They were never as good as the edible ones.

I wrote an article a few months back about how to survive the shift to a cookieless world and why in some ways I’m actually excited about the shift back to a more traditional marketing approach.

Since then I’ve had a few people ask me if I’m concerned about the changes. What will it mean for campaigns? What will it mean for trust in marketing? What will it mean for sales?

I don’t think anyone should be remotely worried about losing cookies.

The truth is, they never really worked that well anyway. As far as tracking performance is concerned, what a cookie tells you is that a particular device did a particular thing. A cookie does not tell you that one thing led to another, it shows correlation not causation, and as anyone who has used retargeting ads will tell you there is a very high likelihood that at least some of the people who supposedly purchased as a result of seeing your ads would have done so anyway.

Indeed, a lot of the time people see your ads as a result of purchasing. Not what cookies were intended to do but a direct consequence of their inadequacy.

Cookies Are The Easy Fix

The marketing world adopted cookies as the default way of tracking because of technical simplicity. They are easy to install, simple to understand, have no noticeable impact on user experience, in most cases they are free, and they give all marketers the ability to say “I did this and achieved that”. For the uninitiated, this simple statement is incredibly convincing.

There are better forms of tracking. I’m not talking about these platforms that claim to fix attribution issues (I’m very happy to discuss what I think about these for anyone who cares to listen, hint; they are not all they’re cracked up to be), rather tools like split audience testing, econometrics analysis, on/off testing, even UTM tracking to a certain extent can be more useful. The problem is, they are all time-consuming and therefore expensive. Why pay a data agency to run a proper analysis when you can just install a cookie?

Cookies Do Not Account For Incremental Sales

In 2015, now almost 10 years ago, I was running advertising for Topman and was invited by Facebook to run an uplift test to ‘prove’ the value of advertising on their platform. The reps from Facebook had been keen to get us to spend more and their cookies were showing fantastic performance, but the big bosses at the parent company were questioning why the overall sales seemed to be unaffected by how much was spent on the platform. So we worked with Facebook to split the audience 50/50, half would be exposed to ads over a 4 week period and half would not see ads.

Basic testing logic then applies, the results were added up to see how many people in the exposed group made a purchase and how much was spent compared to those in the unexposed group.

A month later we got the results. They showed clearly that the people who saw ads did spend more with Topman. They made more purchases, they spent more money. However, not by much. The cost per incremental sale was staggering, the amount of money spent on ads over the 4-week period generated such a small uplift that they would have lost money on every single order many times over.

Cookies showed a ROAS of 5-8x. The uplift test showed a real ROAS of less than 0.2x.

This isn’t to bash Facebook ads, plenty of studies I’ve run have shown the opposite, some with even better results than cookies would suggest, but is just one example of how badly cookies can get it wrong. It also shows how time consuming it is to track properly

Another example just this week, much simpler. We sent a report to a client saying “great news, you’ve had a 40% uplift in leads this week”. The reply came back straight away, those extra 40% are all spam.

A Step Back To The Dark Ages Of Advertising?

The perception in the advertising industry is that we’ve spent the past 20 years benefitting from perfect data, and we’re about to step into the dark ages where we have no idea what’s going on.

My opinion is that we’ve never had perfect data, we never will, and it takes many years of experience, layer upon layer of understanding, and a healthy dose of statistical knowledge to be able to know when bad data is good enough to use. A far simpler future awaits us where everyone is aligned in knowing what we don’t know and being forced to compare and contrast multiple data sources, rather than blindly following such a flawed system.

There is work to be done. There are new systems to install, new methods to learn, training to change our mindsets after years of blindly following cookies. But no one should be worried, long term, we’re all going to be a lot better for it.

Get Started