Episode 71: The Michigan Method: A Strategy for Scaling Ad Campaigns

scale facebook ad campaign

Are you having problems controlling your Facebook ads and keeping your conversion costs in check as you scale?

Take back control from the Facebook ad platform using the Michigan Method, the “multivariate testing technique on steroids.” Ralph reveals how to use this strategy (and the origins of its name), so you can get the best ad results. This episode goes above and beyond the tactics discussed in Episode 69.



  • How to test multiple Facebook ad methods at once, so you can find the perfect ad as you scale.
  • How you can avoid ad cannibalization and audience oversaturation when using the Michigan Method.
  • When to “starve your ponies and feed your stallions” so you get results fast and don’t waste money.


Episode 33: The Ad Grid: How to Build Campaigns that Convert and Scale
Episode 69: How to Structure a High-Converting Facebook Campaign
Episode 70: 5 Facebook Ad Features We’re Using and Why
Keith’s $1 Trail for Facebook Ads University (for the webinar)
Episode 71 Transcript (swipe the PDF version here):

Keith Krance: Hello, and welcome back to Perpetual Traffic, Episode 71. We have all three of us today. Molly, Ralph, and myself. How are we guys doing?



Molly Pittman: Woo!



Ralph Burns: Good.



Keith Krance: Good, good, good.



Ralph Burns: Real good.



Keith Krance: Yeah, I’m in Austin. Molly’s in Austin. Ralph’s in Boston, up at Cape Cod. Me and Molly are like 25 minutes apart.



Molly Pittman: Miles away.



Keith Krance: All good, all good. We’re stoked to go through another episode with you guys. On today’s episode, we’re going to be doing a follow-up to Episode 69. Episode 69 was “How to Structure a High Converting Facebook Campaign.” Today, Ralph is going to be revealing the Michigan Method. In a minute he’ll tell you where that came from, and why it’s called that.



  Today’s going to focus on questions like: “How do I control what’s happening with Facebook? Sometimes they’re not showing all of my ads. They’re not displaying all of my ads. I’m having a tough time scaling out my campaigns.”



If you’re in that situation. If you’re having problems with controlling how many impressions Facebook’s giving different ads. Or just in general, if you’re raising your budgets and your conversion costs are skyrocketing. You feel stuck. You feel like there’s no way to do it.


  Within the agency, Ralph and the team have been testing a lot of different methods. Not only by working with different clients, but also with talking to a lot of the high-level people we know in the industry as well. I think, Ralph, the best way to get started is by explaining what the heck the Michigan Method is, and where the heck did that come from?



Ralph Burns: Good question. By the way, Ryan Deiss actually told us to keep that as a name. Initially, we thought it wasn’t cool enough. Ryan said, “Yeah, it is cool. Because people are going to go, ‘What the heck is that thing?'”



Molly Pittman: That’s probably why you clicked on this episode to listen.



Ralph Burns: Absolutely. Thanks, Ryan!



  The Michigan Method. Let me tell you a little bit about what it is. Because the first time we created it is where the name came from. I’ll leave that hidden for right now. The goal of the Michigan Method is to control what we refer to as the “uncontrollable.” If you’ve been running Facebook ads for any period of time, or maybe if you’re just starting, Episode 69 is a good place to start on this. This is definitely a next level strategy above and beyond what Episode 69 talks about. That episode talks about testing independent variables to see what works. Today, we’re talking about taking back control from Facebook ad platform, so you can get the best result possible by testing as many things as possible in a scientific way. For the base level, definitely listen to Episode 69.



  Once you start getting ROI, now you’re in the position where you can actually start to scale. Whatever you’re doing, your message is starting to work. It’s starting to resonate with your audiences. That’s where the Michigan Method really comes into play. This is not a starting strategy. I’m going to say that over and over again here, so you don’t get confused. This is a lot of information and it can get very, very confusing. For base level, go back to Episode 69. And work from there.



  What we are always trying to do with our advertising inside the agency is to get what our customers really want: to scale their ads up. They want to scale their Facebook spend, and they want to maintain their ROI, if not expand that ROI. In our terms, we’re looking at Return On Ad Spend (which we call ROAS), which is really ROI based upon how much you’re spending. That’s the main goal for all of our customers who come on board with us.



  Back when we were first starting, and I give a lot of credit to this to guy on our staff, Vladimir Gonzales.



Keith Krance: Vlady!



Ralph Burns: Vlady the Unyielding, as we call him in the agency. He is a creator of the Michigan Method, along with myself, but I give him a lot of credit for coming up with this. He noticed there was a campaign where we spent anywhere between $5,000 to 10,000 dollars a day. We kept spending more. We kept expanding our budgets. We would say, “Let’s start off with a $50/day budget.” Then we would expand it to $60/day and then wait three days. Then we would expand it to $75/day and then wait a few days.



  We noticed whenever we would start growing our budget, which we now refer to as budget scaling, our CPAs (cost per acquisition) started to expand as well. Not exactly the type of thing you want. This is typical with Facebook. Maybe they’ll get this fixed at some point in time, but in the here and now, at the end of 2016, that’s the way it is. We realized that was an issue. We remembered when we first started working together about three years ago when Facebook had campaigns and just ads. There was no ad set. If you remember back to those days of the dinosaur, it seems like way, way back now.



Molly Pittman: The good old days.



Ralph Burns: The good old days.



Keith Krance: Back before they had newsfeed ads.



Ralph Burns: Yeah, absolutely. A lot of changes since then.



We had one customer, and I think this was our first customer that we actually worked on together, Keith. This customer would give us 30 or 50 or 100, if not 200, interests and test each one independently in its own campaign. One campaign, one ad.


Molly Pittman: Wow!



Keith Krance: Sometimes it was 250 interests. This was before lookalike got in.



Ralph Burns: Yes.



Keith Krance: This shows you how much interest research you can do. This is a client knew his audience. He also spent a lot of time on Facebook looking at targeting potentials. He would just build out a Google spreadsheet and give us the interests that he wanted to target. But, we’d split those up into different ads.



Ralph Burns: Yeah. What made it even worse, sorry, but this was a nightmare. Let’s say they gave us a spreadsheet of 250 different ads, the “Like Grid,” I think it was called, Keith, I forget exactly the name. But what it made it even worse is that for each campaign and ad—remember, no ad set back in these days—there was a different affiliate link. Not only did we have to create 250 ads and 250 campaigns, but we had to create 250 individual ads with their own unique URL. A nightmare scenario if you’re doing this on your own.



  That’s when we started to enlist the help of Vlady. He actually came on board specifically for this customer. The beauty of this method was that for each one of these ad sets, the customer only wanted us to spend $5 or $10 a day. That was it. Two-hundred-fifty times $10 a day, let’s say, that’s a fair amount of daily ad spend. Then we had a CPA goal. I think in that case it was $3-$5 for a lead. We would pause the ads as they were going over our CPA. But the point was, way back then, we had a lot of control.



  Fast forward three years. We’re now in this campaign that we’re spending $5k-$10k a day, and we don’t have any control. Vlad, to his credit, said “Hey, remember way back when, when we used to do this stuff? Our first campaign we ever worked together. We would do this 250 campaigns, one ad per ad set. Why don’t we start doing that?”



  I said, “Yeah. And why don’t we test three or four different variables on the ad itself?”



  He said, “Yeah, that’s even better. Then why don’t we test the placement? Not necessarily mobile, desktop, and right-hand column in one campaign, but we separate those out and create individual campaigns.” Now we’re testing placement. Somebody responds to a desktop ad differently than they would a mobile ad, or differently than they would a right-hand column ad. So let’s separate those out into three different campaigns.



  Inside each one of those campaigns, we would then test 30 interests. Thirty interests times three campaigns, now you’re talking a lot of ad sets. A lot of interests being tested. But, we wanted to take it one more level. Let’s say one of those interests in those 30 interests is Tony Robbins. Another interest may be Ryan Deiss. Well, there’s probably some overlap between those two, to a certain degree, but we want to test them independently. We also want to test what message, what image, creative, what ad copy, resonates best with the Tony audience versus the Ryan Deiss audience. We said, “Well, why don’t we create three variations of ad copy and then three variations of the image?”



  In this case, actually, the first one we did was three videos. Doesn’t matter what your creative is. It’s three videos, or three creatives, images. Then three different ad copies with slightly different hooks. Then we would target that exact same audience. Now we’re looking at three campaigns times 30 specific interests in each ad set, times nine ads. Nine variations.



  Then we started to look at it. We’re like, “Oh my god, each campaign is 270 ad sets.” If you can do the math, 30 interests times nine ads. But each ad is targeting one of those 30 interests.



Keith Krance: Each one of those is its own ad set.



Ralph Burns: Yeah, so in each campaign, you now have 270 ad sets.



  It’s nine ad sets. Each one of those nine ad sets has the same interest. Then you go to your next one. You have your Ryan Deiss one. There’s another nine ad sets, nine variations of ad. It’s one ad per ad set. That’s the key. One ad per ad set. We did this for all 270 for desktop. I believe that’s the first one we did. Then we duplicated it inside Power Editor and did the same thing for right-hand column. Then we did the same thing and duplicated it again for mobile. Now you have three different campaigns, one for mobile, one for right-hand column, and one for desktop.



  In each of those three campaigns, you have 270 ad sets. One ad per ad set. Two-hundred-seventy times three equals 810. I said, “810.” I googled “810.” The first thing that came up was the Flint, Michigan area code. That was the start of the Michigan Method.



  Our working title was, “Hey, let’s call this the Michigan Method.”



  Vlad said, “Yeah, great. Perfect.”



  Problem was that 270 ad sets inside Power Editor actually crashes the Power Editor. Poor Vlad…



Molly Pittman: Ralph broke Facebook.



Ralph Burns: He’s got 810 ad sets. The poor guy. He said, “Ralph,” I remember he sent me a message in the morning. “I’ve taken down Power Editor. I think I broke the entire platform last night creating this method.”



We soon realized that even though we were calling it the Michigan Method, we now realized that the maximum you should have in any campaign is 200 ad sets. We typically will do a little bit less. We might do 20 different variations of interests, and then nine different variations of ads, which makes 180. Still, that’s a lot. That’s the whole evolution. The long drawn out story of what the Michigan Method is. We set the campaign live, and right out of the gate, it did extremely well.


Keith Krance: I love it, I love it. We can break down the math in a little bit. For a second, I want to go up to a 30,000-foot level. Let’s talk about the Facebook algorithm in general and why this actually works. A lot of you might remember a few years ago when there was a big craze, a big buzz about getting penny likes. Remember that? You’d run Like campaigns, and the goal was to get new fans for a penny or less. One cent per fan. People would build up these fan pages. Like a Tom Brady fan page or New England Patriots or Seattle Seahawks or University of Michigan or University of Florida. Then they would start to sell t-shirts. They would sell t-shirts to these raving fans in the sports niche, or anything like that. I remember when that was happening. What happened was they figured out the algorithm was the key to all of this. To have super big audiences and small budgets of $5 per ad set, maximum $10. Then you would try to have really big audiences.



  If you think about it, for this to work, that’s still a part of the formula. There are some principles that you have to understand with Facebook. Now, also realize that the algorithm changes. And this will probably change in 90 days or six months, which is why you want to continue listening to this podcast and be part of our members’ area, if possible. So you can have your questions answered.



  The point is, Facebook loves large audiences in general. Especially if you have a lot of data. If you are generating conversions consistently. If you’re using the website conversions algorithm we talk about a lot and that we use in 80 to 90 percent of our campaigns. Now, Facebook isn’t going to put your ad in front of everybody in that audience. They’re going to start segmenting. They’re going to start doing a lot of the targeting for you. But, when you’re first starting out, you don’t want to start out with huge audiences because you want to discover which audiences are really working.



  For those of you who are the fact finders and do everything exactly like we say. You’re digging in. You’re looking for every specific detail. You’ve got to get out of that for a second. I want you to sit there for a second. Close your eyes and just think, “Okay, I’m going to listen to this episode, and I’m going to take the key principles out of this. Then I’m going to take my specific situation and I’m going to do what I can. If I know I only have a $30/day budget or a $50/day budget, then I’m going to adjust this principle.”



  If you understand that Facebook works well when you give it more time and a bigger audience to go out and find the right people. That it can work better in the long run. But there are other times or situations where you’re in the discovery phase and you want to have smaller audiences, so you can see how these different interests do against each other.



  The other thing that it’s interesting is I know that we’re testing different interests, like Tony Robbins against Ryan Deiss. You have a lot of different ad sets, and it’s great because you get to test them against each other. But guys who understand split testing, they’re pulling their hair out right now, like, “That’s not a true split test!” Because there’re 500,000 people like Tony Robbins and 500,000 people that like Ryan Deiss. There’s actually an overlap. There’s probably 100,000 of those people that like Ryan Deiss and Tony Robbins.



  Yes, you can do all the exclusions if you want, if you want to do a true split test. But sometimes we found that it’s easier to manage if you just have them on their own. Tony Robbins on his own. Ryan Deiss on his own. And not do any exclusions. Because, for some reason, it seems to work better. Yes, there’s overlap so that split test that you’re doing is not perfectly accurate. But for some reason, it just works better and it’s easier to manage.



Ralph Burns: Yeah, without question. You hit the nail on the head, but this is flexible.



Keith Krance: Yeah.



Ralph Burns: I don’t think we do the same Michigan Method twice in the agency. We always do it slightly differently. For example, we might only have one video asset. We might only get one video from a customer. We’ll split test maybe 3, 4, 5, 6 different ad copies, which is fine.



Keith Krance: Yup.



Ralph Burns: The point is that we’re not necessarily looking for a winner, per se. Yeah, we are, but we’re also testing what works and what doesn’t. The thing about the Michigan Method, or anything related to Facebook ads, and Molly can certainly attest to this, is that you fail an awful lot. You figure out what doesn’t work. The point is to understand that if you did set up a Michigan Method, and 80 percent of your ad sets failed or they were over your target CPA, that’s okay. Let’s say you wanted to get a $2 cost per acquisition, and the majority of them are $3-$7. Pause the ones that are above your CPA and that’s really simple.



  The Michigan Method allows you to see all the ad sets are on your screen, in front of you. The day after you launched this thing, you can just pause the ones that aren’t getting you the result that you want. We call that, I think this is from Perry Belcher, “starve the ponies and feed the stallions.” Feeding the stallions is the scaling method. I don’t know is if we’ll have too much time to get into that here today, but for early optimization, remember that not everything you put on Facebook is going to work. If you get 30 percent of your ads to work really well, you’re in the hall of fame. It’s like baseball.



Molly Pittman: Ralph, I couldn’t agree more. I think what’s beautiful about the Michigan Method is the structure. It’s setting you up to be able to test a lot of variables against one another. In Episode 69, we talked about a very basic way to test. That’s the method you should use if you’re just starting out. If you have a little bit more experience, the Michigan Method is a wonderful way to set up your campaign so you can test multiple variables at one time. Whether that happens to be different interests or placements, different bidding strategy, whatever you’re interested in, whatever you think will move the needle the most for your campaigns, the Michigan Method is a great way to do so.



  I was actually able to hand over the Michigan Method to an intern this summer. He set up and scaled a campaign that lasted for a few months, just based off of this strategy. It’s a great foundation. It’s a great way to set up campaigns and really test different variables so you can find the perfect ad.



Ralph Burns: If you’re familiar with the term of multivariate testing, this is really multivariate testing on steroids because you’ve got so many different variables. Let’s say it’s 180 ad sets times three different placements. You’re looking at 540. You’re not necessarily looking for, “Okay, well it’s the Tony Robbins interest, and it’s Ad Copy 1, Image 2 that works best.” That’s great. If you want to resolve that, and say that’s what you want to do, but our real methodology behind this is to gain control of the uncontrollable.



  Facebook will tell you, “Hey, throw all your placements in one ad set and split test four or five different ads on the ad set level.” That’s great when you’re first starting. I recommend you do that because it does give you information. It gives you some data, but it’s not a true split test. Because, as you know if you’ve been running Facebook ads, if you do have 3-5 ads in an individual ad set, Facebook picks one or the other, unless you have a massive budget for that ad set. Plus, now Facebook has the split testing feature, which we talked in the last episode.



  Getting back to basics here, this is a way to scale up and to get ROI. It’s the best method we’ve seen to be able to do that. But like Molly says, it is really flexible. Look at your data and figure out what you want to do. Maybe you only do ten interests or five interests. Then you put all your different placements in one campaign. I have somebody on staff and that that’s the way that she prefers to do it, which is fine. She gets good results that way. As long as she doesn’t mix her image ads with her video ads. That’s the one thing we want to make sure that we keep out of that.



  The point is that it’s flexible. You don’t have to spend $9,000 a day to use the Michigan Method. You can do it for 100 bucks or so. I’d really encourage you to make sure you get the basics down first. Make sure that your message, your hook, and your avatar are dialed in first by listening to Episode 69. Then you can move on to this one after that.



Keith Krance: Ralph, just to recap on the math, and then we’ll come in to some other specific questions that I think people might be wondering about—like audience size and budget.



  With the initial case that you talked about, just break down the math again. Don’t worry about the other two ad campaigns. Then let’s bring it down. Let’s say someone wants to either do a quick campaign or they don’t have a huge budget. What would be the best way to start, and then what are some suggestions for budget and audience size?



Ralph Burns: Yup, absolutely. If you’re confused by the first part, go back and listen to it again, if you can. I’ll try and break it down a little bit easier here. Because the way that we build these now is a little bit different than the way that we first started building them. Folks in the agency will actually create their first nine ads with one interest in one placement. They’ll create one campaign. Let’s say it’s desktop, for example. They’ll label it “desktop.” That will the first campaign. Then a mobile campaign. And also a right-hand column campaign. As of right now, we don’t use Instagram on this. We don’t use audience network. Those are for other types of opportunities.



  So, starting basic. You create a campaign and create one ad set. In that ad set, you choose one interest. Let’s say it’s Tony Robbins. Then you create your first variation on the ad level. So now you have one campaign, one ad set, and one ad. Let’s say you wanted to test three different ad copies: maybe one with a question mark, maybe one with a statement, maybe one with an ellipsis at the end. A shout-out to Russ Henneberry on that one. Big fan of the ellipsis.



Molly Pittman: And Johnson Boxes.



Ralph Burns: That’s right and Johnson Boxes.



You create that first ad. So it’s Tony Robbins. It’s Ad Copy 1, Image 1. All right, so you write that. Then all you do is you duplicate that ad set eight more times. Same interest. Same Tony Robbins interest. Now you have nine ad sets, nine ads. On your next ad go in and relabel it. Use the same ad copy. Ad Copy 1. Then do a different image. Image 2. Then on your third ad set ad, you go into Ad Copy 1, Image 3. Now you’re varying three images with the same ad copy.


  Then you do the same thing for next three. You change up your copy. Ad Copy 2 now for your fourth ad set and ad. Then maybe Image 1. So you’re using the same image that you used before. Now you go down one more. You create your fifth ad set ad, which is Ad Copy 2, Image 2. The next one is Ad Copy 3, Image 2. Then you’ll do the same thing for the rest of them. The rest of the three, which would be Ad Copy 3, Image 1; Ad Copy 3, Image 2; Ad Copy 3, Image 3. That’s if you’re using nine variations, which is typically what we’ll try to do. In your case, you might want to do four variations, two ad copies, two images. It’d be Ad Copy 1, Image 1; Ad Copy 1, Image 2; Ad Copy 2, Image 1, Ad Copy 2, Image 2. That’s it. You’ve got one ad per ad set, and that’s really the way that we structure it.



  Now, we found it’s vital to you use smaller budgets with the Michigan Method. We typically will go anywhere between $5-$20, $5-$30 a day in each individual ad set. You could go below $5, if you wanted to, but we haven’t tested that a whole lot. The key to this is the audience size. Your audience size has to be at least a million, which is your potential reach. That’s the little thermometer that you see inside your ad set where it says potential reach is 2.7 million, or potential reach is 300,000. You want to make sure that your potential reach is at least a million plus. In many cases, we’ll go 37 million in that. Maybe 70 million. Maybe we’ll leave it wide open, which is not necessarily a best practice, but the point is this method really only works if the audience size is large enough. If the potential reach is large enough. That way, we don’t see any cannibalization between all the nine variations.



Molly Pittman: That way you’re also not saturating your audience too quickly.



Ralph Burns: Correct, absolutely. With a $5 a day budget, you’re not getting a million-person reach. You’re getting maybe a couple hundred people a day, or maybe a thousand people a day. That kind of thing.



Keith Krance: Remember, we’ve talked about this before on previous episodes. If you have a question: “Hey, should I try this audience size? Is this audience okay?” A lot of people ask questions like that. If that’s your question, then the answer’s going to be try it. Create another ad set and test it out, and see how it does.



Ralph Burns: Super simple. Twenty-four hours after you launch this sucker, you want to go in and you want to starve those ponies. Just kill the bad ad sets. Kill your bad ads. Don’t get attached to them. These are not your children. You want to pause them. Even though you really loved Ad Copy 3 and Image 2, maybe that ad is just sucking.



Molly Pittman: The variations that win, in my opinion, are always the ones that I’m least excited about.



Ralph Burns: So true.



Molly Pittman: If you refer back to Episode 33 about the Ad Grid, it’s a system we use for creating ad images, copy, and figuring out your avatar. All of this works really well with the Michigan Method.



Ralph Burns: Yeah.



Molly Pittman: They really go together. It’s cool. We’ll come up with the ideas for different creatives and different hooks and angles we want to take with the ad copy. Then when we test. Like I said, it’s usually the ads that I call the boring ads, they’re the ones that usually win. And I’m like, “Why?!”



Ralph Burns: Yeah, I know. “I love that one so much! It’s sucking so bad.”



Molly Pittman: Right! “Why didn’t you like my creative ad?!”



  So it’s always worth a test, like Keith said. If you’re going to ask “Should I do this?” Just do it. Even if you’re testing it for $5. It’s always worth a test.



Ralph Burns: Yeah.



Molly Pittman: You never know. And we can give you guidelines and strategies. That’s why we’re here, but we’ll always encourage you to test. The Michigan Method is a great way to set your campaigns up so that your testing as many elements as possible.



Ralph Burns: Yeah, I mean there’s sort of a joke inside the agency. Through our project management system, we all give each other a hard time. Any ad that I either write or help write inside of Michigan Method and I’m really psyched about, the joke is that’s going to be the worst performer. It typically is the case. It’s the one that I think, “Oh, this is just so great. This is my best ad copy ever!” I look at it, and it has a $17 CPA for a customer who wants a $5 CPA. Obviously, that is where we pause. We pause them right out of the gate. As soon as they’re two to three times above what our target CPA is. So we don’t lose the customer any money.



  Molly brings up a great point about the Ad Grid in Episode 33. Definitely go back to that one. Because, in many cases, we’ll create Michigan Methods, not to get even more complicated, we’re talking three campaigns, 20 interests, and nine variations of ads. In many cases, we’ll create individual Michigan Methods for three separate avatars. We’ll then create nine different campaigns, all with its own independent copy. Talk about testing on steroids! That’s crazy. But we get results really fast because you just never really know what combination of ad copy and image is going to work for which particular avatar. We have lots of customers who have very specific avatars that are very diverse. Testing through the Michigan Method independently is a great way to get the best result that we want.



Molly Pittman: That’s amazing.



Keith Krance: Love it, love it. Great stuff, great stuff.



Just so you know, we recently did a webinar inside Facebook Ads University, our continuing education members area. We went through this same strategy on the screen. So you could see stuff a little more visual. To access the webinar, you’ll need to sign up for Facebook Ads University. We’ll put a link in the show notes where you can go do a trial. It’s a 10-day trial. You can also go to dominatewebmedia.com/try. It’s a dollar trial. That will give you a dollar trial into Facebook Ads University, which is, like I said, an ongoing education where we’re doing live webinars.


  We also have our Fast Start Training inside there. It’s not the full blueprint course, but it’s a fast start, short seven-module training. Then you can go in, and I believe it was in September, in the recorded webinar section, we did a webinar on this exact thing. Ralph and I did. So if you want to go check that out, I highly encourage it.



  Things are changing fast. Same reason you should be, I believe, a member of DigitalMarketer Lab. That’s where you’ll finding up-to-date things, what’s working right now.



  All right. Good stuff, dude, good stuff. Once again, head to the show notes for the resources.



Molly Pittman: Yeah.



Keith Krance: And we can’t wait to talk to you next week.



Ralph Burns: See you.



Molly Pittman: Thanks, guys.



Thanks so much for joining us this week. Want to subscribe to Perpetual Traffic? Have some feedback you’d like to share? Connect with us on iTunes!

iTunes not your thing? Find us on Stitcher or at DigitalMarketer.com/podcast.