Your book, measured as it moves.
Define a measure once as a view of your book. It maintains itself from there. Written premium, loss ratio, concentration, and the ranges that flag when any of them drifts.
A metric is a view, not a report. Point one at your book, say what it divides by, and it maintains itself from there.
Name it, then shape it.
A metric is flat or ratio, and rolls up as a sum or a weighted mean. Four combinations cover most of a portfolio plan.
Every leg is a view.
Edit each side in place: filter it, hop out to related records, and pick the column it aggregates. A ratio is two of these and one number.
Choose how it slices.
One column is the value it aggregates. Another partitions the result, so the same metric breaks out by region, class code or line.
Cut it into a release.
Metrics are drafted, released and pinned like every other piece of configuration. The portfolio screen reads the released one.
“Every team had its own written premium number. Now there is one definition, it is versioned, and nobody rebuilds it on a Monday.”
The measures a book is watched by. Four shapes cover most of a portfolio plan, and each one is the same machinery pointed at a different view.
Written premium.
Sum the premium on every policy in a bound state, broken out by line, class code or region. The total moves when the next policy binds, not when someone rebuilds the report.
Loss ratio.
A ratio metric divides one view by another. Incurred loss over earned premium, summed on both legs inside the same slice, then divided once.
Concentration.
Share of insured value sitting inside one region or one postcode. The stack shows which buckets are carrying it, and what is still in the pipeline behind them.
Pipeline and bound.
Every reading splits by the state a policy sits in, and states carry roles: pipeline, won, lost, declined, expired, cancelled. One metric answers what you wrote and what you are still working.
Guidelines that flag themselves. A measure becomes a guideline the moment something happens when it moves. Give a reading the ranges you underwrite to and it starts doing that work.
Ranges you set
Give a reading the bands you underwrite to. Each band carries its own score, and a value outside every band falls to neutral.
Referral or knockout
A value landing in a flagged range records a referral or a knockout. It is recorded against the risk, never quietly folded into the score.
The book, on the risk
A book-level reading can be the input behind an appetite score, so a bucket you have already filled shows up on the next quote that would fill it further.
From a number to what moved it. A reading that surprises you is worth nothing until you can see what is behind it.
The rollup.
One value per slice, composed the way the metric was defined: a sum for a flat measure, a weighted mean for an average, a real division for a ratio. The numerator and denominator totals sit next to it.
The breakdown.
The same slice split two ways at once, by the kind of record and by the state it sits in. Ratios do not stack, so the stacks are drawn on the additive parts underneath them.
The filters.
Narrow to one object, one status or one partition and every tile and every stack recomputes against that slice alone. Nothing is re-exported and nothing is rebuilt by hand.
One graph, measured end to end.
The numbers on this page are the records the rest of the platform writes. Here is where they come from.
Bring your portfolio plan.
Send us the guidelines you review each quarter and the report you rebuild to check them. We will define them as metrics on the call.
Usually 30 minutes, with an underwriter on the call.