At BlueCloud, our mission is to drive impactful solutions by leveraging cutting-edge technologies and data-driven strategies.
One of our recent projects was centered around lead capacity smoothing for LendingTree, a leading platform in the financial services sector that relies on matching leads with lenders. The challenge was multifaceted, with a focus on optimizing how lenders manage their daily capacity for leads.
This article explores how we helped LendingTree harness the capabilities of Snowflake's AI Data Cloud in building a cutting-edge Machine Learning (ML) model to streamline lead capacity, optimize lead-to-lender matching, and drive significant revenue growth.
Managing Lead Capacity
One of the major challenges LendingTree had been struggling with for years was something we call “capacity smoothing.”
LendingTree operates through a network of lenders, each with their own definition of “campaign” - how they prefer to filter potential leads. These campaigns come with specific attributes, such as loan type and credit score, and can match up to 50 different characteristics. Each campaign has a daily capacity limit and can only accommodate a certain number of leads per day. One lead, however, can be matched with up to five campaigns simultaneously.
Lenders’ daily capacity would often fill up quickly at the start of the day. This rapid consumption meant that leads matching their campaigns later in the day were left unmatched, reducing their competitive edge in the market.
To make things worse, the manual process of following up with leads through phone calls additionally strained the smaller lenders’ operational capacity.
Scott Totman, Chief Product & Technology Officer at LendingTree explained: "A key part of the challenge was figuring out when to send leads. For example, bigger lenders could handle more volume later in the day, so we didn’t need to flood them with leads in the morning. The goal was to 'smooth' the lead distribution—spreading it out evenly across the day while factoring in lender capacity and customer preferences."
Building a Lead Capacity Smoothing Pipeline with Machine Learning (ML)
To solve the challenge, BlueCloud developed an ML solution focusing on capacity smoothing through better lead segmentation and prediction accuracy. The aim was to distribute lead matches evenly throughout the business day and impact business growth, rather than having capacities filled early in the morning. The resulting lead distribution smoothing improved customer and lender experiences.
Totman added:
“What we wanted to do was reimagine how we distribute leads. For example, instead of bringing all ten leads on a small lender at once, why not send them one lead per hour? This way, they can keep up, provide better service, and still stay within their capacity. We also considered time zones and customer segmentation. Who’s the best match for this small lender? Who might work better for a larger one? We even looked at consumer preferences—some people want a lot of interaction, while others prefer minimal contact.”
Choosing the Right Machine Learning (ML) Model
As data scientists, we have to face the truth: not every ML model is a winner.
We may build bad models—it’s inevitable. And when that happens, we don’t want to waste our time trying to salvage it. Instead, we throw it away and start over.
We did exactly that during this project.
Chase Romano, Senior Architect, AI/ML, Solution Innovation, at Snowflake, added: “We had to be cautious because if the model didn’t smooth out the leads effectively, it could mean lost revenue. It took a few tries, but once we found the right model, we were able to run it every day and start generating ROI. Balancing lender satisfaction with revenue growth was critical."
The resulting model, built and tested in Snowflake's model registry, was selected for its strong performance in producing accurate hourly predictions.
Pipeline Development in Snowflake
Snowflake’s ability to handle vast amounts of data in a scalable and efficient way was integral in making the entire process smoother.
LendingTree's lead and campaign data were stored and processed in Snowflake, which allowed us to create an ML-driven pipeline. This pipeline was designed to predict the distribution of leads and campaign traffic based on historical data on lead and campaign behavior and ensure a balanced match rate throughout the day.
“The key here is speed and efficiency. When you can build, test, and deploy models quickly, right where your data lives, the benefits are huge. So, whether you’re focusing on AI, ML, or a combination of both, think about how you can leverage Snowflake to streamline your predictive analytics—it’s a game-changer."
Lead Segmentation for Real-Time Predictions and Streamlined Lead Distribution
One of the most technically challenging parts of the project was dealing with the dynamic nature of the campaigns. Lenders frequently changed the filters and attributes of their campaigns, making it difficult to predict future matches based on campaign characteristics alone.
To address this, we implemented a lead segmentation approach that used historical data and attributes to create embedded segments for leads, enabling real-time predictions.
Hakan Sirin, Machine Learning Engineer at BlueCloud, and a pivotal contributor to the project's success, explains: “We designed a pipeline to generate hourly traffic predictions for these segments and distribute them across campaigns. A complex stored procedure ensured that segment-level traffic predictions were accurately mapped back to campaign-level estimations. Again, Snowflake's platform played a key role in executing accurate hourly predictions, sending updates to LendingTree’s backend system. This helped us achieve near real-time data throughput in a cost-effective way, which is typically a challenge for many systems.”
Dimensionality Reduction and Predictive Modeling
With over 5,000 campaigns and millions of leads, we applied dimensionality reduction techniques to manage the data's complexity and create manageable segments. Each segment's historical traffic was analyzed over the past two years, and Snowflake's hourly task feature was used to predict future traffic for the next 48 hours.
Implementing and Optimizing the Matching Engine
The next step was integrating these predictions into LendingTree’s existing infrastructure. As the client’s backend was developed with .NET, the BlueCloud team assisted in adapting the selected algorithm, which determined which campaigns would receive leads based on available capacity and matched attributes.
To test different scenarios, we worked with LendingTree’s simulation engine, which allowed us to optimize the matching algorithm for maximum efficiency.
“The revised algorithm introduced modifications that postponed certain matches strategically, ensuring that lead capacity was not used up too early in the day. This adjustment resulted in a significant increase in total selections during business hours (8:00 AM to 5:00 PM),” says Hakan.
Fueling Success: Precision Lead Targeting for Superior Revenue Growth
LendingTree can now leverage advanced analytics in ways they have never been able to do before.
“By learning as much as we could about both the lenders and the consumers, we were able to spread leads more evenly throughout the day. So even by 4:00 PM, our partners could still be getting leads and making meaningful connections without feeling overwhelmed,” says Totman.
This ML-driven capacity smoothing solution brought several benefits to LendingTree:
Increased Lead Selection Rate
Extending campaign durations and smoothing out capacity led to a 1% increase in total daily selections allowing for better utilization of available leads and higher satisfaction among consumers and lenders.
Revenue Growth
The simulations indicated that this 1% increase in selections could generate significant gains in annual revenue for LendingTree.
Totman adds: “It isn’t just about cost-cutting anymore—it is about taking a holistic approach to driving revenue. I think this project laid the foundation for real growth moving forward.”
Even for a company of its scale, this is a substantial gain.
“Visibility into spending is critical. With tools like Snowflake and partners like BlueCloud, it's not just about knowing the numbers; it's about having real control. We can track every single query, know exactly who's running it, and even set parameters to shut off queries if needed. That kind of granularity is a game changer,” says Chase.
By joining forces with Snowflake and BlueCloud, LendingTree cut database costs by 40% and slashed report runtimes from 5 hours to 10 minutes—a 97% improvement. It also gained account-level cost tracking, a single source of truth, and clear data lineage.
Enhanced Operational Efficiency for Smaller Lenders
By spreading out lead matches throughout the day, smaller lenders now have more opportunities to connect with potential clients and stay competitive on the market.
Accelerating Impact with Feature Engineering
Finally, it’s not just about the model itself—what drives success is the ability to access and leverage the right features. The entire project was a six-month-plus effort—lots of feature engineering, trial and error, and A/B testing.
Achieving Goals with BlueCloud’s Borderless Delivery Model
At BlueCloud, our Borderless Delivery Model isn’t just a framework; it’s a game-changer. By connecting clients to top-tier talent worldwide, we ensure innovative, scalable, and cost-effective solutions tailored to their unique needs.
"At BlueCloud, our strength lies in bringing the world’s best minds together to solve our clients’ most pressing challenges, no matter where they are." – Bill Tennant, Chief Revenue Officer, BlueCloud
Here's how we make it happen:
Strong Partner Validation
As an elite Snowflake partner, BlueCloud doesn’t just deliver solutions—we inspire confidence. With our proven expertise and global talent, we empower clients to modernize systems, optimize costs, and achieve strategic goals.
“Working with BlueCloud is truly rewarding because it allows my team to collaborate with customers across the globe. It’s inspiring to see how our collective insights not only help our customers succeed but also strengthen our own expertise along the way," says Totman.
Optimizing Machine Learning with Snowflake
One of the standout features of Snowflake is its ability to seamlessly manage complex data infrastructures effectively and cost-efficiently, enabling easier access and more scalable solutions. By removing much of the overhead involved in handling the backend infrastructure, Snowflake allows businesses to focus on building effective ML solutions.
“Don’t overlook ML. It’s easy to get swept up in the Gen AI buzz, but there’s so much value in classic machine learning. And with Snowflake, you can now move your ML workloads directly into the platform. Running predictive analytics at scale in Snowflake can save you a ton of money because you’re eliminating the need for constant data movement. No moving data to external platforms, no extra overhead—it’s faster, more efficient, and just makes sense,” adds Romano.
One of the most important aspects of this project was dealing with sparse matrices—an out-of-the-box technology that enables the creation of matrices with many zero values, optimizing the matching process for a large volume of leads and campaigns. For Landing Tree, Snowflake’s platform provided the essential backbone for processing large datasets efficiently.
"We could now train models directly in Snowflake, do all the feature engineering, scale up compute as needed, and then scale back down for smaller tasks. "This approach not only made things faster but also saved a lot of money. And when we went live, we added hundreds of features to the model," explains Romano
“By using Snowflake for data processing and predictive modeling, BlueCloud helped the client not only to improve lead distribution but also drive measurable profit increases.”
But it didn’t stop at spending management. Snowflake’s advanced data-sharing capabilities enabled instant, secure data sharing without the hassle of cumbersome file transfers.
Without Snowflake’s infrastructure, the project would have faced significant challenges in terms of data access, scalability, and integration.
A Blueprint for Future Success
The capacity smoothing project for LendingTree is an example of how strategic data science and machine learning can address complex business challenges and drive tangible business outcomes, all while ensuring scalability, flexibility, and cost efficiency.
Romano adds: “Getting AI and ML workloads into production can be tough. When LendingTree partnered with BlueCloud, it was a relief—I could trust them to take the lead without needing to 'retrain' anyone. They didn’t just stay updated; they actively leveraged new features. Their efficiency allowed us to run multiple implementations simultaneously, making a huge difference.”
.jpg)
“By partnering with BlueCloud and Snowflake, we not only achieved a seven-figure improvement by solving this problem, but we also did something much bigger - we built a scalable pipeline that allows us to tackle future challenges efficiently with a repeatable framework. And the beauty of this approach is that the system is designed for experimentation—quick iterations, learning fast, and moving forward.”
The lead capacity smoothing solution sets the stage for more advanced applications.
--
Read the full success story to learn more about how we helped LendingTree optimize its data management and analytics capabilities following its successful migration to Snowflake AI Data Cloud.
Explore BlueCloud’s ML services to learn how we can help you enhance your data analysis and data-driven decision making.
