Skip to content

Banding

You have a continuous value like driver age or sum insured, but your rating structure needs age bands or value brackets. The Banding node turns a column of raw values into a column of bands, using rules you define. It also works with categorical values - grouping many fuel types into "Standard" vs "Green", for example.

When to use

  • Converting continuous values (age, mileage, sum insured) into discrete bands for your rating tables.
  • Grouping categorical values into broader categories.
  • Preparing inputs for a Rating Step that expects banded values.

A Banding node takes a single input and outputs it with one new banded column per factor. One node can band several columns. The panel has two tabs: CONFIG and COLUMNS.

The CONFIG tab

The INPUT chip at the top names the node's connection; its × removes the connection. Next to it, a line says which rows the counts and charts describe:

  • All rows · and the row count, when the node's input data is cached: every count below is for the whole dataset.
  • "Not cached · Refresh this node to count all rows" or "Cached data is out of date · Refresh this node to count all rows" until you refresh the node; "Caching the data…" and "Counting…" while that happens.

The factor list

Each banded column is a factor. The list shows one row per factor, named by its output column (or its input column until it has one), with a dot (green when complete, amber when not), the number of rules, and a bin to remove it (when there is more than one). Hover a row for what an incomplete factor still needs: "No input column", "No output column" or "No rules yet". Click a row to edit that factor below the list.

  • The search box finds a factor by its output or input column; All and Issues (with the number of incomplete factors) filter the list.
  • The + button beside the filter adds a factor, starting as Numeric.
  • Drag a row onto another, or use Alt+Up and Alt+Down, to reorder: the factors are applied in list order.

Editing a factor

Field What it does
TYPE Numeric bands numbers and dates by breakpoints; Categorical maps exact values to groups. Numeric is unavailable for a column that is neither a number nor a date (its tooltip names the column's type). Switching type puts the current rules aside, and switching back restores them until the pipeline is reloaded.
INPUT COLUMN The column to band, chosen from the input's columns (each with its type). Choosing a column sets TYPE to match it - Numeric for a number or date, otherwise Categorical, clearing the rules if the type changes - and fills in OUTPUT COLUMN as the column name with _band added, unless you have typed your own.
OUTPUT COLUMN The name of the new banded column.
DEFAULT (unmatched rows) The band given to rows that match no rule, with the number of such rows beside it once the data is counted ("12 of 5000 rows"). Leave it empty to give them a missing value.

With the data counted, a Numeric factor shows a histogram of the whole dataset with the band boundaries drawn over it.

Numeric breakpoints

A Numeric factor with no breakpoints offers two ways to start:

  • Generate even bands asks for a Start, End and Step and fills in evenly spaced breakpoints, each labelled by its range (such as 18–25); the last band ends at End, so it may be shorter. On a date column, Start and End are dates and Step is a number of days, weeks, months or years. The fields start from the data's lowest and highest values when they are known. Mistakes are reported in the box, such as "End must be greater than start".
  • Add manually adds one empty breakpoint.

After that, BREAKPOINTS (N) lists them, with Generate to replace them with generated ones (its fields start from the settings the current breakpoints imply). Each row has:

Column What it does
Up to (incl.) The band's upper boundary. The band includes the boundary itself, so 25 covers everything up to and including 25. Leave the last row's boundary empty for an open-ended final band.
Band name The label rows in this band get.
Matches How many rows fall in the band, once the data is counted.
  • Add adds a breakpoint, and the bin deletes one.
  • A boundary out of order is shaded with a warning, such as "Breakpoint 3 is out of order; enter a value greater than 45."
  • The copy button at the foot of the table (Copy banding as TSV) copies the breakpoints as tab-separated text with the headers Up to and Band name. Pasting tab-separated rows into a cell fills the rows from that cell down.

Dates work the same way. On a Date or Datetime column, write each boundary as a date (YYYY-MM-DD) or a date and time (YYYY-MM-DD HH:MM); every boundary in one factor must be the same kind, and the table warns about a boundary that is not a date or is of the other kind. A date covers its whole day, even on a Datetime column, so 2024-03-31 includes everything up to midnight at the end of 31 March. Times are read in the column's own time zone. This suits rules that change on a date, such as a rate change for policies starting from 1 April.

Categorical rules

A Categorical factor shows the column's Available values as chips - every value in the data, with its row count, once the data is counted, or the values in the preview until then ("Connect data to see values" when there are none). With more than ten values, Filter values... narrows the chips. Click a chip to add a rule for that value; values already used are greyed.

RULES (N) lists the rules, with Add to add an empty one:

Column What it does
Value A value of the input column, matched exactly.
Maps To The group that value is given. Pressing Enter in the last row's Maps To adds a new rule.
Matches How many rows have the value, once the data is counted.
  • The bin deletes a rule.
  • A value listed twice is flagged, such as "Duplicate value "Petrol" in rules 1, 3".
  • Copy banding as TSV copies the rules with the headers Value and Maps To; pasting tab-separated rows adds them as rules.

The COLUMNS tab

The COLUMNS tab chooses which columns the node passes on (see Working with any node).

Example

A node that bands driver_age and groups fuel_type:

  1. Connect your data to a Banding node and choose driver_age as the INPUT COLUMN. TYPE switches to Numeric and OUTPUT COLUMN becomes driver_age_band; change it to age_band.
  2. Click Add manually, then Add twice, and fill in the breakpoints: Up to 25 → 18-25, 65 → 26-65, and an empty Up to → 65+. Type Unknown as the DEFAULT.
  3. Click the + button to add a second factor, choose fuel_type as the INPUT COLUMN (TYPE becomes Categorical) and change the OUTPUT COLUMN to fuel_band.
  4. Click the chips Petrol, Diesel and Electric under Available values, and set their Maps To to Standard, Standard and Green. Type Other as the DEFAULT.

Before and after:

BEFORE                              AFTER
| driver_age | fuel_type |          | driver_age | fuel_type | age_band | fuel_band |
|------------|-----------|          |------------|-----------|----------|-----------|
| 22         | Petrol    |    →     | 22         | Petrol    | 18-25    | Standard  |
| 45         | Electric  |          | 45         | Electric  | 26-65    | Green     |
| 71         | Diesel    |          | 71         | Diesel    | 65+      | Standard  |

For a rate change from 1 April 2024, band a start_date column with Up to 2024-03-31 → Before April 2024 and an empty Up to → From April 2024.

What falls to the default

Rows with a missing value get the DEFAULT, as do categorical values you haven't listed. Without an open-ended final breakpoint, values above the last boundary get it too.

Every rule needs a name

Saving refuses a breakpoint without a Band name, a categorical rule without a Value or Maps To, and a boundary or value used twice in one factor, with a message naming the rule.

In the pipeline file

The node's settings are stored in a JSON sidecar, config/banding/<node name>.json, which the node's decorator in the pipeline's .py file names: @pipeline.banding(config="config/banding/<node name>.json"). The sidecar holds factors, a list with one object per factor.

Setting in the editor Stored as
TYPE factors[].banding: "breakpoints" (Numeric) or "categorical" (Categorical)
INPUT COLUMN factors[].column
OUTPUT COLUMN factors[].outputColumn
DEFAULT factors[].default (null when empty)
BREAKPOINTS factors[].rules: a map from each Up to boundary to its Band name, in order, with "" as the key of an open-ended final band
RULES factors[].rules: a map from each Value to its Maps To group
COLUMNS tab selected_columns

With no editor control: factors[].rightClosed, for breakpoints. true (the default) makes each band include its upper boundary; false makes it include its lower one. The table header shows it as (incl.) or (excl.), but only the sidecar can change it.

{
  "factors": [{
    "banding": "breakpoints",
    "column": "driver_age",
    "outputColumn": "age_band",
    "rules": { "25": "18-25", "65": "26-65", "": "65+" },
    "default": "Unknown"
  }, {
    "banding": "categorical",
    "column": "fuel_type",
    "outputColumn": "fuel_band",
    "rules": { "Petrol": "Standard", "Diesel": "Standard", "Electric": "Green" },
    "default": "Other"
  }]
}

See also: