Target Customers with Clairvoyance: Predictive Marketing Can Tell You When They’re Ready to Buy

What would you do if you knew what was going to happen tomorrow? Or next week?

What if you could know what the people you interact with were going to do before they did it? Would you change your course based on that knowledge, perhaps to make estimating the click-through rate for new ads much easier?

Of course you would!

It’s been a long time since marketing put the best signs in the window, to attract customers walking past. Putting an emphasis on your business’s online presence will increase your potential customer base exponentially. The flip side, though, is that online consumers have a much easier time switching from one retailer to another. Knowing what happens next could give you the edge you need.

Knowing the future is, of course, impossible. But predicting it based on research on trends? That’s what good marketers have been doing for years. Until recently, it’s been based on the gut feelings of marketing executives. Now, things are a bit different.

Using consumer analytics data, a bit of machine logic, and a bit of technical know-how, marketers are now able to use predictive marketing to “know the future” of their customer interactions.

What is Predictive Marketing?

It may sound a bit like being psychic, but there is a pretty significant difference between being psychic and using predictive marketing. Psychics use fast talking to lead the subject to a place they may not have wanted to go. Predictive marketing uses data, like psychographic marketing, to determine where people want to go, so you can meet them there.

We need to go deeper…

Predictive marketing requires a comprehensive set of customer data that will be used to build models which can predict the future behavior of these customers. These models, paired with the right machine logic, can identify customers who are likely to switch and can then target them with deals (that may entice them to stay). These models can also identify likely customer issues so that you can either address them proactively or be ready when the customer complains.

In order to maximize the effectiveness of this practice, we need to understand our relationship with our customers. It’s not a linear journey – it’s a cycle.

The Customer <> Business Life Cycle

The days of blind brand loyalty are fading fast. We can no longer assume that selling to a customer once guarantees repeat business. Rather, we need to maintain an awareness of the customer throughout the lifecycle.

The customer business life cycle has four stages, and predictive marketing can play a part in each phase.

 

 

1. Customer Acquisition

Customer acquisition is the first phase of the life cycle. This involves running ads, targeted offers, et cetera. Predictive marketing, like ad click prediction, can help generate effective campaigns to get customers initially when combined with psychographic profiles. Using analytics, we can identify the customers who will likely purchase our products or services, as well as the best ways to connect with them and which offers will entice them.

2. Customer Upsell

This particular phase is one that anyone who has ever bought anything from Amazon or Walmart.com may not realize they’ve seen. Anytime you see “Customers also bought…” or “You may also like this…”, this is predictive marketing, based on the behaviors and actions of previous customers.

3. Customer Service

Too often, this phase is undervalued. Once the register door is closed and the door chimes, as the customer exits, we count our money and wash our hands of the customer. This could hurt in the long run. If a customer has a poor experience dealing with our customer support, they’re unlikely to have a good feeling about our company and are unlikely to stay with us when the cycle turns again.

Here, we can use analytics to predict product failures and can take proactive action to address these failures. We can also determine the communication channel the particular customers will find most comfortable to work with (like phone, chat, or email). 

4. Customer Retention

Once customers begin looking to buy again, it’s our job to retain them. 44% of companies focus more on acquisition than retention, despite the fact that increasing customer retention rates can increase profits from 25%-95%.

One piece of data that is not readily available is where a person is in their particular life cycle. For example, if we know when a customer’s phone contract ends, we can ramp up offers to them. There’s no point in targeting them right after their contract starts.

Replacing SEO with Predictive Marketing

Marty Weintraub, the founder of AimClear, said in an interview with EvolvingSEO:

 

“whatever data latency period you have… you can reduce the amount people search for it by reminding them that they’re gonna search.”

 

In other words, if you are able to target customers before they start searching for your product or service, you won’t find yourself leaning as heavily on SEO and keyword searches to attract your customers.

Coupling what we know of customer trends (based on their analytics), along with Ad click prediction algorithms (the process of estimating the click-through rate (CTR) for new ads), we can optimize our ads.

Ad Click Prediction

Estimating the click-through rate for new ads is a surprisingly complicated process. Using historical click-through data, and running it through a number of different models, we can predict (with decent accuracy) which ads will be the most successful.

What variables should you care about?

If you show me an obnoxious ad before I finish this coffee, I’m less likely to click through (and more likely to destroy my phone, in a caffeine-deficient rage).

 

As with any scientific process, we start by identifying the variables. What are the things that will change between the ads that we’ll use, to determine which ones are more successful? Some of these are obvious, while others can be a bit more obscure.

  • The position of the banner. This is already something that is not approached blindly, but some sites may limit your options of ad placement. If their placement does not generate clicks, it’s not worth our investment dollars.
  • The site posting the ad. Depending on the content of the site itself, your ad may or may not get the attention of your desired audience. If the site has a heavy ad presence, your ad may be lost in the mix. If the site features content that allows for quick click and back visits, your ad may never be seen.
  • The device that the ad is being viewed on. Whether it speaks to the type of ad and how it’s viewed on a mobile device vs a larger glass device, or it has to do with the mentality of a user when they’re browsing on their phone, the type of device the consumer is using can impact whether or not they click through on your ad.
  • The time of day. Yes, even the time of day the consumer sees your ad could influence whether or not they actually click through. It could be the lack of coffee making people less appreciative of your ad in the morning, or the lowered inhibitions during a late night surf session, but the time of day has been shown to influence CTR.

Once we plug these variables into the varying ad click prediction models, we can see which scenarios will help us achieve the highest results. But which model should we use?

Choosing the Right Model

I’ll admit, this is a misleading header. The truth is, there really isn’t a reason to only use one prediction model at first.

As time goes, you’ll determine which prediction model gives you the highest accuracy. There are two different types of prediction models you should be aware of, however.

  • Individual Models – These models like the SVM, CART, KNN, and Logistic Regression models use a single model to evaluate the data.
  • Ensemble Models – These models, including the Random Forest and Boosted Tree models, combine elements of other models to evaluate the data. These tend to have higher levels of accuracy than individual models.

Conceptually, marketing hasn’t really changed all that much. It’s always been about guessing what the consumers would respond too. There were educated guesses, to be sure. In the past, they were based on the experiences and gut feelings of experienced marketers who could see trends. Now, they’re based on the infamous “big data”. As long as we have that data (see sidebar), it’s the closest thing to seeing the future, without using a crystal ball.

SIDEBAR:

There’s been no shortage of drama surrounding the collection and sale of data on the internet. With the public outcry heating up, the EU has responded with the General Data Protection Regulation (GDPR), while the US has ramped up discussion on the matter.

This doesn’t spell the end of using data analytics in our marketing, however. Part of the problem is the mystery of this data. When consumers know their data is being used in some way but have no specifics, they’ll create them themselves, and their imagined specifics are likely worse than reality.

Many of these regulations simply mean that sites will have to disclose what data they’re using and for what purpose. While there may be an initial refusal, many will realize that this data is being used to cater their experiences for them and will likely agree to it.

The other aspect of this is the data retention policies. Rather than holding on to data for long periods of time, personal user data will have shorter retention rates. Again, though, this isn’t necessarily bad. As the customer <> business life cycle shortens, the value of data decreases, the older it is. Being required to use our data quickly could give us an edge on our competition, or give us insight into trends before we would otherwise have awareness of them so that we can increase customer conversion.

At the end of the day, we will have to change how we do things. After over 20 years of doing things one way, that will be hard. But, it doesn’t have to be bad, and it certainly doesn’t spell the end of marketing as we know it.

Leave a Reply

Close Menu