Shoppers are creatures of habit. Approximately 60% of shoppers start their journey at online stores that are familiar to them, and the other 40% starts on their search engine. These habits do not mean, however, that they are going to complete their purchase at the website where they started. Approximately 22% – 43% of shoppers visit more than two online stores during their buyer journey.
They visit other online stores because they seek access to a better selection, product availability, more product information, or want to compare shipping options. Remarketing leverages these touch points as opportunities to motivate shoppers to change their habits and consider your online store offers.
A typical remarketing campaign shows advertisements to people who have visited your website. It works by “tagging” your visitors with a small piece of data, a cookie, that contain details about the visits to your online store and the items they were viewing. This cookie will then trigger the display of your advertisements on other websites.
Table of Contents
Why Predict the Prospect’s Purchase Probability?
Studies show that remarketing campaigns are more effective than email or search when it comes to conversions. But at what cost? Not all visitors who view and click on your advertisements are genuinely interested in making a purchase. Your remarketing costs, on the other hand, encompass all the views and clicks from all of your visitors, impeding a healthy ratio between the Customer Acquisition Cost (CAC) and the Customer Lifetime Value (LTV).
Predicting customer behavior and personalizing the banner ads will help with optimizing your remarketing campaign, a challenge that is not easily solved by segmenting customer behavior. These predictions are the product of Exponea’s Machine Learning capabilities; a subset of artificial intelligence that uses statistical techniques to give computers the ability to detect patterns in data without being explicitly programed (i.e., they appear to “learn” from data). This algorithm detects patterns in historical data by looking for commonalities in the online behavior of remarketed customers who purchased an item within the last 30 days. In order to understand the type of patterns the algorithm is detecting, we need to consider the current research on customer shopping behavior.
Understanding your Prospect’s Shopping Habits
The marketing analytics firm Criteo surveyed 2,023 adults in the UK about their online browsing and buying behavior. This survey uncovered the following characteristics of the retail customer’s shopping journey:
- Two-thirds of consumers say they have a specific product in mind and look for just that.
- The willingness to make impulse purchase decisions is directly related to the perceived value of a product category. Groceries have the most impulsive purchases, and electronics the least.
- 20% of shoppers in the Food & Grocery category add products to their carts over time for a weekly or bi-weekly purchase. Product categories that know a high brand loyalty, typically Health & Beauty and Apparel & Accessories, also have the most site-loyal browsers. Each product category has its own purchase decision cycle. Shoppers of the electronics category consider the most options when making a purchase. They evaluate an average of 19 products over a timespan of 10 days before making a purchase. Highly brand loyal and low consideration product categories such as Health & Beauty have an average of 8 products over 6.6 days.
|Product Category||Average Purchase Cycle||Number of Products Viewed|
|Baby Care||7.3 days||10|
|Health & Beauty||6.6 days||8|
|Apparel & Accessories||8.4 days||13|
|Toys & Games||7.1 days||13|
Designing Effective Remarketing Strategies for your Product Categories
These outcomes reveal some significant findings about the degree and manner in which customers are open to outside influence. An effective remarketing strategy understands that product categories represent a balance between a prospect’s consideration process, willingness to make impulse purchases, brand loyalty, and perceived value. For instance, Food and Groceries is a low consideration category with a low perceived value. Customers are therefore more willing to make impulse purchases. Remarketing strategies that are centered around abandoned cart reminders are highly suitable for this type of product category. Display advertisements can present a combination of cart items, product recommendations, and complementary products.
Electronics, on the other hand, being a high consideration, high-value category, would benefit from a remarketing strategy that motivates customers to spend more of their consideration process on your website. Successful display advertisements present products that reflect an awareness of their search intent and display recommendations based on their interests.
Identifying the Purchase Probability Sweet Spot of a Remarketing Campaign
The Machine Learning algorithm associates detected patterns with purchase probabilities. The below chart gives a birds-eye view on the algorithm’s output, which shows that there is a direct correlation between customers’ purchase probability and their site activity. This chart, however, does not explain how a particular amount of site activity relates to a purchase probability.
In order to have a better understanding of these activities, we need to take a deep dive and examine the underlying behavioral patterns. The table below illustrates what a purchase probability range means in terms of behavior.
|0-20||Incidental browsing behaviors underpin this probability range. Customers either check out the product categories and leave, or check the product pages, sometimes adding products to their carts to check prices. They often do not return to your website to buy the product they have in mind.|
|20-60||This is the range where your online store is actively competing with other websites in your customer’s consideration process. They return to your store throughout the duration of this process. The frequency of their visits is decreasing if the competitor’s offers are more attractive and increasing if the customer intends to buy the item at your store. In this case, we also see a narrowing down in the number of recurrently viewed items.|
|60-100||Customers in this range often use your online store as a starting point for their shopping journey, but they also use your website in a manner that is different from the other probability ranges. What stands out is the number of devices that the customer uses in the consideration process. Their journey may start on their phone or tablet, but switches to a laptop or desktop computer when the purchase is intended to be closed.|
Implementation and Outcomes
One of our clients is a mid-sized fashion e-commerce retailer offering fast-fashion clothing for men, women, teenagers, and children. They use Facebook as their platform for their remarketing activities, which allows them to personalize their advertisements by uploading customer lists that are segmented by the interests they expressed when browsing their website. This setup delivered our client a return on advertising spend of 16.15% with a conversion rate of 3.26%. We hypothesized that these figures would noticeably improve if we would filter out the customers who have a low purchase probability. We uploaded two months of historical customer data to our predictions algorithm. We applied the detected purchase probability patterns to the ongoing stream of customer events. Customers who were detected as having a purchase probability higher than 20% were added to the customer lists and uploaded to Facebook. This resulting setup led to an increase in return of advertising spend to 19.01% and the conversion rate to 5.38%.
Further Considerations for your Remarketing Strategy
With an increase in return of 15% for the same advertising budget, this result demonstrates that predictive remarketing improves the efficiency of your advertising spend. This particular setup with Facebook Audiences also allows us to explore other feasible alleys of advertising and remarketing optimization. Facebook offers “Look-a-Like” audiences. These audiences consist of users who share traits, such as location, age, gender, and interests, with your online store customers. Advertising to this audience is shown to be a useful way of reaching new prospects.
Customers with a history of low purchase probabilities could be excluded from the “Look-a-Like” audiences, therefore improving the conversion rate of your marketing campaigns. Purchase probability predictions can also prevent over advertising. Customers with very high purchase probabilities are already convinced of your brand and offers. Advertising to these customers may not only be redundant, but it also can turn them off from your business as they become annoyed and bothered. Excluding them from your remarketing campaign will further reduce your advertising spend while improving your profitability.
Remarketing can be a remarkably effective method for competing with other e-commerce businesses as long as your strategy recognizes your customers’ intent and consideration process. Each product category has a distinct purchase cycle and consideration process. Our Single Customer View can provide you with insights about individual customer journeys, the products those customers consider, and the manner in which they use their shopping cart. Once your remarketing strategy reflects a balance between your customer’s consideration process, their willingness to make impulse purchases, brand loyalty, and perceived value, it can be actioned by creating display advertisements with the right combination of products. Further, the prediction algorithm can help you disseminate these advertisements to the right customers. As we saw with our fashion e-commerce client, implementing this prediction algorithm can lead to a significant improvement of the ratio between your Customer Acquisition Cost (CAC) and your Customer Lifetime Value (LTV).