Is AI-based ad management really superior to manual optimization? And what does that mean for us as marketers? These are the questions Samuel Andratschke, Director of Social Media Advertising at CROSSMEDIA Düsseldorf, is asking.
Recently, I attended a digital social commerce summit in Berlin. Innovations, trends and exciting future visions were presented and discussed. At the end of a day brimming with input, I sorted through what I had heard and found that, generally speaking, there had been a great deal of agreement on the outlooks and best practices of the respective major platforms in social media advertising. Even more so, these best practices apply equally to other digital genres. Malicious voices might refer to this as a strategic one-size-fits-all approach, despite the enormously fragmented digital advertising landscape that is the current status quo.
Everything was better in the good old days – but was it?
“Everything was better back then” is a phrase we often hear. There can be no doubt that a certain romantic yearning is inherent in us humans, at least it certainly always plays a role in hindsight observations. Yet, doesn’t every proverb hold a truth at its core? Therefore, the real question is whether the above also applies to digital advertising, or whether an objective view will declare recent years’ progressive developments to be the better age.
I started in digital advertising back when campaigns were still set up manually, in an extremely granular way. Target groups were separated according to gender, region and various age clusters, and these were set up in combination with the target group’s appropriate interests. The rule used to be: the more line items a campaign had, the better it could be controlled and optimized. Bids were monitored manually, daily and scaled up or down correspondingly, so as to ensure the best possible performance. With the emerging hype of real time bidding campaigning in the early 2010s, early indicators already emerged of what was technically possible, and how and where the digital market landscape was headed.
Today, nearly 10 years later, digital strategies on platforms are completely divergent from the approach outlined above. Over time, the tools’ algorithms have become increasingly more capable of distinguishing and clustering user interests, and then applying the insights gained about the users in the targeting of the campaigns. This always requires a large number of signals that users leave behind with their digital footprint – either on the open web, or in-app in the walled gardens. The following can be stated: Unlike the previous approach of working with as many and granular line items as possible, it is now as many and detailed signals as possible in user behavior that will enable campaign activation, together with functioning machine learning. To guarantee the systems the most flexibility, it is recommended to ideally set up all target groups and targeting within a single campaign setup. Thus, manual control over performance is taken out of the equation to a large extent. Bluntly stated, the optimization possibilities are limited to scaling the daily media budgets for the specified target achievement up and down.
Algorithms calculate, humans consult
You may now be asking yourself whether this is better or worse? One benefit that cannot be denied is indeed the time saved in setting up campaigns. Likewise, the algorithms optimize the set specifications tirelessly and 24/7, also during holidays; sources of human error are minimized, and sick days do not exist. However: an algorithm can only ever work well if it is dictated a clear set of objectives that correspond to its specifications. Unlike the account manager, it is not able to interpret customer requests in a versatile way and respond to dedicated tasks that take more KPIs into consideration. In theory, clients should ideally commit to a single, overarching goal; in practice, though, they still often ask for the all-in-one solution we all know only too well. The task of a client consultant or agency would seem obvious: show the relevant customers what tools and systems can and cannot do. However, this sounds easier than it is – because there are usually several stakeholders who decide on marketing budgets and who, at least in part, have a different focus on objectives. Based on my experience, the number of briefings with at least one additional KPI exceeds those with a single and strictly-formulated objective. What is more, if business KPIs are not exclusively broken down to available data segments of the systems, machine learning will reach its limit just as quickly.
A shift in the task profile
So, how do we deal with this, and how does this setup change the role of the account manager? And, are we in danger of being replaced by the machine – as is the case, for instance, in a number of skilled trades? It is important to understand that direct campaigning in the respective platforms will become simpler and will probably be reduced even further in the future. The job profile of online marketers will inevitably increasingly shift toward an advisory role and away from direct machine-room expertise. Marketers should think more strategically and analytically – and not just at senior level. The time gained through the use of AI and smart bidding can be put to good use in daily business. Where account managers used to spend several hours a day actively tweaking the parameters to optimize campaigns, they now need to observe and analyze performance and target achievement in detail and derive appropriate cross-media recommendations for action. In this regard, the challenge is to understand to what extent one or more objectives can be achieved completely via fully-automated optimization of the systems, or where certain compromises are necessary. As such, the development of adaptive and custom-fit solutions will become even more of a core task in digital advertising communication.
Utilizing benefits and remaining open to change
Where possible, the full power of smart bidding and machine learning can and should be utilized. Clear briefings and unambiguous KPIs that are used for measuring are the key prerequisite here. However, not everything is always unambiguously black and white, and we often find ourselves moving in gray areas. In these areas, using experience and a good understanding of how the systems work is important for us, so as to derive the best possible adaptation of a strategy. Refusing AI as a general principle, and working according to the motto of “I’d rather do all the optimization myself” is not helpful. We are experiencing rapid developments in all areas of life. Yet, we are seeing them especially where manual processes are being replaced or supplemented by machine learning. The whole thing will gain even more momentum in the future and cannot be stopped. This makes it all the more important to accept this change in a non-judgmental way, because trying to fight it and running the risk of being considered “yesterday’s news” will help neither our customers, nor us.
What Do You Think?