AI In Marketing: Where And When It Can Make A Difference

"Data-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Dana Hayes, president at ShareThis.

Today’s CMO is tasked with the challenge of understanding a far greater number of channels, platforms and technologies than ever before. Couple that with the never-ending flow of data coming from every device, method and channel and it’s a recipe for data-processing disaster. The right investment can determine whether a CMO lasts less or more than the average 18-month lifetime.

Artificial intelligence offers fascinating possibilities for marketing. While it’s still in its infancy, the power is in the hands of marketers to push for answers to the hard questions. Marketers looking to invest in new technologies must know how and why they’re going to apply them and evaluate how they will solve specific pain points. By working with teams made up of traditional marketers, who focus on the practical applications or technical investment, and more technically savvy computer scientists, who will be responsible for building out and deploying new solutions, CMOs can make far more informed decisions.

For AI or any new technology to make an impact, it should be able to solve a key pain point. I see a few particular areas where AI can improve performance and reach goals.

Productivity and fraud

As marketers’ functions expand, AI can power better decision-making and help solve the data overload problem that most marketers face. Data’s depth and speed are the biggest issues. Adding a digital brain to the problem helps identify perfect storms that will free strategic teams from the basics of just corralling the insights.

Ad fraud is also an opportunity to apply machine learning to weed out fraudsters reducing the time spent manually building whitelists and blacklists. While AI-powered protection might be helpful for the most common patterns, it is still not a replacement for diligence.

Machine learning can help flag suspicious IP activity. Flagging things such as time spent, scrolling and mouse movements can help identify and train AI to scan for similar patterns. While certain types of bot traffic are predictable and can be identified easily, others are adopting more sophisticated fraud tactics, which is where AI-powered fraud tools can be useful for mitigation.

But marketers also need the right data and truth set so they can trust the accuracy of the AI tool they use. Just because a powerful tool is available doesn't mean it's always the best choice to solve a particular problem.

Reaching the right customers

Many programmatic platforms apply AI to the decision-making process for which impressions should or should not be bid on. AI uncovers insights in the massive data streams in the digital ad-buying process that can be acted upon in real time. Even for assessing and determining audience data, AI can help go beyond what humans can do. With most big data sets, scale and speed can deliver new insights. In the end, AI serves as a tool to keep up with the dizzying array of resources that marketers can leverage.

For instance, marketers can reach the right customer who will convert the fastest and not waste media spend on the masses. Non-relational features and data points are perfect fodder for marketers to create custom marketing experiences on the fly. AI allows marketers to go beyond off-the-shelf data and leverage first-party data with third-party data sources. AI can pinpoint individual user features to leverage all forms of acquisition campaigns.

Alternatively, marketers can expand their pool of target customers by leveraging high-quality data for lookalike modeling. While lookalike modeling has been around for some time, a deeper level of AI coupled with a rich third-party data set refreshed in real time can help expand and select additional prospects. I believe it is far more accurate to call it “act-alike” models, if the power of AI is truly being applied. High-quality external data, such as weather patterns, stock performance or sports team scores, offer opportunities to develop effective, personalized campaigns.

To determine when a marketer might want to leverage AI-powered data, marketers must start with their KPIs and what they know about their customers. Crowded categories, impulse buys and researched purchases will benefit greatly from AI-powered insights. Table stakes in today’s data-enriched world, audience expansion requires at least a baseline level of machine learning.

Marketers need to find prospects that do more than just look like their best customers – they must also behave like them, and that requires multiple touch points and more powerful AI. The depth of connection points in a lookalike model is a great place to start when assessing different solutions.

Improving the customer experience

One of marketing’s primary functions is tied to the customer experience and ensuring that customer interactions are positive. The need to surprise and delight customers will never go away, and AI can help anticipate the needs and wants of customers by learning from, and acting on, past or parallel behavior.

With this deeper knowledge, practical AI applications, such as chatbots, voice recognition and automated translation, enable marketers to create more relevant experiences for customers. Microsoft, for example, launched Skype Translator to give users a near-real-time translation of languages. Applied to a marketing function, chatbots can provide a human-like connection in any language and answer common questions about a recently released report or a customer service issue, or even assist with writing blog content.

At the most basic level, if a brand has a product that leads to questions or might require live engagement, investment in an AI-powered chatbot is a must. If this is the first time they’re dipping their toes into the AI waters, the road might be smoother by partnering to trial and launch key features that provide more of a concierge to basic information about the product.

By their nature, chatbots are steeped in AI, and each player offering services will vary in its capability, but the key is the execution. What is missing in most cases is collaborative problem-solving on how best to deploy chatbots. It starts with determining the key channels and functions that could be automated. For instance, Airbnb or Spotify leverage Twitter to answer customer service issues and elevate areas of concern, while Kayak uses Facebook messenger to find flights and schedules.

Chatbots also offer an opportunity to differentiate or even drive transactions. Dominos allows customers to order via Facebook Messenger, and consumers can share funds on PayPal via Slack. While transactions require more integration tools, it all starts with that creative, collaborative execution between the teams that know the customers best and the teams that create the bridge for the execution.

As with most work, especially those involving data, the larger the sample size, the stronger the insights gained. To be done effectively, machine learning requires very specific questions, massive amounts of data and time to solve a given problem. If marketers miss that step, they risk achieving what some might call a black-box effect, where there is little understanding of how insights were determined.

And, finally, marketers need to remember that analytics, actionable insights and meeting business goals are the best measures of success, far more than the use of the latest bells and whistles. AI holds tremendous potential, but it won’t be of immediate importance to everyone. It won’t fix a bad data set or bidder, nor will it save flawed infrastructure.

It will, however, speed up processes and make connections faster and with more precision. If marketers are considering making the investment, they should take a step back and focus on how AI will affect their ability to better understand and engage with their customers, which will, in turn, move the needle.

Follow ShareThis (@ShareThis) and AdExchanger (@adexchanger) on Twitter.

 

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>