ABC Classification • fpaR

Introduction

In business and financial analytics, the ABC classification method is widely used to categorize items, customers, or other groups according to their relative contribution to a total metric.
The abc() function provides an easy way to classify a group’s member by their relative contribution of either their transaction counts or the sum of a numeric variable such as sales margin or revenue.

Allows for custom break points (e.g., top 10% = A, next 40% = B, etc.)
Works on both tibbles and database objects
Returns a segment object, which is processed by calculate() to produce a table of results

How It Works

abc() requires a grouped tibble or lazy DBI object using dplyr::group_by() to specify the group composition that drives the contribution

Value Capture

If .value is provided, then that column is aggregate per group member; otherwise, it counts rows

Category Values

Provide the break points that are used to set the cumulative categories
Each break point will get a letter category starting with ‘A’
If you want to see the stores that make up the top 40% of revenue follow by top 70% and then 90% you should put in c(0.4,.7,.9,1)

Table 1

Argument	Description
.data	A grouped tibble or DBI object (using dplyr::group_by())
category_values	A numeric vector of breakpoints between 0 and 1, representing cumulative proportions for ABC categories.
.value	Optional. A column to sum for categorization. If not provided, the function counts the number of rows per group.

# Example

contoso::sales |> 
   dplyr::group_by(store_key) |> 
   fpaR::abc(
      category_values = c(0.4,.7,.9,1), 
      .value = margin
      )

── ABC Classification ──────────────────────────────────────────────────────────

Function: `ABC` was executed

── Description: ──

This calculates a rolling cumulative distribution of variable and segments each
group member's contribution by the break points provided. Helpful to know which
group member's proportational contribution to the total.

── Category Information ──


• The data set is grouped by the margin and segments each group member by their first NA entry to define their cohort margin
• This creates cohort ID that each member is assigned to eg; January 2020, February 2020, etc
• The distinct count of each margin member in the cohort is then tracked over time

── Calendar: ──

• The calendar aggregated NA to the day time unit
• A NA calendar is created with 1 groups
• Calendar ranges from 0 to 0
• 0 days were missing and replaced with 0
• New date column NA was created from NA

── Actions: ──

✔Aggregate

✖Shift

✖Compare

✔Proportion Of Total

✖Count Distinct

── Next Steps: ──

• Use `calculate()` to return the results

────────────────────────────────────────────────────────────────────────────────

The function returns a segment object

This will summarize the series of actions and meta information about your data
Use calculate() to generate the ABC classification table
This will return a duckdb DBI object use dplyr::collect() to return a tibble

contoso::sales |> 
   dplyr::group_by(store_key) |> 
   fpaR::abc(
      category_values = c(0.4,.7,.9,1), 
      .value = margin
      ) |> 
   fpaR::calculate()

Table 2

store_key	abc_margin	cum_sum	prop_total	cum_prop_total	row_id	max_row_id	cum_unit_prop	category_value	category_name
540	78124.11	78124.11	0.04102100	0.04102100	1	57	0.01754386	0.4	a
610	74603.84	152727.95	0.03917259	0.08019359	2	57	0.03508772	0.4	a
510	69455.50	222183.44	0.03646933	0.11666292	3	57	0.05263158	0.4	a
80	67880.51	290063.96	0.03564234	0.15230526	4	57	0.07017544	0.4	a
270	56722.52	346786.47	0.02978356	0.18208882	5	57	0.08771930	0.4	a
440	56500.03	403286.51	0.02966674	0.21175556	6	57	0.10526316	0.4	a
450	54941.68	458228.19	0.02884849	0.24060405	7	57	0.12280702	0.4	a
550	53874.43	512102.61	0.02828810	0.26889216	8	57	0.14035088	0.4	a
490	51347.82	563450.44	0.02696145	0.29585360	9	57	0.15789474	0.4	a
650	49467.08	612917.52	0.02597392	0.32182752	10	57	0.17543860	0.4	a

This table contains grouped data with various metrics, highlighting the contribution of each group in terms of both value and transaction count. Below is an explanation of the key columns and how to interpret the results:

Understanding the Results

Store 540 has a margin of $7,812.11 (“ABC Margin”), which accounts for about 4% (“prop_total”) of the total margin across all stores
The “cum_sum” column tracks the running total of values (e.g., revenue or count) for each store, showing the cumulative sum up to that row
The “cum_prop_total” column shows each store’s contribution as a percentage of the total margin as you move down the table
The store with the highest contribution has a “row_id” of 1 and is assigned to the first category segment (‘A’) via the “category_name” column
The “max_row_id” shows that there are 57 additional stores in the same category (‘A’)
The “cum_unit_prop” column tracks the cumulative contribution from a transaction count perspective, similar to cum_prop_total but at the unit level
The category_value and category_name columns define the breakpoints you provided, assigning stores to categories (e.g., ‘A’, ‘B’, ‘C’) based on their cumulative contribution

This is summarized in Table 3 below:

Table 3

Column_Name	Description	Example_Values
cum_sum	The cumulative sum of the specified values (e.g., revenue, count, etc.), aggregated per group. Represents the total value up to that row.	1000, 2500, 4000
prop_total	The proportion of the total for each row's value. Shows the percentage of the total represented by the current row's contribution.	0.10, 0.25, 0.40
cum_prop_total	The cumulative proportion of the total, showing the running total percentage of the entire dataset as you move through the rows.	0.10, 0.35, 0.75
row_id	The unique identifier for the row, often used to track or identify specific rows in the dataset. Typically sequential ID or index.	1, 2, 3
max_row_id	The maximum row ID in the current group (if grouping is applied), representing the total number of rows in the group.	5, 5, 5
cum_unit_prop	The cumulative proportion of the unit values, similar to cum_prop_total, but typically used when the unit is treated as aggregate.	0.10, 0.30, 0.70
category_value	The category value that corresponds to the cumulative proportion break points (e.g., top 10%, top 40%, etc.). Based on the break points provided.	0.4, 0.7, 0.9
category_name	The name of the category assigned to each row based on the cumulative contribution. Categories are represented by letters (A, B, C, etc.).	"A", "B", "C"