For today’s marketers, the challenge isn’t acquiring customer data — we have tons of it! The real struggle is extracting actionable insights from the data that can help guide your decisions as a marketer.
Unfortunately, many businesses still rely on basic information to target their marketing messages, such as gender, location, age, and income bracket. That’s simply not enough in today’s hyper-competitive world.
You need to understand your buyers at a deeper level.
At Zaius, we wanted to provide insights that empower marketers to reach customers with personalized, relevant, and timely communications. To do that, we built a predictive model that analyzes customer behavior to look for patterns that will predict customer behavior, specifically the time to next purchase.
The ultimate goal is not just to analyze the data and tell you someone is likely to convert in the next week or month, but to find opportunities for you to take the right action at the right time to improve conversion rates.
Here’s how it works and why you should care.
Understand your buyer intent
We all know that understanding customer intent is very effective. The behavioral campaigns that everyone should already be running are triggered by “heuristics” (a fancy word for rules) that capture a handful of ways that customers signal intent. For example, if a customer adds something to cart but doesn’t buy it then they’ve demonstrated strong intent that we can action via an abandon cart campaign.
But, do we know all of the ways that customers show intent? No! We’re discovering and putting new ones to work all the time. Browse abandon and brand fan are other heuristics we’ve discovered as we’ve observed customer behavior and run experiments.
What if an AI could find more heuristics for us? While we slept? Things we haven’t thought of or noticed yet? That’s really what the Zaius model is doing: Discovering new heuristics by observing your customers and then surfacing the level of intent those heuristics represented by labeling your customers with their likelihood to buy.
How does the time to purchase model work?
The Zaius time to purchase model looks at leading indicators that a shopper is likely to purchase. The platform ingests the data from customer interactions with your brand across channels and devices. This includes actions such as:
- Read a blog post
- Subscribed to your newsletter
- Clicked an ad
- Viewed a product page (and how many pages per visit)
- Added items to their cart
- Made a purchase
It also pulls in data about the acquisition channel, whether Google Ads, social, or newsletter, and the properties of the particular products they viewed. For example, a customer who browses shorts might have a faster time to purchase than one who browses a winter coat, especially if they are visiting your site in the summer.
The model then looks at sequences and paths, how likely they are to lead to purchase, and how quickly. For example, a customer’s time to purchase might be quicker if they click on a social media post then read a blog article and then visit a product page compared to if they enter a product page via a Google ad then subscribe to your newsletter and then read a blog article.
It also takes into account the timing of these actions. If a customer opens your newsletter and closes it again two seconds later, then they are less likely to make a purchase. In comparison, someone who opens your email newsletter three different times over the space of a week and then forwards it to a friend is more likely to purchase.
This doesn’t just apply to existing and repeat customers — it also makes predictions for first-time shoppers. How? Using identity resolution technology, we can take into account the interactions a buyer had with your brand historically, back when they were just browsing your site anonymously.
Customized for your unique business
This isn’t a one-size-fits-all model. It treats each brand’s customer touchpoints and events in a unique way. For example, one company we work with collects granular data on how far into a video a customer has watched and how much of a blog article they’ve read. For their brand, those are strong indicators of purchase intent.
The algorithm we built is smart enough to be able to interpret that unique data without any analysis before input and then figure out what it means. As you change or add customer touchpoints and events, the model analyzes them in the context of different purchase paths as well. There are a number of ways that the algorithm understands your unique business.
1. Changes in strategy or messaging
If your brand changes strategies, the model reflects new paths and new behavior as they are introduced in your business. For example, if you have honed your messaging and start to personalize and target better, the model updates to reflect the way your customers are interacting with your brand now compared to last year.
2. Purchase velocity
The Zaius model also takes into account the velocity of purchases for your particular brand. Two different customers may be taking the same actions with the same high intent, but the conversion rate for different products may be faster or slower. Take the example of a mattress brand. If a customer has just bought a mattress, then even if they read your blog every week, the time to next purchase is most likely a few years. In comparison, someone who purchased a shirt recently and reads your blog each week may have a much shorter time to next purchase.
The fundamental building blocks of these two ecommerce retailers are the same, but the actions may mean completely different things. Zaius finds the rules that fit the specific circumstances. And the more data you feed it over time, the smarter the algorithm becomes. As you grow and get more examples of each path to purchase, the better predictions you can make.
How can marketers use this model to improve campaign ROI?
Using this model, we can predict the likelihood of each purchase – whether a customer is extremely likely, very likely, likely, or not likely to buy.
This prediction determines the next actions marketers should take and the type of campaigns they should be sending. For example, if a customer is extremely likely to purchase, you could send them an up-sell or cross-sell email, but if they are very likely to purchase, consider sending them a discount code or an offer of free delivery for that extra push.
If they are not likely to purchase, then you need a different approach. Consider building your brand story and explaining the value of your product rather than pushing for the sale right away.
As the model updates regularly, you can see if customers move from not likely to purchase to likely to purchase and switch up your messaging accordingly.
The model is able to predict when your customers are going to purchase and, specifically, how many days until the next order. Using that information, you can better target customers by sending everyone who is likely to convert this week a coupon or more information on the product they are browsing. For someone whose time to purchase is greater than five days, you should consider a softer approach to try to win them over as a brand fan.
Ultimately, the Zaius time to purchase model can help you:
- Harness your data to better understand customer needs, and send more personalized and relevant marketing messages
- Design and automate campaigns that better reflect and cater for customer behavior and interactions, and adapt these campaigns as new data becomes available
- Improve customer satisfaction and boost ROI by providing useful and relevant content that delivers a better buying experience