Archive for the ‘Testing’ Category

Earning by thinking outside the box

Wednesday, July 9th, 2008

Read a great blog post from someone named PunditX that I found following one of my backlinks.

It was about how he made $95k in 2 months marketing a product all of us are aware of but few probably
knew it had an affiliate program - Adsense.

His post spells out just what he did and it seems perfectly reasonable to me. I can think of a dozen or more ways to market the product beyond what he did.

The ’secret’ to his success is that he did not assume “oh that must be saturated” or “that will not work”. he went out and made it work.

He even states he tested 35 different landing pages - which is probably a huge driver in his success. Not being lazy is a critical component to affiliate marketing success.

Too many in affiliate marketing assume that they have to follow the crowd or that their AM knows whats best - the entire time I was focused on affiliate marketing I never had an AM I spoke with regularly. There are some awesome deals on CJ noone has ever heard of and most of the CPA networks are in business for themselves first.

I know a guy who markets $.05 clicks to small niches that then turn into Ebay sales. I know another guy killing it in the Wal-mart affiliate program marketing just a few select products via PPC. There are a million things that are not dating or ringtones or <flavor of the month> that are both sustainable and scalable and that do not require super technical skills.

The best advice I can give anyone wanting to get into this space is to get outside the zone of the affiliate blogging ‘experts’ and start thinking for yourself. If you never read another blog post about affiliate marketing, you will probably be better off than if you read the top 50 affiliate bloggers religiously. The honest truth is that most affiliate blogs are full of junk and one hit wonders.

Testing is for the Rich

Friday, May 30th, 2008

Cannot tell you how often I get a PM or email or AIM message from someone who says they just cannot figure out how to scale or make any money, that all the campaigns seem so saturated, <insert excuse here>.

First thing I always ask is what kinds of things they are they testing - not the offer, but rather A/B type tests? Invariably the answer is a variation of “I plan to start once I find something with a really good conversion rate.”

I am here to say that Conversion rate is not born, it is grown - very slowly over time, with hard work.

Lets take an example of how someone can make a difference by testing.

  • Start with a 3% CR which is considered about the industry average…
  • Assume $.60 CPC which seems absurdly low to me, but lots of you guys are paying a lot less…
  • Assume a payout of $20 per sale - seems fairly generic, lots of products pay more, lots less…
  • If these numbers seem totally wrong to you, just set something up at breakeven for what you are seeing and build a spreadsheet…

After 100 clicks, you would expect to have 3 sales with $60 in revenue and $60 in cost - a total waste of time, right? Just move on to the next thing is what almost all people would do.

But the affiliate who likes to test might try a new headline, maybe some new ad copy, different hero shot or calls to action or any of 100 other things…

Say that this affiliate was able to increase conversion through testing by just 5% each month. Just to show how possible this is, we are often looking for 10-15% increases and seem to find at least one if not several every month! So 5% seems like you are not really trying hard to me…

So the 2nd month, you would earn $3 more than you spend for each 100 clicks - a 3% ROI. Not going to quit the day job, but positive progress. At the end of the year, just finding one 5% increase each month the ROI on this “breakeven” campaign works out to be 42% which is almost certainly worth keeping.

This assumes that you do not manage to increase your CTR or quality score or otherwise lower your CPC which is very doable through testing.

Now say, we have a very clever affiliate who manages to find 10% conversion rate increases each month rather than 5%. At the end of the year, this guy has a 185% ROI from the campaign most of you would have walked away from.

Now if you are getting that kind of ROI, you have probably passed up the affiliate network and gone direct, which will be a free 10% lift, plus increased their payout above where they started, potentially significantly, which allows you to pay more, thus increasing your position, your test bandwidth and frequently your conversion rate in a beautiful cycle.

When you are starting out it is not ridiculous to think you might get 20% or more lifts from some tests, and trust me when I say that front loading the conversion increases significantly improves the end results.

All of the above being said, I am not trying to say that any campaign can be a winner. Nor am I saying that you should stick with a loser. I am merely saying that most people fail because they do not test their way into being a winner ad discard lots of things that would have worked with a proper test plan.

Domain and Error Page ad’s workaround

Saturday, November 17th, 2007

To follow-up on a recent post about domain and error page exclusion ability, we think we have a possible workaround that will enable us to bid on content, domain ads and error ads separately.

We have enough evidence to suggest that the conversion rate on these items are vastly different, and also are susceptible to fraud on different levels as well. As part of our philosophy we always want to bid on the lowest common denominator whenever possible. Match types are in different campaigns, almost all keywords are in their own adgroup etc.

We just put this live so I cannot guarantee that it works, and in fact am sure that Google’s screwy content system is likely to mess it up in some way, but in theory this will work.

We start with adwords editor and copy and paste the enitre content account into a spreadsheet. We then alter the URLs (that is our tracking mechanism) and upload the account twice into campaigns called error ads and domain ads. At this point we have 3 identical campaigns with same bids, ad copy etc.

We then turn off domain and error ads in the original campaign so that it is now content only. In the domain group we turn off error ads so that it is content and Domain ads. Similarly we turn off domain ads in the error ads group leaving it with just Error ads and content.

At this stage we need to eliminate the possibility of content ads showing in the 2 new adgroups. The way we accomplish this is by running a performance placement report and then negative matching every domain that shows up in both the error and domain groups.

This leaves us with 3 campaigns, one that is content only, another that is Domain ads only and a 3rd that is Error ads only. I can now adjust bids and monitor performance for each type of ad and bid according to the return I am seeing from these different tranches of content.

LIke I said it just went live and is not fully battle tested, but on paper it should work.

Fun with Content

Friday, October 5th, 2007

I have been paying attention to content recently and it has really been paying off nicely.

One of the things I have been doing is making sure that my ads are prominently displayed on websites that are getting traffic for the keywords I am bidding on in the major search engines. So I was spending some time clicking the natural listings and then looking for adsense on them.

I was mentioning this to a friend of mine who we can barely talk to me without him suggesting a script or automating stuff. The amount of stuff I do manually drives him up a wall. So a few days later in my inbox I had this nifty little tool that basically allowed me to upload a list of keywords and then the script would automatically check the first few pages of search results on Google, Yahoo and MSN and then tell me which pages ranked for what terms and which pages allowed adsense ads.

I quickly fired it up and ran it for all my best keywords only to find that my crack content team already had us pretty much showing up on these placements. There were a few sites where we went and fine-tuned for better position or to show up on, but for the most part we were on top of things.

Then a few weeks later I was in a training class for some new hires and we were covering a list of words we are not allowed to bid on for competitive or legal reasons and one of the trainees mentioned it was too bad because we were probably losing a lot of profitable volume. A few hours later, back at my desk, I got an IM from my friend asking how his tool had worked out.

That was when I decided I needed to fire up the tool and look at my pretty extensive list of prohibitive keywords. Eureka! I was all of the sudden able to target dozens of websites that ranked for terms I would love to bid on and could not. These clicks came at bargain prices and converted really well.

Moral of the story… Look to the content network to pick up volume on keywords you are not able to bid on for whatever reason. Then doublecheck to make sure you are showing up on the sites that rank for those keywords.

Super Crunchers

Sunday, September 16th, 2007

Sorry about infrequency of posting…. To be perfectly honest, I have been working on a lot of stuff that I won’t share because it is either ultra proprietary or just pretty dull. I made a decision at the very start not to post just for the sake of posting…

I was walking through a bookstore the other day and ran across a book written by one of my favorite Forbes columnists named Ian Ayres (sure wish they would have an article in each issue). The book is called Super Crunchers and although it has very little actual math in it, it is about how people from all walks of life from advertising to politics to medicine are using math and large data sets to make decisions.

One of the things I loved most was that he actually named the book by buying adwords on a few relevant keywords with the proposed names of the book as the title for the ads and then measuring CTR to determine what got the best response.

The book is less about the math and more about the ideas and implementation based upon math. I think it is like $20 bucks retail and $16 or so if you are an Amazon freak. As you probably already know, it is my opinion that if you are not testing you should not be in the business of online marketing.

If you are serious about this business, you will not learn a single thing about online advertising, but you might make a few million dollars from the brainstorms and idea that come from this. Which makes it a must read in my book.

Google Premium Testing

Monday, March 26th, 2007

Google has been testing the color of the premium boxes on Adwords. This is the first time I have seen it.

Yellow Bar

The interesting thing to me is that in the screenshot, the yellow box looks clearly yellow.

However on my monitor, the yellow box on an actual search is only visible when you look at the screen from an angle. When you view the screen head-on the box is completely transparent and the three ads look like they could be natural results. Yes, the formatting is a little different, but I look at this stuff all day everyday and my initial thought when I found this was that this was a search without premium listings. I only noticed it because I walked away from my desk.

Is this an optical illusion? Are others seeing similar things? Is this an unintended consequence or a very specific color designed to appear transparently? Is it a chance outcome based upon my monitor settings, angle of the monitor, and lighting at my desk?

As an advertiser, I very much like being in the blue box, but I am somewhat concerned if people are being tricked into thinking that my ad is a natural result.

How granular should you track?

Friday, March 9th, 2007

I see no reason why every single keyword should not be in its own adgroup. Why every single ad copy should not have its own display URL. Or why each match type should not be tracked differently.

Yes, it is a pain to set-up. Yes it is anal. Yes it matters. It can mean tens or hundreds of thousands of dollars a year difference.

The difference between singular and plural keywords can be huge. It is not out of the realm of possibility for the plural version to be a home run, while at the same time the singular version of the keyword could be bleeding you dry. Or the broad match is getting killed and the exact match is a huge cash cow.
(more…)

Writing Ad Copy For Idiots

Thursday, February 8th, 2007

An interesting thing I have discovered across my career in marketing is that I am not my target market.

I am not the most eloquent or educated person around, but I feel I have ok communications skills and I have always considered myself a little bit above average, but by no means the smartest person in the room. Drunkest, yes. Smartest, No.

What never fails to shock me is that the stuff I make fun of and ridicule in the creation process almost always seems to be the best performing campaign.
(more…)

Testing, Testing, 1.2.3.

Friday, February 2nd, 2007

It is absolutely amazing to me when people say “I tried XYZ, and could not make it work, you must have been lucky.”

I am not lucky, I am diligent, I take calculated risks, I learn from my mistakes. I can only think of one instance where the first thing I did came out of the gates a runaway winner. Just about everything since has lost money or broken even right out of the gates.

If you give up a niche without first testing at least 4-5 landing pages, as well as several different ad copies then you are not trying. Sure I have had a few flyers that crashed and burned that cost me some money, but it is rare the field that I cannot make work to some extent.
(more…)

Content is King?

Friday, January 26th, 2007

We recently made a strong push into Google content in a new account and noticed some interesting things.

  • Our volume in each account was fairly similar
  • Similar keyword sets
  • One account focused on many keywords in a single adgroup
  • Other account focused on 1 keyword per adgroup
  • We have yet to see a page where both accounts content displayed, despite different Display URL’s
  • (more…)