Speedyloans: Reduced cost (-30%) and increased engagement (+1%)

About Speedyloans

Speedyloans is an online lending platform that provides fast, accessible loan advances to users across the US. With a high-volume user base and millions of outbound communication events, optimizing engagement workflows and improving conversion targeting are crucial for revenue and operational efficiency.

Challenge 1: Millions of Emails, Rising Costs & Domain Reputation Risk

Speedyloans relied heavily on large-scale email campaigns to bring users back and encourage them to complete loan applications. However:

• Sending millions of emails was expensive
• Deliverability issues started affecting domain score
• A significant portion of these emails went to users unlikely to convert

The team needed a smarter, data-driven way to prioritize outreach without hurting engagement metrics.

Solution

To solve this, we analyzed historical campaign and behaviour data to understand:

• Which users were most likely to re-engage
• What patterns predicted successful conversions
• Which attributes (demographics, behaviour, timing) correlated with high ROI

Based on these insights, we built a weighted scoring model in Python that ranked users by re-engagement likelihood.

The scoring incorporated:

• Past loan behaviour
• Email engagement patterns
• Product interaction signals
• Time since last activity
• High-value attribute combinations

With this model, the team could send emails only to users above a certain score threshold, maximizing return while cutting waste.

Result

The new scoring-based targeting strategy led to:

• 30% reduction in email costs
• 1% increase in re-engagement (despite sending fewer emails)

This enabled Speedyloans to maintain strong engagement while significantly reducing operational spend and improving domain health.

Challenge 2: 100+ Attributes With No Clear Prioritization

Speedyloans was tracking over 100 user attributes that influenced conversion and revenue outcomes. However:

• Manual analysis was slow and often inconclusive
• Interactions between attributes were not obvious
• The team lacked clarity on which combinations mattered most

They needed a scalable way to find patterns that predict high conversion probability.

Solution

We applied the Apriori algorithm, a powerful association-rule mining technique, to uncover:

• Attribute pairs and combinations that strongly correlated with conversion
• High-probability user segments hidden beneath surface-level metrics
• Patterns that manual analysis would miss due to dimensional complexity

The algorithm revealed:

• Key behavioural sequences
• High-value demographic clusters
• Attribute combinations with strong conversion lift

This allowed the team to refine targeting, segmentation, and messaging with precision.

Result

Speedyloans shifted their focus toward users most likely to convert based on these attribute combinations:

• Improved targeting accuracy
• Reduced time spent on exploratory analysis
• Higher ROI across reactivation and acquisition campaigns

The company now uses these insights as part of their ongoing targeting and segmentation strategy.