Improving sales effectiveness by predicting students' enrollment
Talent Garden
- Customer case
- Data Engineering
- Data projects
- AI & Data Science
Talent Garden provides masterclasses and training programs to students, engaging with them through various online and offline touchpoints. Online interactions include completed contact forms and information requests, while offline touchpoints involve meetings and calls with Talent Garden’s sales team. Throughout the customer journey, from initial contact to final enrollment, Talent Garden collects extensive data*. With a wealth of raw data at their disposal, they sought to improve their enrollment strategy and the effectiveness of their sales team. To achieve this, they asked us to develop a data model that could better predict the likelihood of a new contact eventually becoming a student.
*Talent Garden has always operated in accordance with GDPR regulations, working with anonymised data.
Approach
The dataset consisted of raw, unstructured data, with significant gaps, as many leads did not progress to the next stage of the customer journey. While there were thousands of leads, only 15% advanced to a meeting with the sales team, and just 2-3% ultimately enrolled in a masterclass or training program.
To extract meaningful insights from the dataset, we had to get creative, employing a trial-and-error approach to select and combine the right data. Key types of data we processed included:
- Client-related data: job position, industry, motivation, and educational background
- Course-related data: start/end dates, pricing, course names, and delivery formats
- Sales interaction data: meetings, calls, and lead scoring
- HubSpot data: page views, email interactions, and website visits
We began by cleaning and enriching the dataset, then created new features to enhance predictive power. For instance, we developed a feature based on time, showing that the shorter the time since the most recent contact, the higher the likelihood of conversion.
We selected a model for the classification task, trained it using Talent Garden’s data, and fine-tuned its parameters. Additionally, we experimented with several other models and combinations to determine if any could yield better results.
Next, we focused on feature reduction. Initially, we had around 80 features, which was too many for an explainable model. We streamlined this down to 20 key features. Features like course pricing and email interactions proved to be highly significant in predicting enrollment.
Result
Talent Garden can now predict the enrollment of their leads with 80% accuracy. After the first contact, the accuracy is already at 70% and it improves dramatically after stage 2 of the customer journey, when they often meet the sales team at the office. Using this information, the sales team invests their time in the leads that have the highest chance to convert.
Future
While the result is satisfactory, there is significant room for improvement. We can further enrich the dataset by applying more precise methods to address missing information. In addition to the features already included in the data model, Talent Garden possesses a large amount of textual data, such as interview notes and call records. By leveraging natural language processing (NLP) techniques and models, we can extract valuable insights from this data.
Want to know more?
Joachim will be happy to discuss the possibilities of predictive data models for your organisation. Our experienced team of Data Engineers will be happy to get to work on your data issue.
Commercieel Manager Data Engineering+31(0)20 308 43 90+31(0)6 23 59 83 71joachim.vanbiemen@digital-power.com
Receive data insights, use cases and behind-the-scenes peeks once a month?
Sign up for our email list and stay 'up to data':