A marketer’s toolkit should include predictive artificial intelligence tools that are constantly learning and being refined, so that customer intent is properly captured. Intent marketing, when done well, means that the marketing is done at the right time, right place, with the right content and the right offer.

Challenges to AI for Intent Marketing

While machine learning has been around for more than fifty years, there are only recent inroads being in the space of AI for intent marketing. There are a few reasons for this:

  • First, AI was harnessed in the eyes of a few. For quite a while, you needed to be a data scientist or required a team of data scientists to do the heavy lifting and work with the platforms that supported these efforts. Now, AI is out of the hands of data scientists and is more accessible to others too. 
  • Second, the old version of AI that has been around forever would work around batch learning without iterating upon its outputs. It would take an input of a huge dataset, such as an Excel spreadsheet, and process some outputs or provide an algorithm but never change later. The processing would happen offline in a static environment with a machine, a layer of software, over a slow period of time, and a “predictive” output that would be stale as soon as it was printed out.

The bottom line: it was out of date. These systems provided outputs, but an important variable was missing: the self-learning element, which was the inability to identify variations and required adaptations based on behaviors. 

We are headed into a territory where there is rapidly-changing predictive data, instantaneous real-time decision making as opposed to the older way, which was a days/weeks/months-long process that provided data points and outputs that weren’t changing.

A real-time solution was sorely needed. 

How Can AI Solve for Wasted Impressions?

In a personal example, I bought boots for my wife before the holidays on Black Friday. I was then retargeted before Christmas, after Christmas, and after New Year’s. What these systems did not understand is that:

  1. I had already made the purchase, which could have been driven by cookies or credit card data
  2. I’m not a woman
  3. The seasonality component was broken

This translates to wasted impressions. Think about it:

  • How many times have you made a purchase, only to still see ads that are irrelevant to you since you’re no longer browsing or surfing for the very product you’ve already bought? 
  • How many of you are selling products on sites like eBay and see your own listings staring back at you in remarketing campaigns? 
  • How many of you are car enthusiasts looking for a new Toyota simply because you’ve viewed a car article six months ago?

For local businesses, online learning, then, is needed. The correct AI analysis will need to look at the variety of data points that are going to a profile, based on the constant source of information across a business’s digital profile across channels (Google, Yelp, Facebook, etc.) as well as product listings and inventory, to see what combination of events are people taking that are most predictive of in-search traffic or click-to-call for retailers or service providers. 

A steady stream of new data must always be evaluated and optimized which will then help with refining outputs to give marketers the ability to more easily address customer intent.

What are the Nuances of Batch vs. Adaptive Learning?

If you take a child in New York City who can eyeball storefronts, restaurants, signs, lights, coffeeshops, and various patterns, the same child in Paris who has never seen Paris before will still understand what a stop sign is, the lights, the cafes, the stores, and the gas stations. But AI doesn’t work that way. If you give AI a map of NYC, AI does not have the ability to replicate its knowledge base in Paris.

Batch learning–that is, the NYC map–is finite and limited. Adaptive learning evolves that data and is truly adaptive to new data sets. It will automatically recognize something new and then categorize and incorporate new versions to the model. It will also publish and optimize against this model.

This means that everything is customized and personalized. As you see on larger media publishers, adaptation occurs based on previous and current behavior. A different homepage or newsletter appears on the basis of how users interact with this data. This is the same concept and is super important from a marketing perspective. This unique digital identity caters directly to the individual.

Bottom Line

Showing users data that’s relevant to their activity and making this decisioning come to life is where we see this bringing us. This moves into spheres such as voice search.

Alexa and Google Assistant will incorporate this data and provide relevant outputs. And it’s a win-win for the customer as well, as I’d rather get a personalized experience than some dumb experience that is completely irrelevant to me. 

Fortunately, there is a lot of software out there that you can leverage to be more predictive over purchases or interactions, which means that your targeting is more accurate, your messaging is more accurate, and you waste less money by not showing the wrong person the wrong information.

Since a consumer mindset is driven by instant gratification, it’s important to focus on smart advertising through adaptive learning.

Related Posts