> ## Documentation Index
> Fetch the complete documentation index at: https://docs.creditbenchmark.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Consensus Probability of Default (PD)

> The raw Consensus PD is mapped to the CCR100 scale for finer-grained credit analysis, providing 101 buckets versus the standard 21 letter grades.

The [Consensus Credit Rating](/methodology/consensus-credit-rating) maps a raw average **Probability of Default (PD)** to one of 21 letter-grade buckets. That's the right format for most use cases — but 21 categories can be too coarse when you need to detect small movements within a grade, or rank entities that sit inside the same CCR bucket.

The secondary `CCR100` scale addresses this. It preserves more of the underlying PD signal by mapping the same raw Consensus PD into 101 narrower buckets instead of 21.

## CCR100 Publication

Publishing a CCR100 value is a two-step process. Step 1 is covered in full on the [Consensus Credit Rating](/methodology/consensus-credit-rating) page — this page focuses on Step 2.

<Steps>
  <Step title="Raw PDs → Consensus PD Average">
    Contributed TTC PDs are averaged across all banks:

    $$
    \text{Consensus PD} = \frac{1}{N} \sum_{i=1}^{N} PD_i
    $$
  </Step>

  <Step title="Consensus PD Average → CCR100 Midpoint PD">
    That average is passed through a separate `CCR100` lookup table:
  </Step>
</Steps>

$$
\text{CCR100 PD} = f(\text{Consensus PD})
$$

Buckets are indexed 1–101, where 101 corresponds to default.

<Info>
  Because this is a table lookup, the published midpoint PD is discrete rather than continuous. In practice it is usually very close to the raw average — but not always exactly equal.
</Info>

## When to use CCR100

Use the headline CCR for credit classification, reporting, and any context where a letter grade is the right output. Use CCR100 when you need to:

* **Rank entities within the same CCR bucket** — two entities both rated `bbb` may sit at meaningfully different positions within that grade
* **Track small movements** — a shift that doesn't cross a CCR threshold will still show up in the CCR100 bucket
* **Feed quantitative models** — the underlying average PD and CCR100 midpoint are more suitable continuous inputs than a categorical letter grade

For field-level definitions, see the [Data Dictionary](/data/data-dictionary).