Ah I see, I think I understand.

So these column normalised, meaning columns sum to 100%. So the heatmap is telling you about the wider market and your performance within it.

#### Toy example with 10K policies

Let’s take a toy example where there are a total of 10K policies in the whole dataset, and and we only have two columns, the `None`

(A) and the `(0, 5K]`

(B). Further let’s say they’re broken down such that.

**Column A.** 9K contracts

**Column B.** 1K contracts

So what do the numbers in your heatmap mean?

#### How many policies does each row represent?

First let’s see how many policies you win with any frequency:

**Column A.** We know that 91.32% of these you never win. That means you win 8.68% of them in some situation. That’s 0.0868 x 9000 = 7821 policies. Split equally among the three rows that’s ~ 2600 policies per cell (excluding never)

**Column B.** Similarly for this column you will sometimes win 0.0489 x 1000 = 489 policies. That’s ~160 per cell

#### So what do the percentage numbers mean?

This heatmat is telling you how exposed you are to the entire market with the view on claim amounts.

Ok so what does it mean to say for the column A policies, you have 3.5% in the top cell? That means that the policies that you win often and don’t make a claim (~2600 policies), represent 3.5% of the entire set of policies that don’t make a claim in the market.

Similarly in column B, the top row with 2.3% means, that the policies you win often and have a claim less than 5K (~160 policies) represent 2.3% of all the policies that have made a claim in that range in the market.

I hope that makes it clearer (and that my maths is correct!)