The Client

A leading higher education organization that provides various degree programs through online and offline classes across United States. The institution spends heavily on digital marketing campaigns that focus on customer acquisitions in an attempt of converting prospects into leads. 

It is crucial for the organization to generate leads that they can convert into students. Leads are created when website users fill out a Recruitment Information Form.

The Challenge

The client had recently seen a decrease in the overall conversion, and they were finding it difficult to convert more leads. New student enrollment was also stagnant. To remedy this issue, the institution tried multiple strategies for 8-12 months. They conducted a bunch of A/B tests in a bid to increase their conversion rate. This helped increase conversion rate initially, but eventually hit a plateau.

Nabler’s Response

Nabler developed a personalized solution that would gauge the intent of visitors to convert based on their behavior on the website.

The Intent Scoring Model analyses user activity, and groups of users, into high-, medium-, and low-intent buckets. High Intent users have a maximum conversion rate of 50-60%; Low Intent users have the lowest conversion rate of one-percent or less.

Based on these buckets, medium-intent users can be targeted in order to migrate them into high-intent bucket. This model can also be deployed to find look alike users who exhibit behaviours similar to those of other users. These user segments are used to improve channel acquisition.

Intent Scoring Step-by-Step

1. Data Cleanse
   As part of solution development, Nabler first cleansed, standardized, and harmonized the client’s clickstream data.
   
   
2. Visitor Behavior Analysis
   In order to understand visitor behavior, Nabler performed a detailed exploratory analysis of the data. Once visitor behavior was understood, we identified a few, key patterns to associate with conversions.
   
   
3. Intent Score Model
   Nabler used insights from the visitor behavior analysis were used to build a mechanism to help the client target visitors by providing them with personalized content. A machine learning model was trained to predict the intent of visitor conversion while they were still active on the website. Further, we grouped visitors with similar intent into the three segments: High, medium, and low intent visitors. Segmentation of visitors made it easy to identify and better target them.

Personalization

After identification of the right segment, the next challenge was personalization. We coupled intent score model with a recommendation engine that personalizes the visitor experience by taking their intent into consideration.

The model was integrated with the client’s website to provide real-time personalization.

Prominent Features:

• Identify different visitor segments based on their website activities
• Increasing website engagement with personalized content
• Provide a list of potential visitors to retarget
• Web-based dashboard that provides actionable insights

Technology & Infrastructure

Nabler developed this solution using the Google Cloud Platform. The following components were used for deployment:

• API Hosting: Cloud Run
• Analytical Data Set: Google BigQuery
• ML objects Storage: Cloud Storage
• Compute: GCP Compute Engine
• ML Results: Google BigQuery
• Dashboard: Google Data Studio

The Outcome

The client deployed Nabler’s solution on their website over a 30-day testing period. During this time, we monitored the Intent Score outcome on different user segments. Following the testing period, the client deployed the model, along with recommendations, across all sections and visitors. Experiments were designed on Tealium Audience Stream.

The client has seen a 15% increase in conversion rate on these segments compared to rest of the website traffic. Overall conversion rate has increased by 30%.