"There’s an inherent drive for advertisers to get their hands on more first-party data"

In an exclusive interaction with Adgully, Tanmay Mohanty, CEO - Media Services, Publicis Groupe India, shares his view on how currently the platforms/ marketers operate and in his opinion how marketers should think of reach and performance planning. He also speaks on how the dependency on walled gardens will reduce going forward as marketers are looking for audience led planning.

Algorithms are difficult to define and the processes are complex. Could you demystify Algorithms a bit, especially in the context of Search and Social Media?

There are two definite ways to look at algorithms from a marketing point of view:

Machine Learning (ML) – This is where the platforms look at building algorithmic processes for delivering the content you prefer. This is based on your behaviour in the past to predict the future.

Artificial Intelligence (AI) – This is where the platforms use algorithm data across multiple data points to make decisions on what communication to display to you, not based on your past.

In marketing, the context reference – whether it is look-alike, bidding optimisation, propensity modelling or recommendation engine to cross-sell and upsell – we use a mix of machine learning and artificial intelligence to deliver the best results.

With the growing use of automated decision systems such as machine learning and artificial intelligence, how should marketers think of reach and performance planning?

As marketers, we already use AI and ML to deliver personalised marketing. McKinsey Global Institute found that their use in performance drives additional sales and an improvement of 10-12 % in forecasting. The use of algorithms to drive fast insights and the best action is driven mostly in top of the funnel marketing.

How will the dependency on walled gardens reduce going forward as marketers are looking for audience-led planning?

The trepidation is that algorithms also bring biases and noise to the decision. One of the biggest concerns still stays in data harvesting and exploitation by the platforms. Automation at scale in marketing also brings in discussion to socio-economic impact of these inherent biases and the impact it has on our culture as whole.

This is less about pushing away from walled garden dependency, more about exploring and finding opportunities to bring owned data and brand goals into the mix and using the information brands have on customers to build custom bidding opportunities. With signal deprecation, this has amplified the cause. Not only is there an inherent drive for advertisers to get their hands on more first-party data in order to continue in a world of addressability, but there is also the new wave of algo-based opportunities aimed at navigating the world of unauthenticated web.

• Auditing both data & partners is important and a key reason why we leverage our proprietary Verified Offering to vet partners. Algorithmic tools are only as good as the data fueling them. Knowing data sources, audit measures, recency, etc., is important.
• Opening the proverbial “black box” of AI pushing transparency – Ask how it works - Ask for the measures being taken to eliminate bias.
• Ongoing conversations with trusted partners to understand what technology standards are in place to mitigate bias in the tools that they offer.

How can strategic outcomes be improved with advanced analytics? How can it reduce bias in decision making?

Content marketers are now struggling with the bias that the platforms bring into the mix. Some use cases:

Auto-bidding delivers on a similar audience and maximises the outcome given the complexity; marketers, however, do not see much improvement in the lifetime value of the customers. Also, audience and insight scoping reduces the scope for growth and innovations and assumes certain oneness and linearity to behaviour and we know humans are not linear nor do they follow the same patterns. So, on the one hand there is personalisation at scale, and on the other, there is overall reduction in the uniqueness of personalities.

In the case of, say lookalike audiences, the marketers’ ability to drive reach and penetration for new customers might get limited by these biases. The decision of efficiency forces these decisions to deliver on the benchmarks and can impact the future growth of the brands. Algorithms and AI/ML are the things powering the campaign cycle in a future state – the focus to date has been thinking about the resurgence in the real-time bidding capacity. But we need to also think about impact to measurement and changing the way we think about mixed media models.

AI can help in both the way we plan, as well as how we improve our processes, by extracting different levels of insights from static data sets, then applying AI and using in real-time bidding.

However, while AI can help with performance and even reducing biases, it can also bring and scale biases at the same time.

IT Minister Rajeev Chandrasekhar had addressed the increasing trend about social media platforms using biased algorithms in Rajya Sabha. How do the platforms/ marketers currently operate?

Well, the concern persists in four key metrics for most marketers. Whether the algorithms create and support societal biases which leads to both data and algorithm bias. Therein, leading to the decision preconception that we see.

Here, privacy, accountability and fairness are the core goals to operate with. That will need the platforms, government and stakeholders to work together to drive transparency and responsible marketing for better future of the core human values.