because the people who are coming are generally the same kind of people. Compare that to something like Facebook where, what I say is, visitors are like fish in ponds. Each pond is a different segment. So it could be mothers between 30 to 45 who live in London. It could be boys between 18 and 24 who live in India and like football; it could be all these different types of things. And each time you make an ad you have to be very aware of which segment you're showing that ad to.
So on day one if you're selling shoes and you're Adidas, maybe you're selling to people who've liked Adidas on Facebook. But by day 100 you've used all those Adidas people and you're now onto maybe people who like hip- hop. And then those segments have very different purchase intents. They have very different views on the brand. And it's very much more difficult to take that ad that you've got on day 100 and compare it to the ad you did on day one because you've got such very different segments, very different groups of people that you're showing those ads to.
So that really boils down to the fact that you can compare Google ads over time. Generally, as you test your advertising you'll click-through rate or your conversion rate or whatever the number is that you care about is going to get better over time as your ad gets better because your only variable, really, is the ad. That is completely the opposite on Facebook. It becomes very, very difficult to compare Facebook ads over time.
And what you see here is an example of if we say the blue line on the left represents, if I'm selling a new Adidas trainers, people who've liked Adidas on Facebook. So it gives me a very high click-through rate and a very high conversion rate on day one, and that slowly tapers off over time. And then I have to go and look for a new segment, and that might be anybody who likes any footwear brand on Facebook. And that gives me the red line, which is like they're pretty likely to convert, and then that tapers off. And then, well, the green line is a segment of anybody who likes fashion or clothing brands on Facebook.' See, each time as we go along we get a completely different segment and we get a segment that's generally less likely to convert even though our ads are actually getting better over time. So when we get to the light blue curve on the right, we have best ad but if I compared its conversion rate to the conversion rate I started with it may actually have a much worse conversion rate. And that is because the segments, the people you're showing the ads to, make such a great impact on the click-through rate or the conversion rate over time. So this is why it's tricky but also why you have to be very careful about measuring the improvements in your ads and your ad text over time.
How to create Facebook adverts
The mechanics of Facebook ads, basically any idiot can do them. And really just go to Facebookstudio.com if you want to figure out how to put ads on Facebook. I'm not really going to spend too much time talking about that at all except just to talk a little bit about the third party tools. These tools are cool when you get beyond the most basic level of Facebook. The software that Facebook provides is really very basic. They have a very difficult job because there's a lot of moving parts to doing Facebook advertising and nobody's really solved the user interface very well. These tools, these third-party tools hook into Facebook via the APIs, and they do a couple of things.
One, they try and make the UI a bit better so you can figure out what's going on when you're placing Facebook ads, but they also try and add a layer of analysis so you can figure out this ad is performing better than that ad. If I just flick over here, this is Qwaya. The reason I like Qwaya is this $79 a month. It is by far the cheapest option I've found. I tried it about nine months ago. It was quite rudimentary so it's on my list to try again, they've got a 30-day free trial. If you get to that point where you've got 50 campaigns running and you think this is a pain in the backside, give Qwaya a try as the first step up. You've got Social Moov. These guys are based out in France. They are definitely the next level up as is Glow. So both Social Moov and Glow will take five percent of your ad spent. Glow is based in London, so if you're in London they're probably good to go with because you can speak to them. And they're focused. They're looking definitely at ROI, so if you have a massive account and you want to improve your ROI and you're based in London, Glow's pretty good. The other one I mentioned was Social.com, which is part of Sales Force. I haven't used it, but if you're using Sales Force already have a look at social.com. Maybe it's a very easy thing for you to plug in.
Facebook adverts tools
So tools aside, the next important thing to look at is the message. I'm just going to run through how I approach building ads. If we take this ad here from Qwaya, I obviously found on Facebook. There are three parts that I care about. I care about the image, I care about the title, and I care about the text. So the image is to draw my eye, which they do quite well with that solid color. The title is, I use it as a yes or a no, like an on and off switch: Is this relevant to me, yes or no? So in my case Facebook ads tool, when I read that, I thought, 'Yes, that's interesting to me.' And then I used the text to encourage the click, really. So the text is doing the, the word I've momentarily forgotten, is encouraging me to click through.
Best Facebook advert images
So I'm going to show you a few ways that I come up with ideas. So one of the things that happens is that as soon as you actually have a method for testing things, you then need to find things to test. If you got a first conversion to the Facebook advertising tips page, you will see I have broken a lot of this stuff down.
This basically just looks at images. And I, over a couple of months was just screen shotting every Facebook ad that stood out. I just screen shot it and dumped it into a PowerPoint document. And then after I had, I think in this case, 55 different ads, I sat down, I started to group them into why these things were stand-out ads. If you look at Facebook, it's just a mass of information, right? An ad that's going to stand out in here, there's got to be a good reason for it. And I decided if I analyze these reasons maybe that can help me with my advertising. You've got a lot of stuff on the visual side around lines. You see this big green line. This Moz ad, this big arrow pointing at the text. People do a lot of stuff around color. People do a lot of stuff around, like making sure that they use this white space to make sure their message is clearer to people. I love this one from The Guardian where they've not only used the white space in the image, they've used the white space in the ad by just having these two words. Perspective. And it goes on. There are a number of different ways people do really interesting things. The way Facebook restricts marketers makes some marketers very, very creative; all this stuff around distortion. People have played a lot with putting text on images. And it goes on and on and on. Have a look through that page. Give yourself some creative inspiration if you like.
And I suggest you start to do the same thing with titles and the same thing with these texts. Make more sales. That's quite interesting. Is that a good way to end it off? Free solar insulation. You know, capitalizing stuff at the start. Making the 'free' bit impactful without then having a whole lot of other text, is that something you can test out? It's very important to collect all these things just to give you a base of things to go and test from.
This is quite recent. I added this one at the last minute and this is going to allow me to go into it in slightly more depth. Using dark posts. What a dark post does is it allows you to create a Facebook post without putting it live so that on your wall nobody will see the post but you can still then do Facebook ads to this hidden post. If you're a company, let's say again you're Adidas and you're selling shoes, you can create five hidden posts. Each post can target a different messaging. Your new shoes are coming out. One messaging is around the price. Another messaging is around the technology in the shoes. Another messaging may be in how quickly you can deliver the shoes. You could send maybe 500 people to each post and see which of those messages resonated the most with people. I could say okay, well, the one about the technology had the most shares and had the most likes and had the most comments, so then let's put that one live on the home page. You can use these dark posts to figure out what's the most effective messaging to put on your home page so you don't waste people's attention with substandard messaging. There's also a lot of very interesting and underhanded things you can do with this, so just let your mind wander.
The method; basically, what are we doing when we are testing ads and what can we learn from it? Just to hammer this point home again, segmentation makes a massive impact. Here's an example from a medical trial in the U.S. I was able to target people by people who had liked pages about the illness. It happened to be an illness that targeted black and Latin women, so I could then kind of target by association as well. So what you see here is this targeted campaign is where people had liked the actual Facebook pages of businesses or charities that helped people who had that illness.
Facebook adverts click through rates CTR
Basically you can see the difference in the click-through rates. Where I had the best targeting, I had the highest click-through rate, which gave me the best cost-per-lead. You see the better conversion rate. Where I kind of had to target by association, anybody who liked a black celebrity, black female celebrity on Facebook or Spanish female celebrity on Facebook, I was kind of targeting by that association, and although it did work, it didn't have quite as good of a click-through rate and it had a much higher cost- per-lead which is what I really cared about. Where you can get this highly- specific targeting and matched up with a really great message, you're going to get the best. I guess that's pretty obvious. But if you know the important numbers, if my cost-per-lead that makes my business work is let's say 75 pounds a lead, then that makes this associations also worth my while. This is kind of the meats and bones of it.
When we're looking at images, before we were using any testing, we were just throwing stuff up, going on gut feeling, saying we think this will work or we think this will work or we read something that some guy did this thing and it worked for him so let's do that as well, but when we started to test we were quite amazed by the big difference things made. Here is one of the things, this is one of the first tests I did, throwing some images in. We can see that immediately that the hands just don't work. So when I get to the next iteration of ads, I'm going to remove the hands and I'm going to do things around people's faces.
The same thing with the titles. Now, if I wasn't testing I could have done with either of those titles. They're kind of the same. You have no way of knowing. But once we've tested we can see there's a massive difference in the cost-per-lead, the cost-per-lead being the lighter bucks. And it's significantly lower where we've managed to ask a question. Right? So we've gone from this position where they're all about the same to going, oh my goodness, this one is significantly better than the other one. So the next time I come to creating the next batch of ads, I'm going to create things around questions. Or I'm definitely not going to use the worst performer and I'm going to pick something really different to test against.
And we found the same thing here with the text that we used.
You can see because we've used multiple variations of the title, the image and the text, I now have a really great idea for the next iteration of my ads what definitely not to include.
Here's another example. This was a test for eating yogurt to help you overcome dust mite allergies. This first eight was our first test: a picture of a dust mite or a picture of somebody eating yogurt with two different titles and two different texts. In the first iteration we found that the bug worked really well, so we decided to test a second bug to see if it was maybe the color or was it the fact that an ugly bug, something ugly is attracting clicks. And in the third iteration, we put in some people who were obviously allergic. What we find is that the click-through rate was best for the bugs. The conversions were best for the bugs, but there were some conversions for this guy with hay fever. The cost-per- leads, this thing we cared about, was really much better when we had these bugs. The key thing is you can test, and I think a lot of people, they know that they can't really go on their gut and they kind of use click-through rates. But we'll see is click-through rates are a really bad way, or a really bad thing to use if you're trying to improve Facebook ads.
In this case we had a body text where we had a clear winner. But in the titles, it didn't really make much difference at all. Here we know next round we're definitely not going to use the worst body text. We're definitely not going to use the bad images. And actually we'll just pick one of these titles and then we'll create another title that's very different and see if by creating, by putting all these very, very different variance into the system we can get big wins all the time.
The results. From the first iteration to the second iteration, we managed to improve the cost-per-click significantly. But the thing I really care about is improving the cost-per-lead. So we've managed to drive down the cost-per-lead. Kind of a by-the-by, if you have this system, what you can do is you can spend a limited amount on your learning rounds and when you hit something, now we've figured out, now we've dialed in, now we know what's going on, then you increase your budget.
This is the same campaign, but now we've gone through all three iterations. You can see we've continued to drive the cost-per-click down and I've continued to drive the cost-per-lead down quite significantly in this case.
And again, as I've gone through each iteration, I've load the budget and I've put most of my budget into the best performing ads. And you can see the conversion rate here: 4 percent. And this was all running, all three of these campaigns were running on the same segment of people. I really managed to up the conversion rate and really drive down the cost-per-lead quite significantly from my starting position. And I think that's one of the real take-aways here, is it's very easy to say, yes, of course we should be doing testing and you do testing, but you can make really massive gains if you do testing correctly. It's not small gains. It's massive gains. By the end of this we had the number of conversions or leads that we needed to but if we'd been going through more and more and more iterations, I'd be able to drive this cost-per-lead further and further and further down.
Here's maybe one of these non-obvious things. Sometimes you think you have something dialed in, that you've got this winner and you're improving this winner like we had with the bed bugs, not the bed bugs, the dust mites. We think dust mites work so let's get more and more ugly dust mites. But sometimes that can take you down a path and you kind of ignore other things.
Here's an example from human history, really. If you went on the right side, you can see those iterations down the right but ultimately they turned into a dead end. Or if you had just taken the right hand path you would have missed all the opportunity that still existed on the left hand path. So the thing is, when you're dialing it in you need to keep throwing in all these wild cards or these really different things in case you can, well with the intention really, of hitting upon something that works significantly better than the path you're going on now.
Here's a great example. After a couple of rounds we thought, man, these ugly bugs, like the uglier we can make something, the better it's going to be. But I had seen that the guy that was sneezing had done okay, so I thought well let's throw in an even more obviously sneezing guy. What we get through here, this click-through rate, the uglier the thing is, the higher the click-through rate. But the thing we really care about is conversions. And it turns out the allergic guy converted better, which gave him a better cost-per-lead, which was the thing I cared about.
Once you get into this, you find that you can really start to manipulate click-through rates but you've got to keep an eye on the number that you really care about. In my case, it was cost-per-lead, and that's the number I had to keep an eye on. If you are improving Facebook ads, or any ads really, just on the basis of click-through rate, you can really come unstuck.
Facebook Adverts Conversion Rate Optimisation
Those are some examples. There are a lot of things that need improvement in the system. The Facebook system is quite new in terms of ads. If you think of the Google system, it's been running for ages now and they've had a lot of people, it's had a lot of emphasis where they make billions and billions and billions. Google has quite a mature, shall we say, pay-per-click offering, where Facebook is really just getting into it. The two things I really need are confidence intervals and even delivery of ads. And these things are quite tied together.
Traditional conversation rate optimization, you get a chance to beat the original ad. You'll see here these are controls, the different pages that were being tested. And you go well, this page works 10 percent better than the standard page but it also gives you a degree of certainty that that thing is correct, and that's really what was missing, or is missing, from the third-party Facebook tools. They can tell me that something is 10 or 20 or 30 percent better than something else, but they don't give me the degree of certainty, and that is a lot to do with the fact that they don't give an even delivery of ads over time. They find it very difficult to work out that degree of certainty. But as a marketer, somebody who wants to know that the decisions that we're making are based on solid numbers, this degree of certainty needs to come into the system. If one of either the third-party tools or Facebook, really, can implement even delivery, then we should be able to confidence intervals pretty easily.
The thing I have found with Facebook ads is whatever Facebook is using to pick winners, it's incredibly aggressive. Even on the Google system you've got a couple of different ways you can set the delivery of your ads. You can just say 'even delivery' and then you make the decision of what's better or worse. Or you could tell Google, you decide, when one ad is significantly up, out performs another ad, just keep showing the good ad. I had a lot of experience over the years. And then using Facebook, they are so aggressive in choosing a winner that quite often, actually I have no idea what the basis is for choosing the winner, because they will not have run the other ads enough to give any type of confidence. So I think that's probably a very fundamental issue that Facebook needs to solve in how they deliver ads for it to be relevant to marketers.
There are also two great opportunities; the dark post that I spoke of earlier. What I've shown here is how you test ads. You've got standard conversation rate optimization software that will test landing pages. Dark posts give you the ability to test that bit in between, test the actual content of your Facebook posts. And I think that's a huge thing for everybody that's marketing on Facebook.
And the other big win, really, is for medium-size clients. There is so much volume going through the Facebook system when you're doing Facebook ads that you can really test your messaging on Facebook and then import that into Google pay-per-click. I have a lot of clients where I can get a lot of really great data out of Facebook just because the ads are shown so many times, where there's just not enough click-through data or enough impression data in Google to know that one ad is significantly better than another. It's much easier for me to figure that out using Facebook, that for my shoes that I'm trying to sell people really like that fact that they get this amazing new technology, and that's a much better seller than the fact that it's really cheap. Of course there are differences, and you would still need to test it but you can use Facebook to give you a really nice bit of insight into maybe where you want to start on Google, as well. I hope that has been helpful. Give me an email, give me a shout on Twitter. Cheers.