New vs. Renewal: why the system can’t tell them apart

Findings from the production data · for business stakeholders

We cannot reliably know whether a policy is a new policy or a renewal. The field exists, but it is almost never asked, often left blank, and cannot be reconstructed from the data we already have.

What the data shows

How today’s active policies are tagged — New, Renewal, or left blank:

Fire insurance (Incêndio)18,300 policies
79% New
20% blank

Only 1% marked as renewal — renewals are essentially never recorded here.

Rental surety (Fiança)5,100 policies
42% New
27% Renewal
31% blank
Capitalization bonds (CAP)2,700 policies
59% New
36% Renewal
5%
New Renewal Blank / not filled in

Why this happens

Why we can’t fix this from old data

We tried to detect renewals automatically — matching the same tenant at the same property over time. It doesn’t work:

In short: the answer isn’t hiding in the data. It was never captured in the first place.

What we recommend

Ask the question when the policy is created. Make “New or Renewal?” a required choice at the moment a policy is issued. The person issuing it always knows the answer.

We can’t recover the past — but we can capture it correctly from now on. No algorithm can reconstruct it after the fact; a person at the point of sale can.

Why it matters: this New-vs-Renewal field decides which commission rule applies. When it’s wrong or blank, the wrong rate can be used — affecting what agencies and TDA are paid, and making renewal reporting unreliable.