“Brand Aware” explores the data-driven digital ad ecosystem from the marketer’s point of view.
Today’s column is written by Sachin Puri, vice president of growth marketing at McAfee.
Not a day passes in the marketing world without someone name-dropping machine learning and artificial intelligence in a meeting. The underlying myth is that machine learning is the ultimate solution to all things optimization, and eventually artificial intelligence will enable machines to take over all marketing jobs.
Would we reach the level of marketing shown in “Minority Report” or the power of machines in “I Am Mother?” Well, that’s a longer evolution and involves multiple elements, such as ethics, emotions, technology and science.
In the short term, marketers will continue to leverage machine learning (ML) to deliver value to the customers in a very meaningful way.
The intersection of machines and humans
I love machines and can vouch that machine-driven algos are great in pattern recognition and finding anomalies at scale.
For example, machines can identify unique purchase patterns and accordingly make better bidding decisions based on hundreds of signals available in RTB. With the advancement of underlying software and hardware technologies and the evolving hackathon culture, the current tech-savvy marketing talent is automating many tasks and decisions that were once made by humans.
Marketing is attracting a ton of data scientists who continue to develop smarter algorithms to make better and faster predictive decisions than previous versions. So, the adoption of machines in marketing is inevitable, and machines are here to stay.
Having said that, there is no replacement for human creativity; in geek speak, the human brain is one of the most complex and enhanced machine learning platforms.
In my opinion, one of marketers’ main jobs is to identify the human truth that drives connections with products and services: the joy of seeing LeBron James and Steph Curry do what they do best, a racing heart just before Messi kicks a goal, the excitement to find an Amazon delivery box on your doorstep or the touch of your recently delivered laser-etched titanium Apple card. You get the drift.
These emotions and associated intelligence aren’t captured optimally by machines as yet and are proprietary to humans; that’s why most of the creative ideas around content still come from humans. Yes, machines can optimize and personalize the final delivery to a specific audience better than humans can. Thus it’s an AND vs. OR when it comes to machines vs. humans.
As the old cliche goes, marketing is truly an art and science; wherein humans and machines will continue to evolve and be more powerful together.
Key areas to boost marketing by machine learning
Audience targeting: The ability to target the right audience at the right time and in the right context is key to delivering the promised return on investment for marketing dollars. One such approach is to first leverage first-party data to build microsegments and associated purchase likelihood score based on machine learning models. Next, marketers can use available look-alike models offered by almost all activation platforms; some demand-side platforms (DSPs) even allow onboarding proprietary models. And last, marketers can capture and connect the performance data to associated systems for ongoing learning at scale. This self-learning is one of the core tenets of machine learning and crucial to improving the accuracy of machine learning models.
Media bidding and optimization: For each microsegment, both cost per acquisition (CPA) and lifetime value (LTV) need to be measured and optimized in real time to ensure optimal campaign results. In general, both LTV and CPA vary by the audience, time and platform; thus the decision about how much to bid or even whether to offer a promotion quickly becomes complex, requiring machines to self-learn and automate the process.
Many advertisers have developed in-house bidding models as their secret sauce to maintain their competitive advantage. However, the general trend, especially with the rising requirements for user privacy, is for platforms to share less data with marketers, limiting the impact of in-house bidding models.
For example, Google has stopped sharing position data for paid search, and the life of a display ad cookie continues to shrink. The exchange level bidding by DSPs (cookie vs. segment), search engines (query vs. keyword) and social platforms (user vs. ad set) always had more optimization levers than in-house bidding. Given the current trends, in the future marketers will likely have to settle for BYOD (bring your own data) solutions vs. BYOA (bring your own algorithm) for bidding. However, marketing-mix management would likely stay in house and also likely adopt machine learning.
Dynamic creative: In the world of digital where consumers are multitasking and switching between apps and sites with fast finger swipes, it’s important for marketers to deliver a very impactful and relevant creative for each micro-attention opportunity. For example, as a Michigan fan who loves sports shoes, my likelihood to engage and learn (discovery) about Nike’s Michigan-themed Air Zoom Pegasus 36 shoes while quickly browsing Instagram is quite high.
Dynamic creative, the ability to develop and test different combinations of creatives and copies in real time, is a perfect example of where human creativity and machines complement each other. Essentially, marketers can build a library of images, headlines, messaging and promos, while machines via ML can run tons of multivariate tests to optimize creative performance by creating a unique personalized ad with varying image, copy, headlines, color and call to action.
Many platforms, especially social and search, already offer solutions to run dynamic ads at scale, including Facebook’s Dynamic Creative Ads and Google’s Responsive Search Ads. In addition, dynamic creative optimization and creative management platforms are merging into a unified platform which seems to offer end-to-end solutions to advance productivity via iterative creative generation, translation, formatting, optimization, testing and reporting. This will continue to unlock the power of dynamic creative for display advertising. In my opinion, AI-based natural language processing is already powering chatbots, and it’s only a matter of time before marketers also adopt machine-driven dynamic copy generation as well.
Landing page optimization: The landing page, where the user lands after clicking an ad, email or app notification, is an opportunity to ensure users move down the campaign funnel. This requires presenting the right product, pricing, offer, hero image, copy, color, etc. based on the key attributes known about the user. Again, human creativity allows for an inspiring page design and concept, but machines can likely do a better job of personalizing the page at scale. Most websites and apps are usually modularised, and some leverage machine learning models to optimize performance of overall experience via APIs or other web services.
This is truly where marketers’ first-party data can supercharge the user experience by delivering a highly personalized page. One simple approach is to create different experiences for new vs. repeat customers, then developing ML models for better predictive cross-selling, upselling and promotional offers based on past data of the same or similar customers. There are tools, such as Adobe Target, and techniques, including URL parameter passing and shared audiences, that are readily available to provide a jump start to marketers looking for solutions in this space.
Next steps and suggestions
There are clearly tasks that are better performed by machines, while some tasks have no replacement for human creativity, and the rest are subjective to each company’s business and cultural nuances.
Fully unlocking the power of machines and human creativity requires a cultural and strategic shift, such as developing a holistic view of tech and data, being comfortable with experiments and innovating to build a growth mindset in talent. A practical first step is for marketers to engage in a workshop with cross-functional stakeholders, especially data, creative and tech, to develop a unified strategy and stack.
Another key outcome should include a clear agreement on a build, buy or partner plan and associated 360-degree talent assessment to determine which technology and talent to build in house vs. what to leverage from outside.
In my opinion, talent decisions have been one of the most important differentiators in marketing, and it’s even more important now in this ever-evolving world of data, creativity and machines.