We’re proud to announce that last year we delivered a total of 100 million credit scores to our clients around the world. In other words, it means that Big Data Scoring enabled to make 100 million better credit decisions for our clients – banks, nonbank lenders, telecoms, point of sale credit providers and others.
Delivering such an impressive number of scores has been possible thanks to the trust of our clients towards our scorecards and scoring models. Over the years, we’ve delivered projects in more than 25 countries and seen that our solutions outperform existing in-house or 3rd party models by 20-50% on average, which translates directly to higher EBITDA through improved loan acceptance rates and better credit quality. In addition to pure accuracy outperformance, quite often the flexility and agility of the solution and the team have played an equally important role in switching to a new provider.
The secret behind the constant outperformance of Big Data Scoring’s platform compared to more traditional underwriting solutions can be attributed to 4 main factors:
- More data. The proprietary technology developed by BDS over the past 5 years is able to collect ca 10,000 data points per each individual on the planet. More data means better models, which in turn translates to higher acceptance rates and better credit quality. Adding even more data to the solution is something we work on every day, continuously testing out new data sources and processing the ones that add predictive power to the model(s). Looking at satellite images and better understanding everyone’s hobbies and interest are only a fraction of interesting discoveries done in 2017.
- More flexibility. Quite often we see clients who have very capable underwriting teams who build world class models. However, the overall lending infrastructure (CRM, bank core, etc.) does not allow easy and swift implementation of such models. Also, champion / challenger exercises are still often done in Excel that might take ages and is prone to mistakes. In contrast, in our cloud based Decision Hub, implementing new models takes just minutes. And even more importantly, before implementing new models or policy rules, it’s easy to run real life simulations, what-if analysis and champion / challenger exercises.
- Monitoring. In-house model infrastructure often lacks real time model performance monitoring tools that make it possible to track model performance on daily basis down to single variable level. We regularly see that models which have been implemented do need fine-tuning every now and then due to various reasons (shift in customer segments or economic sentiment, changes in data formats, availability of new data, etc). Being able to spot the need for such changes and fine-tuning the models quickly gives the much desired confidence for the risk team and assures high level model performance for extended periods of time.
- Interdisciplinary knowhow. New and useful data often comes from unexpected sources. Everyone with some knowledge in big data crunching can identify some new and unused data sources, but true value (the cherry on a cake) can be hidden a lot further away. Our team includes PhD-s in a wide range of subjects, such as astrophysics, nano-magnetism, biology and forestry. And surprisingly (or not), specific knowledge from all those fields has helped us improve the models and open completely new sources of data for added predictive power.
- Experience. Last but not least, the industry specific experience it takes to build solid credit scoring models cannot be underestimated. We have seen too many times that pretty much everyone out of university can build a predictive model, but developing models that actually work is a completely different ball game. It takes decades of experience of developing underwriting models and seeing in them real life action across many economic cycles to build algorithms that actually work and deliver then modelled predictive power over a long period of time. For that, our team includes members who have long history with Dun and Bradstreet, Experian and Equifax among others.
To sum up, we first thank our new and long term clients for the trust and cooperation. And second, should anyone want to take the big data journey with us and see how we can help your business, don’t hesitate to contact us (firstname.lastname@example.org)