What is predictive analytics? Learn how forward-looking AI can grow your revenue

---

Ever wish you had a crystal ball or a deck of tarot cards on hand, ready to guide your marketing efforts? 

According to new research, 74% of consumers expect personalized experiences from B2C brands. A glimpse into the future—maybe even an in-house psychic—would surely help. 

You don’t need a psychic on staff to give your business an edge. Predictive analytics is there instead, helping you understand customers’ future actions to empower more relevant online shopping experiences.

Powered by your customer data—and fueled by artificial intelligence (AI)—these forward-looking insights let you anticipate the needs of your individual customers right down to the profile level.  

And by applying those insights to your marketing efforts, you can take your strategy to the next level—improving engagement, reducing churn, and building customer loyalty in the process.

What is predictive analytics and how does it work? 

One of 4 main types of data analytics, predictive analytics helps businesses recognize patterns in their data to fuel forward-looking growth.  

While diagnostic analytics looks backwards, rooting out the cause of data trends, predictive analytics uses AI and machine learning to look forward—seeing how the trends, patterns, and relationships in the data today are likely to play out in the future. 

Predictive analytics isn’t a crystal ball, exactly—the predictions it offers, based on the data you already have, aren’t guaranteed to come true. But by using historical data to determine likely future scenarios, it fuels more future-forward decision-making. 

And predictive analytics tools that incorporate AI insights are able to dig deeper, empowering even more complex predictions and future scenarios.  

3 key benefits of predictive analytics

As a marketer, your goal is to earn revenue to grow and scale your business. All businesses, no matter what the industry, have this in common. AI-driven insights can help you make progress on each of your key business priorities to ultimately drive growth. Here’s how:

Predictive analytics use cases for B2C: turning insights into action

Predictive analytics gives you a variety of new marketing automation metrics to help gauge the health of your customers and your business. But before you can reap the benefits, you need to understand what each insight can tell you. Here, we break down use cases for leveraging each metric in your own retention strategy.

Average time between orders

This refers to the average amount of time that passes between orders for a given customer. Here are a few use cases for this metric:

• Export this data for your key segments so you can compare the average time between orders for specific groups of your customer base. 
• Build segments to group customers with similar buying cycles together. Then, target customers who don’t purchase frequently with offers that incentivize them to purchase more often.

Average order value 

This refers to the average amount of money a given customer spends every time they place an order. Here are a few use cases for this metric:

• Export this data for your key segments so you can compare the AOV of orders placed by your VIP customers against the AOV of other segments.
• Segment your customers based on AOV and then create targeted initiatives aimed at raising the AOV of customers who don’t spend a lot. 
• Branch your flows based on AOV, and encourage customers who usually don’t place big orders to add more items to their cart.

Predicted number of orders

Klaviyo displays the historic number of orders for a given customer, as well as a prediction of how many more orders that customer is likely to place in the next 12 months. Because of the nature of this calculation, the predicted number of orders on a profile may be a non-whole number.

Here are a few use cases for this metric: 

• Build your own loyalty framework in Klaviyo based on how many orders a customer has placed. You can do this by building a variety of segments that group customers based on their historic number of orders. 
• Build segments based on how many more orders customers are likely to place to anticipate who is likely to purchase again and who is unlikely to purchase again. Trigger flows based off of these segments, or retarget customers who are likely to purchase again through Facebook Ads or Google Ads.

Expected date of next order

This date-based property predicts the expected date when a customer will probably purchase again. In Klaviyo, this is the expected date without intervention, but marketers do have the power to influence it.

Here is our favorite use case for this metric, described by Anna Khomenko, product manager at Klaviyo. 

“Expected date of next order should be used for personalization and together with other properties. Why? Let’s look at a scenario with three different customers.

• Customer A has bought something from your brand every month over the last 2 years. If they buy again in March, their expected date of next order will be in April. 
• Customer B has bought something with your brand once every 6 months over the last 4 years. If they buy again in March, their expected date of next order will be in August. 
• Customer C has only purchased once from your brand––and 5 months ago in September. Their expected date of next order was November (because typically, customers for this business purchase their 2nd order within 60 days). 

These are all very different scenarios, and they need different content to help move each of those customers closer to their next purchase. 

• For customer A, maybe you do not need to send anything prior to their expected date of next order because of how consistently they purchase. You could push them to your loyalty program instead. 
• For customer B, you don’t need to wait until their expected date of next order to try and convert them again sooner. You can send this customer some campaigns with A/B tested messaging to see what might get them to convert sooner. 
• For customer C, the expected date of the next order has come and gone, and this customer didn’t buy anything. That’s ok. You can still identify them and send them campaign content or newsletter content to build the relationship with your brand. Or, perhaps they had an issue with their last order. Did your service team properly address that? Is there a path to remedy for the customer?”

See it in action: When men’s personal care brand Every Man Jack adjusted their replenishment flow to send on or slightly before each customer’s predicted date of next order, it was part of a broader AI-powered marketing strategy that helped boost revenue from flows 25% YoY. 

“I trust and value Klaviyo AI because it saves me time, it helps me leverage our customer data to personalize our email timing and strategies, and most importantly, I maintain complete control over how and when it’s used,” says Troy Petrunoff, Every Man Jack’s senior retention marketing manager.

Customer lifetime value 

Within a profile in Klaviyo, you can see a breakdown of the following metrics:

• Historic CLV: the value of what a customer has spent over their lifetime
• Predicted CLV: how much Klaviyo predicts they will spend within the next 365 days
• Total CLV: the sum of historic and predicted CLV 

Here are a few use cases for this metric:

• Export this data for your key segments so you can compare the CLV of orders placed by your VIP/best customers against the CLV of other segments.
• Build segments based on CLV and then use these segments to find new customers via lookalike audiences in Facebook Ads or similar audiences in Google Ads.
• Retarget customers with a high predicted CLV through Facebook Ads or Google Ads.
• Branch flows based on predicted CLV to deliver tailored offers and incentives.

See it in action: The Willow Tree Boutique uses this metric alongside the AOV metric to send campaigns highlighting pricier apparel to segments with demonstrated spending power—customers with a predicted CLV over $500, or an AOV over $150. In their first 6 months using Klaviyo predictive analytics like these, they grew revenue from campaigns 53.1% HoH.

“Since starting with Klaviyo’s predictive analytics, I hardly ever send a campaign without them,” says Jade Richardson, email strategist at Willow Tree’s digital marketing agency Agital.

Churn risk prediction

This percentage calculation shows how likely or unlikely a profile is to make another purchase. Most customers across most industries are one- or two-time purchasers, so most of your customers will likely have a churn risk prediction over 50%. A lower churn risk percentage indicates a higher likelihood that a customer will purchase again.

Here’s a potential use case for this metric:

• Export this data for your key segments so you can compare the average churn risk prediction between groups. For example, it may be helpful to know what your average churn risk prediction is for your VIPs vs. the average churn risk prediction of your entire customer base. 
• Use churn risk in combination with expected date of next order to better understand what kind of marketing might resonate most with a customer and personalize your content—for example, whether you should be targeting someone with a win-back campaign, or whether someone is a loyal customer who would respond better to regular communication.

Best cross-sell date

This refers to the best time to cross-sell to a customer, based on the purchasing behaviors of similar profiles. Here are a few potential use cases for this metric:

• Trigger a flow based on someone’s best cross-sell date to see if you can encourage their next purchase faster. 
• Branch flows based on the best cross-sell date to avoid giving unnecessary monetary incentives to folks who are already likely to purchase.
• Create on-site forms that target site visitors whose best cross-sell date is coming up soon, and present them with a deal designed to increase AOV (i.e., a bundle offer or a buy-two-get-one-free offer).
• Personalize emails, texts, or push notifications with dynamic content based on someone’s best cross-sell date. 

Next best product

This metric predicts the next best product for a customer, based on their most recent purchase. As a customer continues to place orders, this property automatically updates accordingly. 

Here’s a potential use case for this metric:

• Personalize emails, texts, or push notifications with dynamic content based on someone’s next best product. 
•  
• Leverage Klaviyo’s audience sync to serve retargeting ads on social media platforms for the next best product. 

Important note: In Klaviyo, all predictive insights are based on a 365-day forward-looking prediction window. If you are using Klaviyo Marketing Analytics in addition to Klaviyo Marketing, you have the ability to customize your CLV calculation window to be shorter or longer based on your business needs. Best cross-sell date and Next best product are also exclusive to Klaviyo Marketing Analytics and Advanced KDP customers. 

How to avoid biased predictive analytics 

To be able to accurately root out patterns and trends, predictive analytics needs a healthy amount of data to work with. That’s why, to see the predictive analytics section on your Klaviyo profiles, you’ll need enough data to support it, including:

• At least 500 customers who have placed an order with your business 
• At least 180 days of order history, with orders within the last 30 days
• At least some customers who have placed 3 or more orders

With this reliable store of data to draw from, Klaviyo’s predictive analytics can make more accurate predictions for the future. The more historical data you have to draw from, the more accurate your patterns and trends will be.

The best predictive analytics rely on a 360-degree view of individual customer behavior over a long period of time. If you’re using a lot of point solutions and not consolidating data in a central source of truth, or not doing so in near real time, then your predictive analytics is biased to whatever data it has access to. 

This is why Klaviyo connects with more than 350 tools and platforms across retail, restaurants, hotels, and wellness verticals. And, we never delete customer data. What you collect is what we store, unless you ask us to remove it. 

The best predictive analytics use holistic, historical customer data. That’s exactly what Klaviyo provides. 

How Klaviyo helps you predict and personalize with analytics

Klaviyo provides predictive analytics out-of-the-box so you can immediately spotlight patterns in your data, giving you ideas for segmentations, flows, and campaigns to encourage or counter patterns for subsets of customers. 

But with Klaviyo’s advanced Marketing Analytics, you can customize prediction windows and access new predictive analytics, like best cross-sell date and next best product. This makes it easy to understand and take action on the full customer journey, product purchase patterns, and business performance, so that you can implement smarter targeting, merchandising, 1:1 personalization, and marketing optimization. 

Here’s a quick breakdown of how the analytics included with Klaviyo differ from our more advanced Marketing Analytics––and where predictive analytics falls within those offerings (spoiler alert: you get it out-of-the-box).

	Already included with Klaviyo Marketing	Unlocked with advanced Marketing Analytics
Marketing and segment performance reports	✅	✅
Industry benchmarks	✅	✅
Customizable attribution and metrics	✅	✅
Predictive analytics and customer lifetime value	✅	✅ +Custom prediction windows
Advanced personalization with customizable RFM analysis		✅
Advanced, actionable purchase pattern and catalog insights		✅
Customer behavior funnels and cohort reports		✅
Audience and conversion dashboards		✅