Optimization and Split Testing: How LONG and How MUCH Should You Spend Testing Affiliate Marketing Offers?
You should know by now that the #1 most important part of the campaign is the offer.
Remember, the offer is the product or service that you’re promoting as an affiliate.
If your offer sucks, then no amount of copywriting, angles, or ninja Facebook techniques can save your campaign.
Here’s an analogy to help you understand how important the offer is to the whole process.
Imagine that you’re a brand marketing agency.
You create an AMAZING campaign for Ferrari. But for whatever reason, Ferrari decides to end the partnership with your agency.
Sure the offer’s gone, but you still have angles and places to advertise that are proven to work.
You could keep the campaign running if you replace the offer. You have to make sure that the offer you replace it with is just as good.
This can work if you’re working with an offer on the caliber of Lamborghini.
But what if you can’t get a Lambo tier offer? The only car manufacturer that’s interested in this campaign is Fiat.
Sorry, but Fiats aren’t sexy. It’s going to be difficult to get the offer to work using the same “angles” and strategies used in the original campaign.
Does this make sense? I wanted to illustrate how important the right offer is to the process.
Back to affiliate marketing.
You decide to run a “solar energy” lead generation campaign.
How do you find out which offer is the best (most profitable) when you have 10+ options?
The only way to decide what’s the best offer is to conduct a split-test.
You pick a few offers to test, and then you see which one generates the highest revenue. Sounds easy enough right?
Well, a few questions probably come to mind.
You might be wondering…
How many days should you spend testing an offer before declaring a winner?
How much money should you spend testing an offer before declaring a winner?
These are things I wondered about when I first started, and I’ve noticed that these questions keep coming up over and over again.
I think our brains are programmed to want formulas because they help make sense of the chaos.
I cooked spaghetti the other day. I wondered, “How much pasta should I cook for seven people?”
I found a “formula” online that says it should be 2oz of dry pasta per person.
7 people x 2oz = 14oz of dry pasta.
The box told me to cook it for 9 minutes in boiling water for “Al Dente.”
I followed these formulas and I got the portions right! And the pasta came out perfectly!
I’m not a master pasta maker…I just found the right formulas to follow!
Wouldn’t it be awesome if affiliate marketing worked the same way?
Oh man…I wish.
But unfortunately, that’s not the way affiliate marketing works. At least not in this case.
I can’t tell you that you need to test offers for three days before declaring a winner. What if the traffic source you’re using is NOT delivering an adequate amount of traffic?
I can’t tell you that you need to spend $500 before declaring a winner. If you’re testing higher payout offers, then that’s not going to be enough data.
So, what are you supposed to do?
This article’s going to give you my method on how I determine if an offer is the best, and how to use it.
The Only Way to Determine a Winner
There’s only one way to gauge if an offer is better than another.
You need to test long enough and gather enough data for there to be a statistically significant conclusion.
I don’t ever want to hear about “how long” or “how much to spend” ever again!
How do you actually take action on this? You need to use a statistically significant calculator.
By the way, if all this statistically significant talk confuses you, I recommend that you read
Back to Basics: Statistical Significance
How to Use a Calculator
First of all, I rarely use calculators these days. I have enough experience to eyeball my results.
Whenever I cook steak, I need to use a thermometer to make sure it’s medium rare. My dad has enough experience to poke the steak to determine if it’s medium rare.
If you don’t have enough experience yet, then you should use a split test calculator. There are hundreds of them online.
The one I recommend is:
Why? Because it’s simple. Some of the other ones I’ve seen give too many unnecessary options.
(Note: I know someone’s going to ask if they can use X calculator on whatever site. Use whatever you want, but you’re overthinking it. )
This is what you see when you load the website up.
Here are some terms to help us get on the same page.
Visitors: How many clicks there are to the offer’s page. You’ll find this in your affiliate network stats.
In a perfect world, the number of visitors to each offer page would perfectly match. But due to different technologies and variabilities, these numbers are usually different.
Conversions: How many people signed up for the offer. You’ll find this in your affiliate network stats.
Control: The original offer that you’re testing.
Treatment 1, etc.: Additional offers that you are testing.
The order doesn’t really matter, these are labels.
Confidence Interval: Remember that we’re running an experiment. The confidence interval is the percentage that you’re going to be right in the future.
Would you go to a heart surgeon that has a 60% success rate? Probably not. When it’s important you want to be confident in the results.
The standard that most experiments settle for is 95%. This is a key number to aim for.
Now that we have some definitions out the way, lets run a few scenarios to help you understand how to use this tool
Scenario 1: Not Enough Data
In the first scenario, you can see that the confidence interval is at 62%.
What’s our goal again? 95%.
Action: Keep the split test running longer. You probably don’t have enough data yet.
Scenario 2: You have TOO much data
In this scenario, you have a 100% confidence level. That’s bad.
The problem? It’s obvious that the Control offer was performing the best. By running the split test too long, you lost money due to opportunity cost.
If you had all the traffic running 100% to the control offer sooner, you would’ve had more conversions overall.
Scenario 3: Just Right
In this scenario, the control is the winner.
You’ve tested to get enough data, but you stopped it as soon as you reached 95%.
Combining Art and Science into Affiliate Marketing
Testing affiliate marketing offers is one of the most important actions you can take.
But you just can’t throw shit at the wall and trust your gut instinct. The more you employ scientific methods in your campaign testing, the more accurate your results will be.
I’ve always found it fascinating how affiliate marketing is a combination of art and science.
Coming up with angles, copywriting, and designing the landing pages are art. Campaign analysis, understanding the data, and doing proper split tests are science.
Combine both and you’ll be raking in the money.
Featured image by IgorTishenko