Affiliate Sales and Hop Links: Cloaking, Tracking and Hijack Detection

[xiaophil]

There's been a lot of talk recently about the state of affiliate tracking, in particular regarding one certain high-profile product vending site. The speculation ranges from concerns over link theft, malware and cookie-stuffing through to full-blown conspiracy theories about the company itself, and culminating in the "change your id to restart sales" theory.

At least once a week there is a post asking about redirecting, hiding or 'cloaking' affiliate links, and the number of products available which claim to achieve this attests to the fear and concern harbored by many affiliates.

I am going to cover two separate topics here today, and the benefits to you are as follows:

I'll discuss ways of detecting some attempted compromises of your affiliate links. This will either prove that nothing is astray and enable you to sleep better, or give you some hard facts and reason to investigate further any possible inconsistencies.

Also, I intend to put your mind at ease regarding those sales figures which don't seem to follow any pattern: stopping and starting of their own accord and bouncing around even with consistent levels of traffic. This will involve a tiny bit of math, but it's easy enough. People's responses to what I will show you have included disbelief, denial, shock, confusion and more disbelief, but once you understand this simple principle you will gain a new insight and never view your stats quite the same way again.

First up though, let's take a look at some basics and the different ways that someone might possibly attempt to hijack or otherwise compromise your affiliate link.

By the way, I think most of the following tactics are only really popular among the Internet Marketing crowd. I believe that non-IM and especially non-technical niches are pretty much immune to such exploits as the market is oblivous to things like hoplinks, cookies and redirects. The obvious exception would be the malicious scripts as outlined below.


The Drive-By Copy & Paste

The perpetrator hovers over your raw unencoded affiliate link, right clicks and copies the link. They paste the link into a browser and change the affiliate id for their own before proceeding with the hop.

Result: Your hop is not registered as your link has not been clicked. Your commission is well and truly stolen, and this cannot be detected.

Remedy: Use an encoded link or a redirect or both. This means your hop will be registered as the link must be clicked to discover the target of the redirect and hop. A determined thief will probably still prevail, see the next method.


The Hand In The Cookie Jar

After landing on the sales page via your hoplink, the perp researches and discovers the vendor id, clears their cookies and rehops back to the page under their own affiliate id.

Result: Your hop is registered but your commission is stolen. This appears as a reduced conversion rate for that product (hops/sale).

Remedy: Some possible means of mitigating this compromise include: 1) Using a redirect script which removes any obvious affiliate markers from the landing page URL or 2) Offer some compelling bonuses for buying through your link. Both of these methods require care to ensure you remain within the TOS of the product vendor.


The Worm That Turned

It is possible that scripts running as part of virus, worm or other malware including toolbars can hijack affiliate links directly within a user's browser.

Result: Although your link is clicked in the browser, it is not registered as the malicious script substitutes it's own hoplink and redirects.

Remedy: No definite remedies as the methods will vary but in all likelihood such activity can be detected using redirects as described below and probably countered, at least for now, by using encoded hoplinks.



A Hop, Skip and a Jump - But How Many Exactly ?

Out of the many posts I have read here related to affiliate links, only a few appear to have a handle on analytics and tracking, and can therefore back up their suspicions with hard numbers.

Without a robust analytics package in place you are limited to speculation and guesswork. What's more you will soon discover that you are dealing with data that often appears to be random in nature and makes no intuitive sense over small intervals. Only by averaging many samples over a period of time can any useful patterns or metrics be drawn.

As a bare minimum it is necessary to track the number of unique visitors to your offer page and the number of unique clicks on an affiliate link. A very popular and free analytics package is provided by Google. There are others which can be self hosted but for illustration I will assume you are using Google Analytics.

Simply by copying the code Google gives you into every page on your site will enable a wealth of information to be captured and analyzed, and for many this is adequate and they take it no further. However for the purpose of tracking clicks on affiliate links I suggest you take a look at the "events" subsystem. I will spare the details as they are well documented, but adding a tracking event to a link is as easy as:

<a onClick="pageTracker._trackEvent('Category', 'Action', 'Label');" href="affiliatelink">Click Here</a>

with 'Category', 'Action' and 'Label' set to meaningful things for you. This will report every click on that link as an event under Content | Event Tracking in the Analytics control panel.

Another alternative for click tracking is to use PHP, which may be attractive if some form of PHP redirect is already in use. There are many scripts available for PHP redirect and click counting which range in price and complexity. A good one should count only unique clicks per visitor.

It is worth noting that by using a redirect of some kind the click can be counted by JavaScript before a hop url becomes visible to the system. This may be an important point if one suspects a hostile toolbar or other malware of hijacking links within the browser. Even a simple redirect in .htaccess will enable a click to be registered before redirecting to the actual hoplink. This type of link hijacking is particularly difficult to detect and without a tracked redirect will remain undetected as the hop will most likely not be registered by either your analytics or your vendor.

Once your tracking is in place it becomes a very simple exercise to compare your hop count to those of your vendor. Any major differences will provide a sound basis for enquiry.



Lies, Damn Lies and Statistics

Now onward to those crazy sales figures. This appears to be a huge source of confusion for many people who are usually looking for some kind of day-to-day pattern in their conversions (whether they be CPA, hops, sales, opt-ins or whatever).

I'll cut to the chase here, explain a little more and then give an example.

Individual events such as whether a particular visitor will buy a product or not cannot be predicted, but after many such events have occurred we can work out the probability of a future visitor making a purchase.

Say you have had 1000 hops and 10 sales. That's 1 sale per 100 hops and we can pretty safely conclude that, all other things being equal (like traffic source and quality) that in the future each hop would have a probability of about 0.01 or a 1% chance of making a sale.

Simple enough so far right? But here's the clincher that many people miss:

The probability of any particular customer making a purchase is completely independent of any other purchases.

What does that mean? Imagine flipping a coin. The chances of landing on heads is 50% right? Now imagine you flip it 9 times and get tails every time. What's the chances of getting heads on the 10th flip? It's still 50% right, in other words the 'sales history' doesn't alter the chances of the next visitor buying.

The figures only begin to make sense and a pattern emerge once we have a large number of samples and average the results.

How many samples you ask? Well the more the better as this will improve accuracy and reduce uncertainties. It's pointless trying to extract meaningful data from just a few samples.



I'm a Model, You Know What I Mean?

To illustrate and solidify these concepts I have produced a simple mathematical model of a web site.

The model has only two assumptions:

1) The web site receives traffic of between 50 and 100 unique visitors per day, determined randomly.

2) Each visitor has a probability of 0.0025 of making a purchase (i.e a 0.25% chance)

This probability is equivalent to an affiliate site with a consistent quality of traffic getting a 25% click through rate on a hop-link to a sales page that converts at 1%. A fairly believable scenario.

The model was set up to run many iterations each of 30 days.

Over a great many runs the sales would of course average out to exactly 0.25% or one sale per 400 visitors. On a day-to-day scale however this pattern is far from obvious.

By the way, the following effect was recently parodied by Harvey.Segal in a thread called "People with glasses: the stats are in". For those who missed it, here's his initial post:

Originally Posted by Harvey.Segal

Here are some stats on people walking past my house wearing glasses

Sunday: 2
Monday: 1
Tuesday: 2

However, get this, yesterday and today - none !

This does not make sense.

It is completely at odds with the law of averages.

Yes - at odds.

Don't believe others who deny it.

Reminds me: last November there were 3 consecutive days of
zero figures. And this followed a summer when the average
was an amazing 1.5 per day.


Something very fishy going on here - does the word 'conspiracy'
spring to mind ?


I'd like to ask members to check for similar results -
together we can get to the bottom of this and contact the
proper authorities (FBI, Better Business Bureau, other IM
forums etc).

Harvey

I like that, good humor.

Anyway, this simple mathematical model run over many iterations can yield some useful and interesting information, such as:

Average Monthly Visits = 2249.64
Average Monthly Sales = 5.60
Average Sales per Visit = 0.00249 (This of course will converge to exactly 0.0025)

I can also tell you that almost 83% of the days have no sales whatsoever but 81% of the months have between 3 and 8 sales.

On average a sale occurred on only 15.5% of days and two sales on 1.5% of days. 3 and 4 sales a day are extremely unlikely but not impossible, as is an entire month with no sales at all.

Here's a 'typical' month, 8 sales so above average, but look at the pattern - or rather the absence of one:

Day:  1 Visitors:  97 Sales: 0
Day:  2 Visitors:  85 Sales: 0
Day:  3 Visitors:  98 Sales: 1
Day:  4 Visitors:  95 Sales: 0
Day:  5 Visitors:  71 Sales: 1
Day:  6 Visitors:  81 Sales: 1
Day:  7 Visitors:  72 Sales: 0
Day:  8 Visitors:  51 Sales: 0
Day:  9 Visitors:  89 Sales: 1
Day: 10 Visitors:  71 Sales: 1
Day: 11 Visitors:  62 Sales: 0
Day: 12 Visitors:  99 Sales: 1
Day: 13 Visitors:  80 Sales: 0
Day: 14 Visitors:  84 Sales: 0
Day: 15 Visitors: 100 Sales: 0
Day: 16 Visitors:  57 Sales: 0
Day: 17 Visitors:  74 Sales: 0
Day: 18 Visitors:  83 Sales: 0
Day: 19 Visitors:  94 Sales: 0
Day: 20 Visitors:  72 Sales: 0
Day: 21 Visitors:  97 Sales: 0
Day: 22 Visitors:  85 Sales: 0
Day: 23 Visitors:  73 Sales: 1
Day: 24 Visitors:  82 Sales: 0
Day: 25 Visitors:  78 Sales: 0
Day: 26 Visitors:  78 Sales: 0
Day: 27 Visitors:  61 Sales: 0
Day: 28 Visitors:  75 Sales: 1
Day: 29 Visitors:  77 Sales: 0
Day: 30 Visitors:  89 Sales: 0

For the first couple of weeks the webmaster is quite happy, after some hard work promoting their new site they now have some pretty consistent traffic and a sale about every other day. But then things just seem to stop. They wait and watch, log in often to check their stats, notice record daily traffic (100) but still no sales and wonder what the hell is going on. After a week of this they log in to a forum and see that someone has suggested there may be a problem with their affiliate ID. After 10 days with no sales (after all they made 6 sales in the first 12 days of the month right?) they get desperate and decide to create a new ID and change their links, and lo and behold the very next day a sale! So it must have been a problem with the affiliate ID yes? It's obvious right?

No it isn't! The data above conforms perfectly well to a long term average conversion of 1 sale in 400 visitors or 0.25%. It's because of the low number of samples that the webmaster is 'zoomed in' too far and can't see the forest for the trees.

By the way, with 800 uniques a day that same model yields 60 sales per month on average but you could still expect about 13.5% of the days, or about 4 days/month, to have zero sales.



I've tried to explain everything as clearly as possible and hope all this makes sense. It's aimed mainly at the newer folk and people who may not have run into statistics and probabilities before - hopefully someone out there will find this information helpful.

Best Regards,

[Tony Dean]

In other words traffic is fickle?

[xiaophil]

Originally Posted by Tony Dean

In other words traffic is fickle?

Do you mean the quantity of traffic or the buying decisions of the people behind it?

Either way it may be true but that's not what I am getting at.

Let's say hypothetically that you have exactly the same quantity of traffic every day and each visitor has exactly the same probability of making a purchase.

The resulting sales would still look random.

The point I am trying to make is that unless the site has a very large volume of homogeneous traffic then it is pointless trying to look for any kind of pattern in short term data.

[kkchoon]

How much traffic is homogeneous? 300 or 3000 unique visitors?

[ebuyer123]

Bad econ make people THINK twice x 2 before pressing the order button.

More clicks (good for adsense publishers) but less buyers in general.

[SamSnead]

Great way to breakdown my uncertainties, you hit it right on the spot. Plus I guess more than anything it brings back hope.

[joelraitt]

It is a little confusing...so to some it up...in the short term you really dont know squat...but long term with lots of visitors you can sort of predict what will happen?

[GeorgR.]

The point is,

we had debates like this all the time...in regards to "XYZ's tracking is off, they might be shaving. I see irregularities which are NOT to explain statistically." (Eg. two weeks without a single sale after usually having 2 sales/day, things like that).

So...i started marketing with another vendor, or CPA, due to exactly such issues and complaints...but i see exactly the same thing happening.

Non logical sales patterns. Like: A few days in a row constantly 20% conversion rate....then some days with hundreds of clicks and NO single conversion.

As for the "hijacking" and "ID theft problem"....yes, this might be the case in the tight IM niche, if you are an IM, sell an IM product to IMs. But also THERE i think its still far away from a reason to panic.

As for other niches, mainstream niches...you would be surprised how many surfers probably dont even know what an AFFILIATE *is* - let alone would bother to "exchange" their IDs, things like that. Your average "mom at home" shopping for a coupon....or baby clothes....how big do you think is the danger that something happens with your affiliateID?

I personally think this issue is way overrated - it might have way less negative impact on sales than what people actually think.

[kkchoon]

I must admit what Phil said is probably very true!

Just can't blamed the people who think bad about the affiliate network, because when upgrade of browser, new anti-spyware, anti-virus that blocked cookies and some famous affiliate networks such as Clickbank and CJ.

Sometime clickbank did have some problems with their tracking, but they work fast to rectify the problems, at least most of the time.

This triggers many affiliates doubt about their sales and commission. The economy did impact the online sales, some see increased in sale while other see decreased...

There are a lot of factors we need to consider, not just blaming the network, and much much more statistic needed before jumping into conclusion...

Sometimes even with 1000 hops may not be conclusive, but good thing is we can share it here...

[ebuyer123]

Originally Posted by GeorgR.

The point is,....
As for other niches, mainstream niches...you would be surprised how many surfers probably dont even know what an AFFILIATE *is* - let alone would bother to "exchange" their IDs, things like that. Your average "mom at home" shopping for a coupon....or baby clothes....how big do you think is the danger that something happens with your affiliateID?

That's absolutely true...