Real-Time Machine Learning
for Financial Services

How Tide built real-time fraud detection and loan approval ML applications using Tecton’s feature platform

Operational machine learning is challenging to get right, especially when real-time data and online inference are involved—but it can be done.

For Tide, a UK business banking provider for small and medium enterprises, the ability to detect fraudulent transactions and make credit approval decisions in real time is key to business success. 

Watch this 1-hour webinar where Hendrik Brackmann, VP Data at Tide, and Mike Del Balso, Co-founder and CEO of Tecton, discuss the key areas of collaboration the Tide team focused on during their first attempt at building a feature store to support their real-time machine learning use cases, namely:  

• Adding features. Adding features from new topics required significant overhead. 
• Streamlining changes to production.  Approval rates in production were lower than expected than at training time. 
• Improving iteration times. Adding new features required extensive engineering support as Tide focused on creating feature pipelines. 
  
  

Hendrik also shares how his team collaborated with Tecton to realize the benefits below:  

• Easy re-use of high-quality features in different models within the same domain.
• Ease of backtesting. Ability to estimate the holistic impact before deploying changes into production. 
• Faster deployment time (by 50%!) due to the use of pre-built feature types.
• And as a result, more accurate models generating tremendous business value.