The human touch is the real fault in programmatic
Machines don't benefit from hitting their targets. This, as Bohemia Group's Victor Condogeorges explains, is a human bias.
Programmatic has its fair share of technological challenges, but the biggest, potentially unsurmountable challenge is human bias.
Aside from the short term benefits of “efficiency” and “achieving outcomes”, I see the real watch-out in the programmatic game is people and their motivational incentives.
On one side, there are the well covered challenges of fraud, dodgy/unclear supply chains, and sometimes even malice, but lurking in the background are the challenges of human bias. This is a combination of unclear goals, a lack of long-term direction, and the borderline obsessive focus on immediately attributable outcomes. These crappy instructions and measurement methodologies then get fed into the system and guide the machine doing the work.
I find the rules and definitions of success are often scantily thought of and as a result, gaps are left for bad actors to hide and take advantage, either intentionally or for not. The process it seems has been “shit in, shit out” for a long time – and with no qualms raised so long as it backs out to a CPA that is within the target.
Bugs in the bonnet – Even self driving cars have flaws
Are you giving your programmatic platform the best opportunity for success by providing it with all the available data?
If you were to hop into a self driving car and tell it to take you from location A to location B, it would do so taking every signal into consideration: how fast the traffic is going, the surrounding obstacles, the lines on the road, the speed limit, the braking distance, the list goes on. Because of this, loopholes can be found in the programming because it doesn’t think. It simply executes.
This principle is the reason why they can be trapped in a circle so easily.
When you think about it, the signals in the programmatic arena haven’t been entirely mapped out. There are many tangents that can send your campaign in a somewhat myopic direction to achieve that single target you set, neglecting all the other signals that have other impacts on customers. The machine is literally making decisions on only part of the information. Does this sound like a good idea? Do you have confidence in your decisions when you only know part of the information?
A more confronting thought is that while you may have achieved the best CPA, perhaps those conversions were going to happen anyway based on your micro targeting.
Congratulations, you achieved your quarterly target – and stole equity from the future in the process.
I don’t think a machine would have this challenge, unless their targets were set by humans. Machines don’t benefit from hitting their targets. They don’t get a bonus of more wattage or amps to buy themselves a higher shelf in the server rack. This is a human bias.
Funnel cutting tactics – CPA hunting and targeting have side effects
Are individual KPIs set up for each segment or are short term outcomes the biggest focus?
ROI is unequivocally the name of the game however focusing solely on cost efficiency in CPA is biased, thanks to a malignant measurement method.
Squeezing CPA targets can lead you down a path that eventually creates a vacuum for your brand by only talking to the “highly likely to become customers” audience. Sounds like the dream, right? Only talk to those that will buy? Minimise wastage? Let’s take a moment to pop this idea on its head.
Imagine you’re planning a programmatic campaign. You plan to add data targeting to find an audience that is most likely to complete the action required to achieve your CPA target whether it be with first, second, or third party data. This does in fact reduce your wastage on advertising to people that aren’t identified in those audience segments. Sounds like a good idea, right?
To illustrate this visually, this is basically what happens when you add targeting to programmatic, and the targeting is dictated by cost efficiency rather than growth.
One of the problems with a micro targeted audience is that you may achieve your target CPA, but you’ve cut out a large portion of the market not in that audience. Those people may be interested in your product too but you’ve excluded them entirely from the targeting for the sake of “efficiency”.
You’ve completely overlooked the people that would consider your product, simply to focus on people that were already past consideration, all to get the best CPA.
This is one of the many attribution conundrums, and why search continuously seems to top the cost efficiency list in last-click attribution models. But it’s 2018 and we all know this isn’t the way to compare CPA efficiency, right?
Now, considering this “setting and achieving of targets to unlock a bonus” is a problem for humans, how do you expect a machine to work with it and achieve the best result?
Humans are still involved in writing the brief & setting the targets, then adjusting the settings within programmatic platforms. Success is critically dependant upon having as much information as possible and above all knowing what you’re doing, why you’re doing it. It is not dependant on the machines capability.
Writing the brief and understanding customers is the area where effort needs to be invested, that’s where we still retain value over the machines. Once they can do this, you will no longer be needed in the programmatic loop. Remember, the machine just executes, it does not think. Yet.
In summary
I’ll leave you with some challenging questions and even more jarring statements.
- Are you giving your programmatic platform the best opportunity for success by providing it with all the available data?
- Are individual KPIs set up for each segment or are short term outcomes the biggest focus?
If you were replaced by a machine that wasn’t incentivised by quarterly targets, would it be making the same decisions you are today?
NB: If you think you would do it better than a machine, you have probably overlooked something in the brief or may be biased by personal gain.
I think we can all agree there is more value to be had from advertising than quarterly targets, hoops to jump through, and boxes to tick. There is a component we need to consider about investing for the longer term. It may be longer than our tenure but that doesn’t mean we should neglect it as brands often last longer than our time working on them.
If you steal brand equity from tomorrow without replenishing it, you’ll probably end up having to face the consequences if you intend on staying with that brand in the mid to long term. There’s only so much of the hair of the dog you can have before you crash and burn.
To put it bluntly, if you’re solely focused on short-term targets, you have revealed your motivations; that you consider your short term targets more important than the long-term success and future of the brand.
Victor Condogeorges is digital investment director at Bohemia Group. This article first appeared here.
Great article
User ID not verified.