A product built on a flawed assumption

Digital-first life insurers aim to build products that are more personalized and behavior-based. Instead of relying solely on age or historical risk tables, they want underwriting models that reflect real-life health patterns. In practice, that shift depends on one critical factor - access to reliable health data.

The core constraint is data. Most digital underwriting models rely on self-reported health information, a method that was never designed for clinical precision. Customers complete questionnaires in good faith, but the answers cannot be objectively verified at scale. The gap between reported and measurable health data is not primarily a fraud issue. It is structural: the underlying collection method limits accuracy from the outset.

For Lyfery, this was not an abstract industry challenge. Their scoring model, which enables customers to receive up to 50% premium discounts based on health behavior, depends directly on the quality of its inputs.

When incentives are tied to behavioral scoring, the integrity of the underlying data becomes critical. Models built on unverifiable inputs struggle to differentiate meaningfully between low and high risk.

Mihkel Mandre, founder and CEO of Lyfery, explains the thinking behind their shift toward more objective data sources:

"We aim to ensure that as many inputs into the scoring model as possible are data-driven rather than self-reported. The more objective the data, the stronger its impact on the score. That’s also how we manage the risk side."

‍

Where a 30-second face scan changes the equation

Blood pressure is one of the most important indicators of cardiovascular risk and one of the most difficult to collect objectively outside a clinical setting. Many individuals do not know their current reading unless it has been measured at a recent doctor appointment. In digital underwriting models, self-reported blood pressure is therefore often incomplete or inaccurate.

To address this gap, Lyfery integrated Shen AI’s camera-based measurement technology. Using a standard smartphone camera, a 30-second face scan provides an estimated blood pressure reading without additional hardware. Within a behavioral scoring model where cardiovascular health carries meaningful weight, replacing self-reported input with objective measurement materially improves data quality.

Clinical credibility was an important factor. Kadri Haller-Kikkatalo, Chief Health Strategy Officer at Lyfery and a practising medical doctor, conducted her own comparisons before supporting the integration:

"We compared this data with manual measurements and clinical equipment. It is accurate and reliable, more reliable than self-reported data. And importantly, many people don't know their blood pressure unless they visit a doctor's office. This technology could change that."

‍

Lyfery also conducted a live comparison during one of Estonia’s largest investment festivals, testing Shen AI face scan readings alongside traditional cuff-based measurements. The results showed strong alignment across the sample, helping build internal confidence, including among team members with scientific backgrounds.

‍

The 'wow effect' is useful, but it's not the point

When Mihkel Mandre presents Lyfery to customers or partners, the face scan often serves as an entry point. A 30-second smartphone measurement that generates an immediate health reading creates strong engagement. During a live appearance on one of Estonia’s largest morning television programmes, the scan was performed on air, with results displayed in real time (see the clip here).

However, the commercial relevance of Shen AI within Lyfery’s model goes beyond demonstration value. Within Lyfery’s scoring framework, objectively measured data carries greater algorithmic weight than self-reported inputs. Integrating camera-based blood pressure measurement therefore influences scoring outcomes directly. Customers who complete the scan contribute higher-integrity data to the model, which can translate into more accurate risk differentiation and, where applicable, more precisely calibrated discounts. This reflects a broader design principle: in behavior-based underwriting, data quality and pricing integrity are inseparable.

‍

Rethinking what insurance data is actually for

Lyfery’s scoring model incorporates physical activity, nutrition, smoking and alcohol habits, social and mental wellbeing, and participation in national screening programmes. The breadth is intentional. The aim is to capture meaningful risk factors while ensuring that inputs remain measurable remotely and practical for customers to complete.

As Kadri Haller-Kikkatalo explains:

"We picked those parameters in different areas - the most significant risk factors, but also the ones which are actually measurable from a distance. It's a calibrated balance between what we would like to ask and what we need to ask."

‍

The model is not designed as a one-time assessment. Over time, annual data collection builds a longitudinal health profile. Lyfery’s next development phase focuses on returning that insight to customers, contextualizing their position relative to broader population benchmarks and identifying changes that could meaningfully improve their risk trajectory.

In this approach, insurance shifts from static risk classification toward ongoing health engagement. But that shift depends on one foundational condition: the underlying data must be trustworthy. Objective measurement is the prerequisite for building credible, behavior-based insurance products.

‍

Toward a more robust model of digital underwriting

Lyfery is one example, but the underlying constraint is industry-wide. Any digital insurer building behavioral or lifestyle-based underwriting faces the same structural limitation: the greater the reliance on customer-reported inputs, the greater the exposure to data variability and inaccuracy.

Alternative data sources are increasingly accessible. Wearable integrations, passive mobile data, and camera-based health measurement can introduce more objective inputs into digital models. The strategic decision is not whether these tools are available but how they are used. Are they positioned as engagement features, or integrated as core infrastructure for underwriting logic?

Lyfery’s approach demonstrates that embedding objective measurement into the scoring framework can influence portfolio composition, improve risk differentiation, and enable longitudinal health insights over time.

‍