It seems universally accepted by affiliates, networks, tracking platforms platforms, agencies and even advertisers, that “affiliate marketing doesn’t track all the sales it should do”.
This is blamed on a confluence of very different, yet intertwined issues, which I’ve listed here in no particular order of importance:-
Ad-blocking
Issues transferring tracking signals using http/s requests
De-duplication across different marketing channels
Consent requirements
Cross-device journeys
Cross-platform support for web-to-app and app-to-web journeys
Affiliate tracking in embedded web experiences, such as on Instagram or Facebook
Impact of third-party attribution or measurement providers
Third-party cookies are not on this list as all major affiliate tracking solutions have long since stopped relying on them.
But based on the above list of myriad issues it’s safe to assume that affiliate tracking is not 100% reliable. Let’s state for the record that ‘we’ (the affiliate industry) all believe there are transactions that affiliate marketing should track, which it doesn’t. With that point agreed, let’s look in more detail at what this affiliate tracking issue really looks like.
Defining ‘Universal’ and ‘Specific’ tracking issues
Some issues affect the industry in a universal way, i.e. they will impact all advertisers and affiliates relatively equally, while others will exist due to the specific set-ups of advertiser / affiliate partnerships. Here we’ll divide them into groups:
Universal Impact, an issue which affects all advertisers and affiliates
Ad-blocking
Issues transferring tracking signals using http/s requests
Cross-device journeys
Affiliate tracking in embedded web experiences, such as on Instagram or Facebook
Impact of third-party attribution or measurement providers
Specific Impact, an issue specific to an advertiser’s tracking set-up
De-duplication across different marketing channels
Consent requirements
Cross-platform support for web-to-app and app-to-web journeys
You could broadly say that Universal Impacts are industry problems, and it’s down to technology providers with a vested interest in improving the accuracy of tracking to solve them. It has become de rigueur to assume that affiliate tracking is antiquated technology no longer fit for purpose. But that’s a cheap shot, which fails to credit the innovation in tracking that has happened over the last decade to solve issues like third-party cookie reliance and keep pace with hardware development.
Unfortunately tracking innovation has generally been two steps behind the curve and the industry has generally been found chasing its collective tail to plug gaps in tracking technology. An example was the sudden proliferation of ad-blockers in the early 2010’s and Safari’s first iteration of ITP in 2017, which very much caught the industry on the hop. More recently, the affiliate industry’s lack of input to the public codas that have formed the basis of Google’s Privacy Sandbox, in particular the Attribution Reporting API, could mean we have missed another chance to shape a technology that will impact our tracking into the next decade.
The Specific Impact issues are problems that arise from how an advertiser implements their tracking:
How affiliate tracking cookies are treated by an advertiser's Consent Management Platform (CMP) is also a decision made and implemented specifically by each advertiser, as is how the actual CMP operates on the advertiser's website.
Cross-channel de-duplication on an affiliate's sales is an advertiser's decision and can be implemented in a very heavy-handed, unfair way
If an advertiser allows deferred deep-linking - sends users into their app if it’s installed when web links are clicked - but does not track affiliate sales in-app
I present these Universal and Specific impacts to demonstrate that affiliate tracking’s many challenges are not all tributaries flowing from a single point of failure, but a variety of standalone issues with different causes and effects.
There are two important points that link the Specific Impact issues. Firstly, they are the issues with by far the biggest impact on tracking accuracy. Secondly, they aren’t challenges that need to be solved by technology. These issues are born out of strategic decisions, normally taken by advertisers.
Because these Specific impacts are about theory and implementation, rather than technology, the obvious conclusion is that they are the lowest hanging fruit and the points where the industry can make the most gains if we find workable solutions.
Well, yes…and no. It’s tempting to say that advertisers should just not deduplicate or put affiliate tracking behind a consent wall and then affiliate tracking would be infinitely better (and by better we mean track more!). But these decisions are not straightforward choices for advertisers.
It’s pretty obvious that while consent implementation is ultimately the responsibility of each advertiser, these decisions are made at the behest of misguided regulators who (with the exception of loyalty partners) believe there is no legitimate rationale for affiliate cookies to be allowed as strictly necessary. So while consent is an advertiser-led implementation, most advertisers are simply following the advice of regulators.
Cross-channel de-duplication is a much more complicated issue. When we introduce this topic into the affiliate tracking discussion we start to see just how inextricably linked tracking accuracy and the value of the channel really are.
What does de-duping tell us about affiliate marketing’s value?
De-duping affiliate sales against other online marketing channels is an advertiser-led decision. Normally it is done on a ‘last-click-wins’ basis, but other methodologies are occasionally used. And yes, there are some overzealous de-duping policies which are unfair on the channel.
Over the last decade this practice has been the number one reason for affiliate marketing losing out on sales that the industry believes it should be credited for. It is the most impactful affiliate tracking issue of them all - Universal or Specific. You can also very squarely pin this issue to the doors of advertisers. Each advertiser makes their own decision on if, and how, cross-channel de-duping affects affiliate sales. It’s very tempting to think this is the issue that is easiest to solve. Just make advertisers stop de-duping!
But to think this is to miss the point about why cross-channel de-duplication exists in the first place. It is a response from advertisers to a channel that does not engage with enough customers that have not already been on advertiser websites before purchasing. The APMA’s recent State of the Nation survey said that 35% of industry sales are credited to loyalty and discount sites. The PMA’s survey in the US had the same number at a much higher 68%. Coam’s work on how affiliate marketing introduces versus converts consumers suggested that Affiliate engages only a third of the consumers that it converts. Saying this isn’t to degenerate the channel.
Affiliate marketing has already tried to tackle the potential lost revenue from cross-channel de-duplication in the form of tenancy payments, which are an ever-increasing share of channel revenue in-part due to affiliates seeing the need to make-up for lost commission due to tracking. Advertiser-led de-duplication policies are a direct response to concerns that spending doesn’t directly equate to value in affiliate marketing. De-duping is the unspoken message from advertisers that the industry’s biggest tracking ‘issue’ has much more to do with value perception than it does technology.
What is a valid affiliate sale?
This question introduces us to an ethereal but very important term in the affiliate tracking discussion - valid. What is a valid affiliate sale? This term means very different things depending on what side of the advertiser / affiliate / network ecosystem you fall, and it has a big impact on how much we really believe is being lost from this rabbit warren of tracking issues that are supposedly blighting the industry?
It’s true to say that challenges to accurate affiliate tracking do exist. Some are long-standing and some are newer, such as the impact of cookie consent preventing an affiliate tracking cookie from being set.
Every time you throw a new tracking issue into the mix - like consent - the potential for missed sales gets greater. But along the way the industry has also resolved a number of issues that have been problematic for accurate tracking in the past, such as third-party cookies, ad-blocking and the failure of http post-backs. Affiliate tracking is not just a stone rolling downhill, picking up one issue after another as it goes. It is a continuously evolving technology.
The reason the term valid affiliate sale is central to the affiliate tracking dilemma is that affiliate tracking has never been 100% accurate and is not less accurate now than it has been in the past.
I feel like we need to let that point about accuracy sink in and then reiterate it - affiliate marketing tracks, reports, books, charges advertisers and pays affiliates for sales that never had anything to do with affiliate marketing. Post-purchase tracking on the web has never dealt in absolutes, and how that specifically relates to affiliate marketing is really important when we consider to what extent the channel is losing out.
Probabilistic tracking makes good on Affiliate’s shortfall
Affiliate tracking should always be seen as a bucket that networks and technology providers are trying to fill up with as many sales as possible, tracked as accurately as possible, and as deterministically as possible. In our world, deterministic means that there is a hard match between a user interacting with a piece of affiliate advertising and then making a purchase.
But some sales have always leaked out of that bucket due to missed, blocked or non-functioning tracking (see the list earlier for the variety of reasons this can happen). That was as true in 2010 as it is today. To compensate for the potential of missed sales affiliate tracking has used a variety of secondary tracking methods to claim credit for sales. Polite terminology might call these methods probabilistic - a combination of signals like IP address, user agent, device type, location and encrypted personal data that are used to match a user's click to a sale. Collectively these tracking methods have amounted to a ‘best endeavours’ approach to find sales that ‘may have been driven by an affiliate click’ if the hard match of a deterministic signal like a cookie is not present. Probabilistic matching has long played an important role in keeping affiliate marketing’s bucket topped-up. Over the years between 5% and 12% of total industry sales have been tracked using these probabilistic methods, and the practices are still commonplace today.
The challenge in probabilistic matching is its accuracy, or lack thereof. Fingerprinting an online user using probabilistic signals has always been an imprecise science and has generally been getting less accurate over time due to the shortening of IP ranges, fewer opportunities to customise user agents, the Android / Apple mobile device duopoly, proliferation of discount code matching on multi-use codes and also tighter privacy restrictions. This has led to more false-positives and more incorrectly tracked affiliate sales.
The affiliate industry consistently sees tracking as a half empty bucket, leaking from all sides that we are unable to fill up. But that isn’t the reality. Affiliate’s bucket is still full, sometimes with sales that have not actually been referred by affiliate marketing. The perception of affiliate marketing’s tracking problem, and who is responsible for it, is totally backwards. Sometimes affiliate tracking doesn’t work; and sometimes it tracks sales that it shouldn’t. In most cases these two opposite sides of the scale exist in happy equilibrium. Don’t get me wrong, there are some advertiser tracking set-ups that due to a conflagration of the issues described earlier will be under-reporting affiliate sales. But for every one of those there are also advertisers that have the ‘probabilistic sweet-spot’ in their tracking set-up who will be over-indexing in the sales that are credited to the affiliate channel.
What is the scale of affiliate marketing’s tracking problem?
To end, let’s return to the beginning and the key question - does affiliate marketing have a tracking problem?
Yes, in the sense that we have shown there are very real issues that cause affiliate tracking not to function in a way that the industry would expect it to. Cookie consent is a great example of a very real threat to affiliate tracking, which will get a lot worse over time as more advertisers put their affiliate tracking behind consent walls. .
But a key point here is that currently the impact of affiliate tracking issues are not uniform across the industry, and while there are some tracking set-ups which absolutely under-report, affiliate marketing is also doing plenty of work in the background to back-fill for lost sales.
We talked earlier about the scale of affiliate marketing tracking’s problem. Let’s finish this discussion by putting the scale of the issue into context.
I estimate that tracking failures cost the industry about 4% of its total UK spend (£64 million). This estimate does not include the issue of de-duplication, which are sales that the advertiser does not intend the industry to track in the first place, and so should be considered separately.
Crucially I do not believe the percentage has changed significantly in 15 years (although the monetary amount will have gone up as the industry has grown). I would say that up to 3.5% of total UK industry sales (£56 million) could be tracked incorrectly due to false positives based on probabilistic matches. So net-net, I would say the industry is looking at a shortfall of £8 million.
Okay, that kind of money isn’t hiding down the back of your sofa. But I’m also pretty sure it’s a lot less than the industry would like to think.
We are misunderstanding the problem.
Affiliate tracking technology is about continuously evolving to co-exist with how people use the internet, not looking for a ghoulish issue around every corner. The message about affiliate tracking should be that we are doing just fine, and where there is a fundamental challenge to accuracy, we will meet it and deal with it.