Case study

The Hartford cut risk report production from weeks to 2 days

Risk teams waited 2 to 3 weeks for consolidated risk reports built by hand from policy, claims and catastrophe data.

The Hartford

The Hartford · Underwriting and risk

85%less time to produce a full risk report

The workflow before#

The Hartford is one of the largest property and casualty insurers in the US, and its risk teams decide where the company takes on exposure and where it pulls back. The data behind those decisions sat in separate policy, claims and external catastrophe systems, each structured differently for each line of business. Analysts pulled it together by hand into static reports and PowerPoint decks. A comprehensive risk report took 2 to 3 weeks, so by the time executives saw it, the numbers were already old. Metrics refreshed monthly or quarterly. The dashboards that did exist were hard to trust and harder to use, and only 30 to 40% of the people meant to use them did.

What we built#

We designed and built a set of Tableau dashboards on top of The Hartford's data warehouse, each one built around a specific decision. Executives, underwriters, catastrophe modelers and claims teams each get the view they need, all drawn from the same daily-refreshed data. The analysts who used to build decks by hand now spend that time on the questions behind them.

A dashboard for each decision

We built five views: executive risk, underwriting risk, catastrophe exposure, claims risk and portfolio management. We designed around decisions instead of data sources because a risk officer asks where exposure is growing, not what's in the claims table. Each view answers the questions its users bring to it.

Drill from KPI to policy

Executives start from a KPI summary and drill down to policy-level detail without leaving the dashboard. Underwriters see risk scores, geographic concentration and appetite compliance in one place, with what-if parameters to test a change before they make it.

Catastrophe and claims exposure

The catastrophe view shows probable maximum loss by peril and return period, mapped to The Hartford's own insurance territories. Claims teams track loss ratios, reserve development and emerging patterns, so a shift in claims shows up in days rather than at the next quarterly review.

Daily data, scoped by role

Data refreshes daily on Tableau Server with row-level security, so each team sees only what it should and nobody waits for a monthly cycle. Putting security in the data rather than building separate dashboards per team meant one set of views to maintain and one version of the numbers.

How we delivered it

We worked in two-week sprints with The Hartford's risk analysts, showing working dashboards early and adjusting them to how each team actually worked. Old and new reporting ran side by side for a two-month validation period so users could check the numbers for themselves. Hands-on training ran alongside, and that is a large part of why adoption roughly doubled.

Before and after#

MeasureBeforeAfter
Full risk report2 to 3 weeks2 days
Risk data refreshMonthly or quarterlyDaily
Ad hoc risk analysis1 to 2 weeks1 to 2 days
Active dashboard users30 to 40% of intended usersAbout 70%

What's next#

Policy, claims and catastrophe data now feed one governed reporting layer. That sets up a risk assistant where underwriters and risk officers can ask about exposure in plain English, such as how much coastal property exposure grew this quarter, instead of waiting for someone to build a new view.

ShareLinkedInEmail

Tableau Prep flows shape data from the enterprise data warehouse and data lake, including policy, claims and external catastrophe and economic data. Extracts and aggregation strategies keep queries fast across a large portfolio.

Dashboards use LOD expressions, custom geocoding for insurance territories, parameters for what-if scenarios, and filter and set actions for cross-dashboard drill-down.

Tableau Server handles scheduled daily refresh, row-level security, project-based content governance and usage monitoring, which is how adoption is tracked.

Next step

Let's build the case study we tell next.