For obvious reasons, most of us prefer facts we can understand and trust to be valid. The challenge, in our technological times, is that the facts with the greatest validity are often hard to understand, and the most understandable facts are often wrong.

Many of the algorithms involved in processes such as medical diagnosis, self-driving cars and financial fraud detection are more accurate than any individual doctor, driver or auditor. We know this from measuring their success rate. However, that leaves us having to trust things we can’t understand, in preference to things we can.

At its heart, I think the inertia this desire for proven success rates brings to human development is often good. Even if there were algorithms available that might out-perform me in picking a partner or deciding what to eat for dinner to maximise my happiness, I would still like some good proof of their success before putting them in charge; just as I would like to see evidence of fewer accidents per kilometre before getting into a self-driving car instead of one driven by a human.

In the brand measurement business, we find ourselves at a tipping point. We have spent years developing sophisticated algorithms that identify the effects of advertising. And by this we don’t mean how ads trigger instant behaviour and clicks - since that is rarely how marketing works - but rather how advertising changes awareness, attitudes and intentions in the minds of those who take part of it.

Such technology - or our own, at least - is continually being upgraded to make it increasingly accurate in matters such as campaign exposures and their statistical relationship to brand perception, factoring in benchmarks and comparability. Further accuracy, of course, is great, but it also brings further complexity, which can be challenging.

Or it used to be. We have spent as many years arguing for the superior accuracy of the algorithmic approach as we have developing it. This was never an easy task. But as we have sat down lately with some of the world’s biggest spenders in advertising and some of the world’s biggest publishers to explain our recent upgrades, we find we are not challenged on validity - we no longer have to argue.

Why? Have we finally learned how to boil down the complexity of what we do into something that feels easy to grasp? Has our storytelling improved? No, this doesn’t come down to pedagogy.

The fact that traditional approaches have problems has long been as obvious to the market as the fact that human car drivers sometimes cause accidents. But that an algorithmic approach is the better choice and deserves to be trusted accordingly, despite it being harder to explain and understand, comes from one thing, and one thing only: its proven success rate.

Reporting the effects of tens of thousands of campaigns for leading publishers on all continents, with the world’s biggest brands at the receiving end, has added up to an evidence-based industry consensus that this is the better way. The proof of the pudding is in the eating.

It’s with relief we finally see marketers, agencies and media owners coming together to work on privacy-secure, first party data-informed advertising.

And it’s with gratitude that we note that algorithmic metrics like ours are becoming the measures of success in doing so.

The algorithmic approach now has enough usage and application for the industry to know that it is more valid than pre-existing approaches. To pick up the self-driving car analogy, we now have the mileage-without-accidents data that we need.

The day we stop working hard to improve is the day we too find ourselves being defended as the old best practice, until the market realises that a more advanced way is more successful. The better approach will always conquer when it has the success rate to prove itself.

And right now, speaking for ourselves, we can say that brand lift evaluation algorithms are simply the best. They’re just not simple. But the best rarely is.

Also published in: AOP

Podcasts powered by AudioHarvest