📚 What You Will Learn Today:
Have you ever run two Facebook ads that were exactly the same, and one did great while the other flopped? Same picture, same words, same audience, totally different results. It feels like gambling. The Michigan Method is a simple way to stop gambling and start testing. You split your ads into small, clean groups so you can finally see what's working.
By the end of this guide, you will know:
- Why identical ads give different results (it's not bad luck)
- The real reason your cost per lead keeps creeping up
- The 4 steps of the Michigan Method, in plain words
- How to set it up in your ad account this week
- When to stop testing and put your money on the winner
🤔 First, Why Does This Matter?
Here's the problem most business owners run into. Facebook's system changes all the time, and the cost to get a new customer keeps going up.
Worse, the system feels random. You make two ad sets. Same audience, same placements, same picture, same words. One brings in leads for $8 each. The other brings in leads for $31 each. Sound familiar? You're not alone. Almost everyone who runs ads has lived this.
It's Not Facebook. It's Your Setup.
It's easy to blame Facebook. But before you do, look at your own ad structure. Ask yourself: am I running too many things at once?
Most people are. They put 5 pictures, 4 headlines, 3 age ranges, and 6 interests all into one or two ad sets and hit publish. That's like throwing 20 kinds of seeds into one garden bed and then wondering which one grew into the tomato. You can't tell. Everything is mixed together.
A gym owner had one ad set with 8 different ads inside it, targeting everyone aged 25 to 65 who liked fitness, yoga, running, or weight loss. He spent $900 in a month and got 40 leads. Decent. But when his coach asked "which age group is cheapest?" he had no idea. "Which interest brings people who actually show up?" No idea. "Which picture is the winner?" He guessed. He had spent $900 and learned nothing he could use next month.
Facebook Splits Your Audience Without Telling You
Here's the part most people don't know. When you give Facebook a big audience, it does not show your ad evenly to everyone. It picks little pockets of people it thinks will respond and leans hard on them. Marketers call this segmentation, which just means "splitting a big group into smaller groups."
So your "25 to 65, likes fitness" audience might secretly be "mostly 45 to 54 year old women who like yoga" in one ad set, and "mostly 25 to 34 year old men who like running" in the other. No wonder the results are different! Facebook split them for you, and you never saw it happen.
The fix: do the splitting yourself. That's the whole idea behind the Michigan Method.
📊 Mixed Up vs. Split Up
When everything is mixed, you learn nothing. When each ad set tests one thing, the winner jumps out at you.
✅ The Michigan Method: 4 Steps
The method has four parts. You do them in order. Each one removes one kind of "mixing" so your results get clearer.
🎂 Split by Age
Instead of one ad set for everyone 25 to 65, make separate ad sets for each age group:
- 25 to 34
- 35 to 44
- 45 to 54
- 55 to 65+
Everything else in these ad sets stays exactly the same. Same interest, same placement, same ads. The ONLY difference is age.
Why This Matters
Different ages respond to different things, and they cost different amounts to reach. A 28-year-old and a 58-year-old might both want to join your gym, but they click on different pictures, and one group might cost half as much. If they're mixed together, Facebook picks the cheap group for you and hides the expensive one. When they're split, you see both clearly.
A chiropractor split her "back pain relief" ad by age. Result after two weeks: the 45 to 54 group was bringing leads at $12 each. The 25 to 34 group was costing $40 per lead and most of them never booked. She turned off 25 to 34 and moved that money to 45 to 54. Her total leads went up and her cost went down, with the same budget.
📱 Split by Placement
A placement is where your ad shows up. The main news feed. Stories. Reels. The right-hand column on a computer. The Marketplace. Each one is a different place with different people in a different mood.
The Michigan Method says: don't mix placements in one ad set. Test them separately.
- One ad set for Feeds (Facebook and Instagram feed)
- One ad set for Stories and Reels
- One ad set for the right column or other smaller spots, if you want to test them
Why This Matters
Reels placements are often very cheap, so Facebook loves to pour your budget there. But cheap clicks are not always good clicks. A person watching Reels at midnight might tap your ad by accident. A person scrolling the feed at lunch might actually read it and fill out your form. If you mix them, the cheap Reels clicks hide the fact that your feed ads are the ones bringing real customers.
Stories and Reels need tall, vertical videos or pictures. Feed ads work best with square. When you split placements, you can also give each one the right shape of ad. That alone often lowers your costs.
🎯 One Interest Per Ad Set
An interest is a topic Facebook thinks a person cares about, based on what they like and follow. "Yoga." "Home improvement." "Luxury real estate." "Parenting."
Most people stack 5 or 10 interests into one ad set. The Michigan Method says: use one interest per ad set. This way you know exactly which interest is pulling its weight.
How It Looks
- Ad set A: Ages 35 to 44, Feed, Interest: Yoga
- Ad set B: Ages 35 to 44, Feed, Interest: Running
- Ad set C: Ages 35 to 44, Feed, Interest: Healthy eating
A med spa owner always assumed "Beauty" was her best interest because it was the biggest. When she split it out, "Beauty" leads cost $45 and rarely booked. "Skin care" leads cost $19 and booked often. She had been spending most of her budget on the wrong one for a year. One week of clean testing fixed it.
Making the audience too small. One interest plus one age group in a small town might only be 8,000 people. That's okay for testing, but if Facebook warns you the audience is too narrow, widen the location a little or combine two neighboring age groups. You want enough people to get real results.
🖼️ Test Your Ads Inside the Clean Structure
Now that your age, placement, and interest are each clean, you can finally test the ads themselves. Put 2 or 3 different versions inside each ad set:
- Same words, different picture
- Same picture, different headline
- A short video vs. a still photo
Because everything around the ads is the same, any difference in results comes from the ad itself. Finally, you're learning something true.
Keep Ad Examples Honest
If you run ads for a gym or any health business, Facebook has rules. Don't say "lose 20 pounds in 2 weeks." Don't use before-and-after body photos. Say things like "get stronger," "build a habit you'll keep," or "feel better after work." These pass review and, honestly, they work better anyway because they sound like a real person talking.
Change only one thing between ad versions. If you change the picture AND the words, you still won't know which one made the difference. One change, one lesson.
📊 How the Ad Sets Stack Up
Every layer tests one thing. That's what makes the results readable.
🧮 Why This Works (and How Much It Costs)
Here's the math behind the method. If you throw 100 ads into a few mixed ad sets, finding the winner is almost impossible. The Michigan Method gives you a structured, step-by-step way to get clear results. When you control the variables, you can actually learn from your tests.
A "variable" is just something that can change. Age is a variable. Placement is a variable. When you lock all of them except one, the one you left open is the one you're testing. That's how scientists do it, and it works for ads too.
Won't This Cost a Fortune?
Not if you're smart about it. You don't have to test every age, every placement, and every interest all at once. Pick your two best guesses for each and start there.
- 2 age groups × 1 placement × 2 interests = 4 ad sets
- $10 a day each = $40 a day
- Run for 7 days = $280 for a very clear picture
After a week, kill the losers, keep the winners, and add the next thing you want to test.
You're not spending extra money on testing. You were already going to spend that $280. You're just spending it in a way that teaches you something, instead of a way that leaves you guessing.
🏆 When to Stop Testing and Scale
Testing is a phase, not a forever thing. Once you know your winners, you move to the next step: scaling, which just means putting more money behind what works.
Signs You've Found a Winner
- One ad set has had the lowest cost per lead for at least 5 to 7 days in a row
- The leads from that ad set actually show up, book, or buy (not just fill out a form)
- You've spent enough on it to trust the result (at least 10 to 15 leads)
How to Scale
- Raise the winner's budget by about 20% every few days. Big jumps confuse Facebook's system.
- Once you know your best age and best interest, you can combine two or three winning groups into one bigger ad set. Now Facebook's smarts work for you, because you only gave it good ingredients.
- Keep one small test ad set running at all times. Winners wear out. You always want the next one warming up.
Measure What Matters
Cost per lead is nice. Cost per customer is what pays the bills. Track both. And when you count a sale, count it on the day the customer actually paid, not the day they filled out the form. Otherwise your numbers will lie to you.
None of this matters if you're slow to follow up. A $9 lead that waits two days to hear from you is a $9 wasted. Set up automatic texting so every new lead gets a message within 5 minutes. Clean testing finds cheap leads. Fast follow-up turns them into customers.
📝 Quick Summary
The Michigan Method, In One Place
- The problem: Facebook splits your audience for you, so mixed ad sets give random-looking results.
- Step 1, Age: separate ad sets for 25 to 34, 35 to 44, 45 to 54, 55 to 65+.
- Step 2, Placement: feeds in one ad set, Stories and Reels in another. Never mixed.
- Step 3, Interest: one interest per ad set so you know which one is working.
- Step 4, Ads: test 2 or 3 ad versions inside each clean ad set, changing one thing at a time.
- Then scale: kill losers, raise winners slowly, keep one test running.
Remember: You can't learn from a test where everything changes at once. Control the variables and the answers show themselves.
🎓 Your Action Plan
Week 1: Set up the clean test
- Pick your 2 most likely age groups and 2 most likely interests.
- Build 4 ad sets: each one age, one interest, Feed placement only.
- Put the same 2 ads in every ad set. $10 a day each.
- Turn on automatic text follow-up for every new lead (within 5 minutes).
Week 2: Read the results
- Write down cost per lead for each ad set.
- Turn off the two worst. Move their budget to the two best.
- Add a new test: the winning age and interest, but Stories and Reels placement.
Week 3: Test the ads
- In your best ad set, add a third ad that changes only the picture.
- Check which leads actually booked or bought, not just who filled out the form.
Week 4: Scale the winner
- Raise the best ad set's budget by 20%. Do it again in 3 days if results hold.
- Count your sales by the day people paid. Compare to what you spent.
- Pick the next age group or interest to test next month.
The business owners who win with ads aren't lucky. They just stopped guessing. Split your tests, change one thing at a time, and write down what you learn. In a month you'll know more about your customers than most of your competitors ever will.
Ready to Get More Customers From Your Ads?
We set up and run Facebook and Google ads for local businesses every day. The ads, the tracking, the follow-up texts, and the booking are all done for you.
Get Your Free Growth Strategy Session