In this second instalment of our "How do we do that?" series, we delve into the detailed and meticulous process behind creating risk baskets. At Consumer Intelligence, these risk baskets or Unique Quote Records (UQRs) are fundamental to providing nationally representative, accurate, and ethically sourced data for our clients. But how exactly do we ensure these risks reflect the complexity of the real world?
High-quality data doesn't happen by accident; it requires meticulous attention to detail, transparent processes, and rigorous governance. Building from the ground up, we have designed our data systems to fully comply with ESG (Environmental, Social, and Governance) standards as well as GDPR. This foundational commitment means that our data collection and usage practices are inherently sustainable, ethical, and reliable.
Accurately representing the insurance marketplace requires carefully crafted datasets, balancing real-world authenticity with methodological precision. Our aim is always to build a nationally representative set of profiles while also ensuring our real data sources, individual consumers, remain unaffected by our analysis.
We start by identifying real people whose data closely reflects genuine consumer scenarios. To safeguard these individuals, we carefully manage the timing and use of their personal information. We specifically track their real insurance renewal dates, making sure to avoid using their data during their personal renewal window to prevent unintended impact from our mystery shopping activities.
Once the right individuals have been identified, the next step is constructing risk baskets that accurately represent the national picture. This involves meticulously ensuring diversity across critical variables such as age, region, driving history, and countless other nuanced details. Each basket must balance detailed specificity with broad representativeness, requiring significant expertise and precise control.
For over a decade, our risk baskets required expert developers to carefully "hand-cook" these detailed profiles, ensuring internal consistency. For example, drivers can’t have convictions recorded before their licence was issued such details require meticulous manual attention. Recently, we've started to leverage artificial intelligence (AI) to assist our team, enabling deeper precision and efficiency. With over 140 variables for each risk profile, AI tools significantly enhance our ability to maintain data accuracy.
A crucial aspect of our risk construction approach is deliberately including scenarios outside the comfortable core or "Vanilla-verse" of standard insurance practices. By doing this, we aim to encourage insurers to confidently price risks beyond typical boundaries. This inclusivity aligns with our moral duty and our core purpose of building confidence within financial services, making insurance accessible to as broad an audience as possible.
Our approach has occasionally faced criticism: why not recycle familiar, easily managed risks repeatedly? Why complicate matters by embracing more challenging scenarios? Simply put, because accuracy and inclusivity matter. While our method has its challenges and isn't perfect—no method is—our commitment to authenticity and representation remains unwavering. We are clear and transparent about this, rejecting the notion of an easy but flawed solution.
At Consumer Intelligence, integrating machine learning on both the front and back end of our risk construction process has proven transformative. It supports better initial data selection, enhances quality control, and significantly refines the final analysis. This powerful combination of human expertise and technological innovation ensures our data remains robust, representative, and reliably useful.
In future articles, we’ll delve deeper into how machine learning specifically enhances our analytical capabilities. But for now, this is how we create our accurate, balanced risk baskets—today and for tomorrow.