How to use AI to drive targeted customer acquisition in 2020

Today, 80 percent of digital marketers feel growing pressure to meet customer acquisition and revenue goals, saying they feel like they are "running on a never-ending hamster wheel."

And that’s true: You may be overstretching your marketing muscle to get your customers to find you through organic search -- but it’s something you have little control over and is reactive. Traditional advertising media is becoming more obsolete, but the problem is that many businesses still don’t realize how technology can reverse that. Take the example of the automotive industry, where customer acquisition is still very reactive. The majority of customers still find the auto dealer and brand, and not the other way round. A recent study found that car dealers' first point of contact with more than half of their buyers is when they physically walk into the dealership, essentially leaving it to chance that their dealership or brand gets picked for a walk-in. There’s no doubt that companies across the board are rapidly experimenting with adopting artificial intelligence (AI) in various departments, including business performance and automating the human tasks, the low-hanging fruits. But it’s time you thought about bringing it in to enhance your customer acquisition strategy.


Shifting entire business models

Customer acquisition strategies themselves aren’t enough -- unless they are targeted and proactive. We all saw how the video rental industry was upended by the streaming business model. Just imagine going back to a video rental store today, if they were to exist, and being forced to survey the shelves of movies to figure out what to rent next. It is unimaginable in today’s world but it highlights the power of proactive and personalized delivery to the customer based on what the customer wants and when they want. The streaming companies use AI and so can the marketers to transform their marketing approach.

Take Netflix, for instance. After logging in, you immediately get personalized recommendations based on prior actions made on the site. Because Netflix knows your preferences, it’s able to predict and recommend movies or shows you’re likely to enjoy, allowing you to binge-watch interesting new programs on a daily basis. Netflix’s recommendation machine is so good that 75 percent of what users watch comes from these product recommendations.

Seeing these powerful results, it’s surprising that the established marketing model is typically reactive. It relies on the customers finding you. While the digital age has brought advances in terms of incentivization tools, companies are still one step behind. Customers need to take the first step to express interest or visit the site for digital marketing to be effective. Many businesses are still dependent on that stroke of luck.

AI can flip this model around. When powered with the right datasets, the marketing AI can accurately predict the likelihood of purchase, recommend the best moment to approach a customer, and help shape the exact offer that will get the most traction. In short, AI helps to answer the most pressing marketing questions of our times: Who, when, and how to target.

AI-driven targeted conquest

Digitally, it’s fairly simple once the customer has expressed interest in a product to recommend other products or to recommend new products based on their purchase history. Is it targeted conquest? Not really but it still is much further along than the traditional industries.

Let’s take the example of a traditional industry, your typical consumer electronic goods retailer. It spends marketing dollars on traditional media, weekly email and print flyers, and deals on selected products. These get blasted to the general public with a single aim -- that anyone with the purchasing power will pick it up for a walk-in. Were the same retailer to use AI-driven targeted conquest, they could reach out to one customer who is likely to buy a TV with multiple options and personalized offers based on their unique preferences while to another it may reach out with home theater offers and options. Your customers already expect this level of personalization and are waiting for you to catch up.

72 percent of consumers expect companies to understand their unique needs and expectations. So it’s safe to say that AI’s ability to offer targeted and proactive customer acquisition, beyond the traditional digital channels, is the single biggest opportunity that we see in marketing these days.

The powerful build-predict-recommend-learn cycle 

By deploying targeted customer acquisition, you can find the right customer for the right product at the right time and reach out to them with the right offer and messaging. To do that, you just need to follow these four steps.


Start by creating a classification for each product you offer. This means taking five to ten features that would describe it best. These aren’t necessarily the product characteristics as they would be described in marketing literature, but the features that influence the buying decision. After that, pull together historical customer lists for each product.

To get a 360-degree view of customers, you need to dig for external, publicly available data, including demographics, lifestyle, behavioral, and major life purchases data.

These two data sets then need to be stitched together and cleaned. Know that the latter is a very time-consuming task, but an important step for the ability to predict accurately. This way, you can create specific personas of potential customers for certain products and categories. 


Once you have that data ready, the next step is to build the prediction models using statistical, machine, and deep learning models. While your broader business objective is to predict target customers for specific products, this may consist of multiple mini-predictions that cover the trigger point, buying range, features that customers prefer, and other factors.

This step also incorporates selecting features for the prediction models. If you created a clean big-data universe in the previous step, you may have hundreds of attributes. However, only around 10-20 are actually required to feed the prediction models.


This stage is when you apply your prediction models for targeted customer acquisition. You can select various filters -- such as geographical location -- to identify the broad set of customers that they want to run their prediction models on.

By applying the target persona on actual customers in the area, you can identify and score targets based on the extent they match. Look at the account trigger points and product match attributes to find the exact right customers at the stage when they’re most probable to make the purchase.

Now, this is when your true marketing spirit kicks in. The last part of this stage is designing personalized messaging and offers for the customer and deciding on the best channels to engage them. Both of these elements also will come from your mini-predictions or attributes as defined previously


The learning and self-improving capabilities of AI are the technology’s key strengths. If you feel that Amazon and Google recommendations are becoming better and smarter, that’s because they are learning with each prediction cycle.

For targeted marketing AI, the learn stage needs to consist of two elements:

Self-learning cycle: In simplest terms, imagine showing the AI robot its conquest marketing report card, which consists of which customers it predicted right – meaning that they purchased from you. For every wrong prediction, the machine penalizes itself, and for every right prediction, it rewards itself. It learns from the results and uses this intelligence to improve further.

Supervised learning: As a marketer, you need to be continuously doing the same thing too -- see which results are coming better, which are not and what tuning is required.

The learning stage is critical and ensures that the next cycle is smarter than the last one, and that the buyer persona, trigger points, and account for changes in buyer’s behavior are continually reshaped.

Many companies already deploy the build-predict-recommend-learn cycle to great success. One great example is Spotify, with its Discover Weekly feature. After building their algorithm, Spotify is able to learn each user’s music preferences, predict what music is similar to the music the user likes, and recommend a unique playlist every week to each one of the hundreds of millions of subscribers. 

The key to harnessing AI? Overcoming challenges

To fully drive the potential of AI in targeted customer acquisition, you need to be aware of three key challenges: removing bias, completing buying persona, and moving to real-time predictions.

Removing bias

When it comes to dealing with any predilections in automation, be alert. Bias is inherently a real concern, so it may come as a surprise that not all bias is necessarily bad for targeted conquest. That’s because companies may see brand bias in their models, but this may be a real bias required for the accuracy of those models.

The only way to eliminate the unwanted bias is through constant supervision of the algorithms you deploy. This shouldn’t be the sole responsibility of the dedicated data scientist -- but of a team of professionals, including the business experts. The AI output and market feedback should constantly be monitored by the best marketing brains you have on board.

Completing the buying persona

Completing the buying persona in a comprehensive way can be a hard nut to crack. Many companies lag behind in adopting AI in their business decisions because they simply don’t know how to complete data and build the buying persona.

The typical business has customer name, address, and phone number -- but that’s insufficient to carry out predictions and recommendations. That’s why these pieces of information must be brought together and gathered over time to tell a more accurate story of each customer. Don’t hesitate to bring in external data to combine them with your internal datasets. But also be ready to spend significant time in cleaning out the noise in the data

Real-time predictions

Striving for real-time predictions should be your ultimate goal. Analytics and efforts of looking back and reassessing are already commonplace -- and represents only the springboard. The power of AI really comes in predicting how to engage and acquire the customer at the right time when the interaction is happening or needed, rather than analyzing insights from a historical perspective.

A good example of this is seen in the banking industry, where banks have started to use machine learning models to help with their real-time fraud detection. Using a complex set of algorithms to analyze each specific user’s behavior, any suspicious activity is immediately flagged by the machine, sending out an alert to the user to make sure that they are aware of potential fraud on their account.

The early days of AI are long gone. In this decade, it’s time companies stop regarding the technology as a trendy buzzword to prove that they’re "innovating" and instead focus on harnessing its real potential. With its capabilities in targeted customer acquisition, AI can empower vendors traditionally paralyzed by dependency on customers’ proactivity and change entire marketing strategies -- whether they’re in real estate, car industry, interior design, electronics, or elsewhere.

Photo credit: Peshkova / Shutterstock

Vikrant Pathak, is the CEO and chief data scientist for myautoIQ, an AI-powered auto customer engagement platform. Vikrant is passionate about using machine learning and data science for common business decision scenarios & creating real-time augmented intelligence capability, embedded in the business workflow. Before myautoIQ, Vikrant spent 20+ years in the industry, partnering with clients in Automotive, Information Services & Media, Marketing & Loyalty, and Financial Services domains to accelerate digital transformation, monetize data, and create disruptive capabilities.  

Comments are closed.

© 1998-2022 BetaNews, Inc. All Rights Reserved. Privacy Policy - Cookie Policy.