My Tips for Cost Efficient Testing (18)
It's pretty much accepted that the majority of campaigns fail, and those that don't require a good deal of testing and optimisation before they are profitable.
If you can't test cost-efficiently, not only are you going to wind up frustrated, but you're going to be broke too.
Yesterday I was reading this thread on what to test first, and it struck me that if you took a vote from 100 different affiliates, you'd get a lot of different answers for the best way to test an offer.
Some affiliate prefer to test a certain angle first ("Would You Date a Fireman?", "Single Moms Need Loving Too"), work out if it's a good performer, and then take the campaign from there.
Others swear by the principle of testing the offer first, to the point where they won't even build a landing page until 1000 hits have been sent direct to the merchant to gauge a baseline conversion rate.
A few affiliates go by intuition. They have no guidelines, just a hypersensitive 'feel' for what shows promise and what does not.
It's really interesting to hear the different theories over how best to test an offer, and it'd be great to get some feedback on your own personal strategy.
I've listed my own line of thought below. Let me know how it compares!
My Order of Testing
1. Find a suitable traffic source.
For me, the top priority is to find a reliable traffic source, before I even look at offers to promote. If the traffic is junk, or if the bid ceiling is too high to get profitable through arbitrage, then you're working at an instant disadvantage. There's no good in handicapping yourself before you've even touched the offer.
2. Match the traffic source to a suitable niche.
If you know that you've got good value traffic, the next priority should be to match it to a suitable niche. A lot of this is just basic common sense with the major traffic sources.
Mochi Media… gaming niche.
Plentyoffish… dating ads.
JuicyAds… adult ads.
Where it becomes complicated is when you start working with slightly more obscure networks that may be specialised in one field and hopeless in another.
For example... 7Search.
Some niches are very popular on this platform, others are hopeless.
It's a similar case with the major PPV networks. If you're promoting a credit card based offer, you're going to have a much higher success rate on TrafficVance than you are somewhere like MediaTraffic, where short submits are more suited to the traffic.
The nature of the traffic determines the suitable niches. And if you get this wrong, you are cooked before you've spent a single dollar.
3. Find the best converting offer in the niche.
So, I've found a good traffic source, and I've established that it's suitable for dating ads. The next step in my testing is to isolate the two best offers in the dating niche.
Some people will say the best way to do this is to test, test, test.
Well, that's true - but a cheaper way is to use your eyes and ears, then test, test, test.
Look for examples of existing campaigns that are already running on your chosen traffic source. Model the similar attributes, including the offers they promote.
Listen for chatter of the most popular offers on forums like this, in network newsletters and in correspondence with your affiliate managers.
Use that information to isolate the two offers that are most likely to succeed.
That's not to say you should ignore the rest of the offers, but I like to trust in the judgment of the people who are already spending money.
Once I have my two chosen offers, I'll split test them, aiming for at least 1000 clicks each before I make a decision on which to keep.
Note: Remember to split test the merchant's Landing Page variations, not just the offer. I found this out recently with BeNaughty when one landing page was converting at 6.5% over the rest at less than 2.5%. Doesn't take a genius to calculate that a campaign previous breaking even leaps to a 100% ROI just by discovering the best page to send your traffic to.
4. Find the best angle
A lot of affiliates, after isolating the best traffic source/offer will jump straight to testing landing page variations.
Even though I rank testing the angle as fourth priority, it is still MUCH more important than the type of landing page you use, the copy, the headline, and your images above the fold.
I place a much higher priority on finding the best angle to promote an offer than I do on the content of the landing page. The angle is what distinguishes your entire sales funnel. It is how you attack the marketplace, and the message you use to monetize demand.
If you use an overly general angle, ("Want a Girlfriend?"), you risk being beaten out of the market for irrelevance. If you use an overly niche angle, ("Want to date a washed up Asian Beach Whale in California?"), you risk alienating too much of the market and making arbitrage impossible.
Your angle is… everything.
From a creative perspective, before I think about my landing page variations, I will think long and hard about potential hooks that will dictate the creatives and LPs I use. Assuming all other variables stay the same, it's your hook that is often the difference between success and failure.
I recommend testing at least 4 different angles on an offer.
Here's a more detailed post on how I come up with angles, using the seven deadly sins as fodder for my imagination:
http://stmforum.com/forum/showthread...-7-Deadly-Sins
5. Find the best landing page theme
Once you've nailed your traffic source, niche, offer and angle - it's time to weld it all together using a landing page that defines your value as an affiliate.
You will notice in most niches that there are nearly always 2 or 3 landing page themes that pop up on a consistent basis.
For example, for those selling weight loss supplements, you'll find a fairly even spread between affiliates that favour the flog (fake blog testimonial) and the farticle (MBC Health News exclusive!). These are two different themes that, if you are familiar with programming, you might group separately as a 'class'. They each have traits that distinguish them.
What I see from a lot of affiliates is a failure to test different classes of landing pages.
It's a common mistake to choose one class (just the flog, or just the farticle, or just the four bullet point hard-sell), and then to create a number of variations of that single class - instead of testing the classes against each other (i.e. the flog vs. the farticle)
Before I start playing with the specific mechanisms on my landing page, I always test the different classes. I want to find the best converting landing page theme before I move on to…
6. Find the best headline and imagery to go above the fold.
These are undoubtedly the most important factors of your landing page.
If you measure landing pages of the same class against each other, you will nearly always find that altering these two variables has the biggest difference on your CTR to the offer.
We could go on forever talking about the best headlines, but suffice to say, it needs to be a bold statement/promise that captures the user's attention and builds on your angle of choice.
The image above the fold is equally important. An underrated factor is the caption that goes underneath your image. Did you know that captions under the main image have the second highest readership on the whole page, second only to the headline? That's damn good reason to make your caption count. Use it to hard-sell your USP.
7. Find the best advertising creatives to drive traffic to your landing page.
Some affiliates insist that the banners and text ads we use to drive traffic are more important than the landing pages we direct it to. I'm not sure I agree, but it's easy to see where they're coming from.
A lot of the time - well, every time when you're bidding CPM - the quality of your creative will determine the CPC that you pay for your traffic. Failing to nail down a good CPC can cause us to lose faith in the rest of the campaign before we've even started. It can be a hammer blow to your confidence when the first reported CPCs are x4 what you expected.
Personally, I try not to worry too much about the creatives until I am happy with the other six factors above. If you make a high CTR banner the cornerstone of your campaign, you will often end up frustrated when the wave of clicks fail to translate in to a good conversion rate. It is beneficial, but it won't make or break your campaign.
8. Find the best copy and CTA for your landing page
Once all of the above variables have been fixed up, it's only then that I'll start playing with the main copy on my landing pages, and testing factors like the wording on my call-to-action, the shape of the buttons, the icons surrounding it and etc.
If you are on a small budget, these factors fall under 'want to test', not 'need to test'. I don't recommend burying your head in them until you have a stable campaign with a proven formula for generating profit.
9. Find the best hours and days to run your campaign.
The last thing I test is hour-specific performance. Are there times in the day where the campaign is converting better? Can I shave out certain hours to bump my ROI? Are there certain days that are not worth running on?
The reason I save this until last is because the data is going to be irrelevant, not-to-mention majorly skewed if I start making day-parting decisions while I'm testing everything else.
Continued below...
So, that's about it. A complete breakdown of the factors I consider while testing, as well as a rough sketch of the order in which I test them. Feel free to chime in if you have a different approach, or want to tell me that I'm doing it wrong!
Final Tip: Hunt in a pack
I know a lot of guys are fascinated by how to test effectively, but they are often held back by the restraints of what they can afford.
If this is you, I recommend setting up a small mastermind group with the objective of hunting and testing in a pack.
One of the benefits of using a small, private mastermind group is the money you can save in testing dud offers, dud angles, dud traffic sources and dud everything else. But only if you share the same testing philosophy as a group.
That last point is important. If you have a small group of individuals testing as individuals, rather than using a shared philosophy, you're going to get inconsistent results. A structured approached is *critical* to cost efficient testing.
$100/day is a healthy budget to find some profitable campaigns. But you can get there even faster by teaming up with 3 or 4 like-minded individuals and turning that in to, effectively, a $500/day budget.
Best post Ive read here. This made my registration worth it.
ya gotta say you could ignore every other thread in the forum and read just this one....
one of the best bit of words I've ever read on how to become a successful aff...
I was on the roof with the rifle and scope.
But you have talked me down with this step by step , logical course of action.
There is hope again.
Whose Martin Osborne and where are the pics of naked girls I was promised in this book I just bought?
Originally Posted by sandyone
Next time, next time I promise to be NSFW...
Great book!
Well written and in a friendly informative tone.
Finch, that is quality mate.
I am sitting here trying to think how to optimize this camp right now and you have given me some actionable steps and ideas are flowing! I owe you a beer!
It's a common mistake to choose one class (just the flog, or just the farticle, or just the four bullet point hard-sell), and then to create a number of variations of that single class - instead of testing the classes against each other (i.e. the flog vs. the farticle)
1. You compare the style of landing page.. ie: flog vs straight sell vs whatever
2. Then you make a couple variations of the winning style
solid gold
Great post! Thank you for the insight!
Great post. Your blog is killer, and I've loved all of your premium post editions.
Macro to micro -- this is how testing should be done.
Thanks finch. This is definitely one of the posts I'll print out and put under my pillow 
great post Finch!
a couple of quick questions if you don't mind:
1) the initial two best offers may be available on a few different affiliate networks, should i test different networks at this stage too?
2) you also said "Once I have my two chosen offers, I'll split test them, aiming for at least 1000 clicks each before I make a decision on which to keep" if the answer to my question above is yes, then doesn't it mean i have to run quite a few thousand clicks before making a decision? e.g. offer A is available on 3 diffrent networks and offer B is on 2 networks. does that mean i have to run 1000 clicks each, 5000 clicks in total before making a decision?
3) (following question 2) is 1000 clicks a hard number or it just depends on traffic source and offers(and thus i should get this number from a statistical significance calculator)?
many thanks and have a great new year!
So, I've found a good traffic source, and I've established that it's suitable for dating ads. The next step in my testing is to isolate the two best offers in the dating niche.
Some people will say the best way to do this is to test, test, test.
Well, that's true - but a cheaper way is to use your eyes and ears, then test, test, test.
Look for examples of existing campaigns that are already running on your chosen traffic source. Model the similar attributes, including the offers they promote.
Listen for chatter of the most popular offers on forums like this, in network newsletters and in correspondence with your affiliate managers.
Use that information to isolate the two offers that are most likely to succeed.
That's not to say you should ignore the rest of the offers, but I like to trust in the judgment of the people who are already spending money.
Once I have my two chosen offers, I'll split test them, aiming for at least 1000 clicks each before I make a decision on which to keep.
Good post. Happy you bumped it up.
I would say just use this calculator and aim for 90%+ confidence(http://www.peakconversion.com/2012/0...al-calculator/), but I wonder what Finch would say.
The goal is getting more for your testing buck so it makes sense.
maybe Finch has been busy stacking all the money and doesn't have the time to reply
other veterans may think "oh this is Finch's thread so better let him answer it"
of course another possibility is i asked too many good questions and nobody wants to reveal too much lol.
any help would be much appreciated!
Heh, indeed so - one of the things about AM is that everyone has their own methodology. Some of us are very mathsy indeed (hi), others work on rules of thumb that have been proven over huuuuuge data sets and years of experience.
It's a lot like professional poker - you get very mathsy players, very "feel"-based players, and the entire range in between. And all of those approaches can be extremely successful.
Speaking personally, and not for Finch, I'd say:
1) Yes, if it's available it's a very good idea to test offers across several networks.
2) Reading what Finch writes, I'd say that he's going for 1k clicks each.
3) This is where our methodologies differ! My guess is that Finch uses 1k clicks as a reasonably hard rule of thumb but varies it a certain extent to take account of different types of traffic (pops in particular). If I was doing a test here, I'd use a Bayesian split-test calculator and aim for a very high certainty (95%, maybe even 99%) before concluding the split-test.
Finch, please feel free to correct if I've got that wildly wrong!
Home > Affiliate Marketing Forum >