Identifying Your Most Likely Buyers


A MACHINE LEARNING USE CASE

Download Now

Get the full use case now in a printable format or continue reading below.

Overview

In this use case, we describe how the Velocidi CDP’s proprietary machine-learning predictive models helped a busy marketing team develop more efficient and effective marketing campaigns by:

01

Predicting buyer intent.
02

Autonomously identifying visitors who are most likely to buy in the next 7 days.
03

Enabling their team to deliver tailored marketing messages to visitors based on whether they are more or less likely to make a purchase.

The Challenge

Before using Velocidi, a direct-to-consumer pet food brand was unable to identify which of their website visitors were most likely to make a purchase.

As a consequence, they were reaching out to all of their visitors, prospects, and customers with similar marketing messages, regardless of the demonstrated interest in making a purchase.

This resulted in wasted ad spend and a sub-optimal customer experience.

The Solution

To tackle this challenge, the CDP was deployed in the brand’s AWS environment, and their website was tagged using a first-party CDP tag. Note that because the first party-tag was setup using the brand’s domain, it is not obstructed by ad blockers and is able to track more effectively than a third-party tag.

The first-party tag then collected event-level data on website visitors and all relevant actions they took on the site. Customer profiles were dynamically created from the attributes and events collected and fed to Velocidi’s proprietary ML predictive model to identify the visitors more likely to make a purchase.

Using all the data insights uncovered by the Velocidi Costumer Data Platform, the model was able to build a complex behavioral profile based on customers who had already converted. As new potential customers perused the website, the model applied a dynamic likelihood to buy score of 0 to 1 based on their real-time behavior.

Then, the machine-learning predictive model segmented all visitors into one of two groups, according to their intent to purchase. Website visitors whose behavior resembled the buyer journey of existing customers scored higher on the 0 to 1 cut-off scale, which meant they were included in the “higher intent” group.

The Results

Velocidi’s ML predictive model automated the creation of an audience segment with website visitors that had a higher likelihood of buying within the next 7 days. This segment, that reflects the visitors in the “higher intent” group, had a size equal to 34% of the brand’s total website visitors and captured over 95% of visitors who would buy within the next 7 days. Not only that but the visitors in the higher intent group demonstrated as much as 40x more likelihood to purchase than the ones in the lower intent group.

Predictive model cut-off scale visualization

Using a 0.2 likelihood to buy score cut-off with the model we created allowed us to gather 95.6% of the actual buyers within a segment that’s only 34% of the whole universe of visitors. This is the optimal customer concentration in the “higher intent group”, that raises its likelihood to buy to 40x higher than the low intent group.

Using this audience, the brand is able to:

REDUCE ad spend by 60% while maintaining sales performance. Or keep their ad spend and boost sales to a new level.

IMPROVE customer experience by removing inefficient retargeting.

NURTURE low intent visitors to increase sales.

DELIVER messaging tailored to each customer’s real-time intent level as it changes throughout their buyer journey.

REPURPOSE their saved budget and use it for other marketing initiatives.

OPTIMIZE discount offers and promotions.

Share this!

Share this!

Download

Get the full version of this machine learning use case in PDF form.

Start growing your brand now

Get a demo and see how Velocidi takes the guess-work out of marketing by making machine learning practical and accessible.