Category Archives: Split Testing

Split Testing with Visual Website Optimizer

This is a guest post and contains affiliate links from the author.

Split testing is one of the most important things in the CPA game, but its something that many affiliates do incorrectly or just ignore (choosing instead to take the lazy route and copy without improving on something). Split testing to be honest, is a pain to do sometimes, creating the versions and loading them up again and again, but not anymore. Recently, I stumbled across VisualWebsiteOptimizer (VWO). This is the sexiest piece of tech I’ve seen in a while. But lets take a step back first.

For the noobs, split testing (A/B testing) is simply comparing two different versions of a page together with a element changed on one and seeing which does better (and obviously continuing to use this). VisualWebsiteOptimizer makes this simple. Before (when I was a young warthog), you’d load up multiple pages with a rotator script and track which did best then rinse and repeat. With VWO, you simply just put some code onto your page, and set up multiple tests in a what you see is what you get editor. You then add in a goal (like actions or clicks if your trying to improve ctr). Here’s a fast video they made that outlines how easy it is.

The best thing for me about this tool is the auto optimization. In their interface I can set thresholds for my test in terms of stats. When those are hit, it’ll turn off tests that are loosing badly automatically, saving me time and money.

Pricing wise, it starts with a free trial (def worth using even if your not gonna buy), then starts at $49 a month, which isn’t bad for a tool that’s turned into something I use almost daily on my campaigns. That being said, the view packages are a bit low, so this is a tool that you would NOT normally use on ppv or other “crap” traffic as you’d end up just throwing away your budget on pop unders.

There are a ton of other features like heatmaps, visitor segmentation, and the ability to have client logins. Get the free trial of this tool, once you try it you’ll be hooked and it may just revolutionize the way you look at testing.

The customer list includes some huge companies: Microsoft, Rackspace, RetailMeNot, Groupon, Aweber, SitePoint, and many others.

Get you some VisualWebsiteOptimizer here.

An Introduction to Online Testing

This is a post written by CPA, an active member of CPA Hustle (no relation). While PPC.bz does not normally publish articles of value, CPA sent this over to me, and since I have no ideas for posts, this will do.

Remember, get your grind on at CPAHustle.com

If you guys enjoy this post, there will be more in this series on proper online testing.

Proper Testing: An Introduction to Online Testing (Part 1 of the Proper Testing Mini-Series)

Online Testing has become very popular in the last two years. No matter where you go, your favorite blog, your network of friends & associates, or even the latest course on affiliate marketing, everyone is screaming, “Test Test Test!”. While it is good to test and crucial to the success of your campaign or business, it has come to my attention that 99% of those actually putting testing into practice are doing it incorrectly. You’re probably thinking, “Testing is easy so how could someone possibly be doing it wrong?”. Well, testing is more complicated than you might think. On the contrary, once done right, you can see your increase in conversions go up by hundreds of percentage points very quickly.

Why You Should Test

Let’s say you have a campaign that you are running that is having success. You are making a good ROI right from the start so why mess with it? While a site or landing page may have success initially, it can eventually fail. It is crucial to test properly and constantly to avoid running into catastrophic errors in the future. You may have heard the quote, “Adequacy is the enemy of excellence”. In other words, sufficiency should never be ‘good enough’. You should never feel satisfied with your current success and always be looking for ways to improve something even further. Satisfaction is not excellence but laziness.

An Example

Let’s say your testing a landing page for a new CPS (Cost-Per-Sale) offer you plan to launch. You are offering a 7-day free trial with a small S&H charge. You are not sure whether you should charge $2.00, 3.50, or $5.95. You might think you should charge $5.95 because that’s more money in your pocket initially than if you were only charging $2.00. Right? Wrong.

We decide to send 1,000 clicks to each price point with a CPC (Cost-Per-Click) of $1.25. Upon completion of the test, which took about 3 days, we find the following:

$2.00: 173 leads – $346.00 in revenue
$3.25: 94 leads – $305.50 in revenue
$4.95: 52 leads – $257.40 in revenue

As you can see from the results above, the $2.00 S&H fee produced 79 more leads than the next price point, an 84% increase in opt-ins. While the initial difference is only $41.50 in a 3 day period, in a month your looking at $415.00 extra. The main point here though is since your product is a continuity product you also have to calculate that into the equation. Let’s say after the 7-day free trial your customer is billed $49.95 per month at a 67% rebill rate. In a month here are the following totals:

$2.00 – 1730 leads = 1159 (67%) @ $49.95 = $57,892.05
$3.25 – 940 leads = 630 (67%) @ 49.95 = $31,468.50
$4.95 – 520 leads = 172 (67%) @ 49.95 = $8,591.40

By using the smaller price point, you have made an additional $26,423.55 by doing a simple price point test. You also have an 84% increase in leads. You would have never known this if you hadn’t tested! In a 12 month period, that simple test will help you bring in an extra $317,000 in revenue. (assuming they stick around for 12 months)

Improper Testing Vs. Proper Testing

Online Testing (such as split testing) is crucial to max out conversion rates for your page. It is important to not speculate too much but to test instead. Let the numbers show you how the customers feel. While testing is so important, it is actually better to speculate than to conduct an invalid test. Invalid testing can produce invalid results that in turn give you a false sense of confidence in how you plan to proceed. If you hadn’t tested at all at least you’d be more cautious and realize that you might be making a big mistake. If you get back data that is invalid or false, you might run with it and make your long term results even worse. Improper testing can become very expensive and ultimately plummet your chances of successfully reaching your target market.

There are a lot of other issues besides the confidence interval or statistical insignificance. You need to be able to recognize instrumentation errors and historic effects. Validity is the most important. If you don’t have good analysis tools and you don’t have a good way to interpret your data and the insights shown, you will often conduct tests and you’ll only learn one thing when you should be learning many things. You should never learn only one thing within the bounds of your test. There will always be all kinds of lateral insights you can find besides the obvious vertical result you initially were looking for.

The concept of a successful test is to not only test the right way, but to also test the right factors that will produce the best results. A successful test is a test that has been designed properly so that the resulting data provides you with an accurate projection that answers your question. All good tests, especially single factorial tests, are based around an initial research question. If your asking the wrong questions, it doesn’t matter if you get an answer since the answer itself is irrelevant to your factual needs. The goal is not to conduct an effective test, but to conduct a useful test.

Final Thoughts

If you were responsible for conducting a price test like the one above, do you really think could you do it properly? Most of you probably said “Yes” which leads me to the following questions; How would you design this test? How would you determine the right price points that you were going to test? How would you setup or lay out the treatment pages? How would you determine if you had a valid sampling size for the test?
I have a feeling most of you couldn’t answer these questions but if you could, please post your answers in the comments section below. During the next few weeks I will be continuing this mini-series on conducting accurate and valid tests for your affiliate offers, landing pages, or web sites. By the time we finish, you should have working knowledge of how to setup proper tests and produce accurate results and deem yourself an expert.

How to Squeeze Money With Prosper202

When you’re paying for clicks, you really want the most bang for your buck. That way, you know which traffic source, which ad copy, which landing page, which whatever is converting or not.

To be honest, you don’t really need Prosper202 if you have your own solution. If you already took the time to write up your own tracking software, then you probably don’t need to read this post anyway.

Fuck it I don’t think anyone needs to read this post because it’s pretty much common sense.

Step 1 – Setup Tracking Campaigns P202 keeps track of your referrers, which is good for shit like Google Content Network. This data applies whether you’re direct linking or doing landing pages. For this example, we’re direct linking. So, set-up your network, traffic source, offers, and get the link code.

Step 2 – Setup Traffic Campaigns Run a campaign on Google Content Network for one week. Set your budget high, make sure your keywords are targeted and well grouped. Bid whatever you feel is right (Let’s say 30 cents). Let that shit run for a week so you collect the data you need. Don’t be a pussy when 2 days later you’re losing half of your money. Just let it run.

Step 3 – Gather Data After a week, you should have some significant clicks, and you should have a pretty accurate CPC (Cost Per Click: What you’re paying Google) and EPC (Earnings Per Click. Now I remember. )

Prosper202 lets you see your conversions by the hour. The reason you run it for a week is because time data is important too. For example, let’s say your offer really only converts when kids are home from school, like that New Hannah Montana Ringtone.

Be careful not to day-part (Setting up your ad campaigns to run at only certain times) too quickly. If you don’t have a couple thousand clicks to the offer in the first week, it’s probably not a good time to start day-parting, because you don’t have enough data and at this point you could be hurting yourself more than helping yourself. But as the days and weeks go by, you can really solidify this data and then begin to start trimming the clicks at certain times.

Step 4 – Trim that Shit! Let’s say for this example… your CPC is $.22, and your EPC is $.19. You’re losing about 3 cents on every visitor. Log into your Prosper202 account, go to Analyze > Referrers. Now you can see which sites sending you traffic, which ones are converting and which ones do not.

For example, you received 1,000 clicks from some website that only got you 2 leads. $220 spent to earn $20 is a clear loser. Knowing that, you log into Adwords and negative match that fucker’s URL.

You have lost 2 conversions in the next week, but you saved $220.

The idea here is to keep doing this week after week. After 1-2 weeks you will be able to trim websites from displaying your Ads that don’t convert. After a couple more weeks, you should be able to stop showing ads at certain times of day, or days of the week entirely. (If necessary of course)

Eventually you’ll have a profitable campaign. There’s a whole lot of other shit to split test, including the actual offer. This can be done within Adwords and Prosper202, and it’s one thing you should always be testing to. This is what fuckers mean when they say “You have to find the right traffic to the right offer.” This is one way how to actually find that out.