2  Tracking Spending


You only manage what you measure. This chapter covers how to capture every dollar that leaves your accounts and turn that list of purchases into something you can learn from.

Tracking is the recording side of personal finance. In the next chapter, we’ll cover budgeting, which is the allocation side (i.e., where money is supposed to go). Without tracking, a budget is a guess; with tracking, a budget becomes a feedback loop.

%%{init: {'theme': 'neutral', 'themeVariables': { 'fontFamily': 'monospace', "darkMode":false, "fontSize":"16px"}}}%%

flowchart LR
    Tracking(["<strong>TRACKING</strong>"]) --"<em>Where money<br>actually went</em>"--> Compare{"<strong>COMPARE</strong><br/><em>plan vs.<br>reality</em>"}
    Budgeting(["<strong>BUDGETING</strong>"]) --"<em>Where money<br>should go</em>"--> Compare
    

    Compare --> Insights("Insights &<br>Adjustments")
    Insights --"<em>Feedback</em>"--> Budgeting 
    
    style Budgeting fill:#48a56a,color:#F5F2E8,stroke:#1C1C1E,stroke-width:2px
    style Tracking fill:#48a56a,color:#F5F2E8,stroke:#1C1C1E,stroke-width:2px
    style Compare fill:#f0cfcf,color:#1C1C1E,stroke:#1C1C1E,stroke-width:2px
    style Insights fill:#86ddcd,color:#1C1C1E,stroke:#1C1C1E,stroke-width:2px

Tracking vs. Budgeting

2.1 Capturing Transactions

Below are a few time-tested ways to capture your spending. If you’re reading this, you’ll also know there’s a sea of apps and digital tools for tracking your spending. It’s important to know people were tracking their spending before the internet, personal computers, and calculators. Regardless of the method you choose, the important thing is to pick one, run it for a full month, then decide whether it works for you.

A spending journal

A small notebook, kept in a pocket or bag. Every time money leaves your hands, you write one line: amount, place, category. At week’s end, you tally. This is the oldest method on the list (Benjamin Franklin kept one). Its strength is the moment of awareness it forces. Writing “$6 / Starbucks / coffee” by hand is a small pause that an automatically recorded card swipe will never give you.

Checkbook registers

Every paper checkbook used to come with a register that recorded each transaction on its own line, with a running balance. Registers work for any account where you can list money going out (debits) and money coming in (credits). One page per month is plenty. The habit of writing each entry keeps you disciplined and honest about the balance.

Cash envelope system

If you’re a fan of low-tech solutions, you can withdraw monthly amounts in cash for groceries, dining, gas and fun money, then split them into labeled physical envelopes. When an envelope is empty, that category is done for the month. Families during the Great Depression of the 1930s used envelopes because they had to. The system has stuck around because the physical, finite nature of cash slows spending more reliably than any reminder ever will.1

Bank statements

Today, we make purchases digitally (via debit and credit cards), so every transaction is already captured. The work shifts from capture to review: once a month, pull the bank and card statements, go through them line by line, and assign each transaction to a category. This is the modern equivalent of the register above, except that the bank has done much of the work.

2.2 Categorizing Transactions

Categorizing transactions can get complicated, but it doesn’t need to be. Downloading a bank statement is easy, but today, transactions might be handled by multiple services (Amazon, PayPal, Venmo, etc.). Transactions like these can be difficult to match up with what you actually bought, so I recommend starting with about 10 buckets and a small, stable list.

A working starter set:

  • Housing (rent/mortgage, utilities, insurance)

  • Transportation (car payment, fuel, maintenance, transit)

  • Food: groceries

  • Food: dining out

  • Subscriptions (everything recurring and digital)

  • Personal (clothes, haircuts, gym)

  • Health & insurance

  • Gifts & giving

  • Fun money (don’t break this down further; protect it from over-tracking)

  • Other / one-offs

Two rules that save hours of bookkeeping:

  1. When in doubt, “Other”. Don’t invent a new category for a single transaction. Review the “Other” bucket at month-end and only promote a new category once you’ve seen the same kind of transaction three months in a row.

  2. Keep the categories stable. Changing categories every month destroys the comparisons that make tracking useful in the first place. A stable set of categories will outlast any rotating set of “smart” categorizations.

2.3 Using Spreadsheets

Digital tracking is the bank-statement method above, but some of the work is done for you. Most banks allow you to export your monthly statement, and they typically include at least three columns: Date, Description, and Amount (negative for money going out, positive for money coming in).

Not all exported bank statements include a category, but you can add it yourself, and we’ll do it in the building categories section below.

The bank_statement can also be copied + pasted into Excel:

bank_statement in Excel

bank_statement in Excel

Excel also has a handy table format, which gives you many tools for sorting and analyzing data.2 I’ll go into using tables more on the accompanying website for the book.

Create the bank_statement data in Python (with polars):

show/hide Python bank_statement 
import polars as pl

py_bank_statement = pl.DataFrame({
    "Date": [
        "2026-01-02", "2026-01-02", "2026-01-03", "2026-01-03",
        "2026-01-04", "2026-01-04", "2026-01-05", "2026-01-06",
        "2026-01-07", "2026-01-08", "2026-01-09", "2026-01-10",
        "2026-01-11", "2026-01-12", "2026-01-13", "2026-01-14",
        "2026-01-15"
    ],
    "Description": [
        "Employer Payroll",     "Rent",
        "Comcast Internet",     "Trader Joe's",
        "Shell Gas",            "Netflix",
        "Starbucks",            "PG&E Electric",
        "AT&T Wireless",        "Spotify",
        "Whole Foods",          "Chipotle",
        "Geico Auto Insurance", "NYTimes Digital",
        "Shell Gas",            "Costco",
        "24 Hour Fitness"
    ],
    "Amount": [
         2400.00, -1500.00,
          -75.00,   -82.45,
          -45.20,   -15.99,
           -6.75,  -142.30,
          -85.00,   -10.99,
          -64.12,   -14.50,
         -125.00,   -17.00,
          -42.80,  -156.78,
          -39.99
    ]
}).with_columns(pl.col("Date").str.to_date())
show/hide 
py_bank_statement
shape: (17, 3)
Date Description Amount
date str f64
2026-01-02 "Employer Payroll" 2400.0
2026-01-02 "Rent" -1500.0
2026-01-03 "Comcast Internet" -75.0
2026-01-03 "Trader Joe's" -82.45
2026-01-04 "Shell Gas" -45.2
2026-01-04 "Netflix" -15.99
2026-01-05 "Starbucks" -6.75
2026-01-06 "PG&E Electric" -142.3
2026-01-07 "AT&T Wireless" -85.0
2026-01-08 "Spotify" -10.99
2026-01-09 "Whole Foods" -64.12
2026-01-10 "Chipotle" -14.5
2026-01-11 "Geico Auto Insurance" -125.0
2026-01-12 "NYTimes Digital" -17.0
2026-01-13 "Shell Gas" -42.8
2026-01-14 "Costco" -156.78
2026-01-15 "24 Hour Fitness" -39.99

Create the bank_statement data in Python (with pandas):

show/hide pandas bank_statement 
import pandas as pd

pd_bank_statement = pd.DataFrame({
    "Date": [
        "2026-01-02", "2026-01-02", "2026-01-03", "2026-01-03",
        "2026-01-04", "2026-01-04", "2026-01-05", "2026-01-06",
        "2026-01-07", "2026-01-08", "2026-01-09", "2026-01-10",
        "2026-01-11", "2026-01-12", "2026-01-13", "2026-01-14",
        "2026-01-15"
    ],
    "Description": [
        "Employer Payroll",     "Rent",
        "Comcast Internet",     "Trader Joe's",
        "Shell Gas",            "Netflix",
        "Starbucks",            "PG&E Electric",
        "AT&T Wireless",        "Spotify",
        "Whole Foods",          "Chipotle",
        "Geico Auto Insurance", "NYTimes Digital",
        "Shell Gas",            "Costco",
        "24 Hour Fitness"
    ],
    "Amount": [
         2400.00, -1500.00,
          -75.00,   -82.45,
          -45.20,   -15.99,
           -6.75,  -142.30,
          -85.00,   -10.99,
          -64.12,   -14.50,
         -125.00,   -17.00,
          -42.80,  -156.78,
          -39.99
    ]
})
pd_bank_statement["Date"] = pd.to_datetime(pd_bank_statement["Date"])
show/hide 
pd_bank_statement
#>          Date           Description   Amount
#> 0  2026-01-02      Employer Payroll  2400.00
#> 1  2026-01-02                  Rent -1500.00
#> 2  2026-01-03      Comcast Internet   -75.00
#> 3  2026-01-03          Trader Joe's   -82.45
#> 4  2026-01-04             Shell Gas   -45.20
#> 5  2026-01-04               Netflix   -15.99
#> 6  2026-01-05             Starbucks    -6.75
#> 7  2026-01-06         PG&E Electric  -142.30
#> 8  2026-01-07         AT&T Wireless   -85.00
#> 9  2026-01-08               Spotify   -10.99
#> 10 2026-01-09           Whole Foods   -64.12
#> 11 2026-01-10              Chipotle   -14.50
#> 12 2026-01-11  Geico Auto Insurance  -125.00
#> 13 2026-01-12       NYTimes Digital   -17.00
#> 14 2026-01-13             Shell Gas   -42.80
#> 15 2026-01-14                Costco  -156.78
#> 16 2026-01-15       24 Hour Fitness   -39.99

Create the bank_statement data in R:

show/hide R bank_statement 
r_bank_statement <- tibble::tibble(
  Date = as.Date(c(
    "2026-01-02", "2026-01-02", "2026-01-03", "2026-01-03",
    "2026-01-04", "2026-01-04", "2026-01-05", "2026-01-06",
    "2026-01-07", "2026-01-08", "2026-01-09", "2026-01-10",
    "2026-01-11", "2026-01-12", "2026-01-13", "2026-01-14",
    "2026-01-15"
  )),
  Description = c(
    "Employer Payroll",       "Rent",
    "Comcast Internet",       "Trader Joe's",
    "Shell Gas",              "Netflix",
    "Starbucks",              "PG&E Electric",
    "AT&T Wireless",          "Spotify",
    "Whole Foods",            "Chipotle",
    "Geico Auto Insurance",   "NYTimes Digital",
    "Shell Gas",              "Costco",
    "24 Hour Fitness"
  ),
  Amount = c(
     2400.00, -1500.00,
      -75.00,   -82.45,
      -45.20,   -15.99,
       -6.75,  -142.30,
      -85.00,   -10.99,
      -64.12,   -14.50,
     -125.00,   -17.00,
      -42.80,  -156.78,
      -39.99
  )
)

We’ll use knitr::kable() to make the output pretty:

show/hide 
r_bank_statement |> knitr::kable()
Date Description Amount
2026-01-02 Employer Payroll 2400.00
2026-01-02 Rent -1500.00
2026-01-03 Comcast Internet -75.00
2026-01-03 Trader Joe’s -82.45
2026-01-04 Shell Gas -45.20
2026-01-04 Netflix -15.99
2026-01-05 Starbucks -6.75
2026-01-06 PG&E Electric -142.30
2026-01-07 AT&T Wireless -85.00
2026-01-08 Spotify -10.99
2026-01-09 Whole Foods -64.12
2026-01-10 Chipotle -14.50
2026-01-11 Geico Auto Insurance -125.00
2026-01-12 NYTimes Digital -17.00
2026-01-13 Shell Gas -42.80
2026-01-14 Costco -156.78
2026-01-15 24 Hour Fitness -39.99

Create the bank_statement data in R (with data.table):

show/hide data.table bank_statement 
dt_bank_statement <- data.table::data.table(
  Date = as.Date(c(
    "2026-01-02", "2026-01-02", "2026-01-03", "2026-01-03",
    "2026-01-04", "2026-01-04", "2026-01-05", "2026-01-06",
    "2026-01-07", "2026-01-08", "2026-01-09", "2026-01-10",
    "2026-01-11", "2026-01-12", "2026-01-13", "2026-01-14",
    "2026-01-15"
  )),
  Description = c(
    "Employer Payroll",       "Rent",
    "Comcast Internet",       "Trader Joe's",
    "Shell Gas",              "Netflix",
    "Starbucks",              "PG&E Electric",
    "AT&T Wireless",          "Spotify",
    "Whole Foods",            "Chipotle",
    "Geico Auto Insurance",   "NYTimes Digital",
    "Shell Gas",              "Costco",
    "24 Hour Fitness"
  ),
  Amount = c(
     2400.00, -1500.00,
      -75.00,   -82.45,
      -45.20,   -15.99,
       -6.75,  -142.30,
      -85.00,   -10.99,
      -64.12,   -14.50,
     -125.00,   -17.00,
      -42.80,  -156.78,
      -39.99
  )
)
show/hide 
dt_bank_statement |> knitr::kable()
Date Description Amount
2026-01-02 Employer Payroll 2400.00
2026-01-02 Rent -1500.00
2026-01-03 Comcast Internet -75.00
2026-01-03 Trader Joe’s -82.45
2026-01-04 Shell Gas -45.20
2026-01-04 Netflix -15.99
2026-01-05 Starbucks -6.75
2026-01-06 PG&E Electric -142.30
2026-01-07 AT&T Wireless -85.00
2026-01-08 Spotify -10.99
2026-01-09 Whole Foods -64.12
2026-01-10 Chipotle -14.50
2026-01-11 Geico Auto Insurance -125.00
2026-01-12 NYTimes Digital -17.00
2026-01-13 Shell Gas -42.80
2026-01-14 Costco -156.78
2026-01-15 24 Hour Fitness -39.99

Once your transactions sit in columns and rows, it’s much easier to answer tracking questions like:

  1. What is the total spent on subscriptions?
  2. What percentage of my income is going to fixed costs?
  3. What was my biggest variable category (one that changes from month to month, like dining out) last month?

Bogle would approve of building a custom spreadsheet because simplicity wins. A well-built spreadsheet beats most apps because it forces you to engage with your numbers.

Two ideas the categorization code above relies on:

Data frames

R’s data.frame is built into the base language; the tibble is the modern replacement (used throughout this book).

Python has no built-in data frame, so you need a library; we use polars (pl.DataFrame).

df <- tibble::tibble(x = 1:5)
import polars as pl
df = pl.DataFrame({"x": [1, 2, 3, 4, 5]})

Pipes / chaining

R has a native pipe |> (also %>% from magrittr/dplyr) for left-to-right reading: data flows from one function into the next.

Python has no native pipe; the same effect comes from chaining method calls, which polars is designed for.

df |> dplyr::filter(x > 2)
df.filter(pl.col("x") > 2)

2.3.1 Building Categories

The bank_statement dataset above includes the columns typically found in a real bank export (Date, Description, and Amount). To get to a Category column, we have to map descriptions to the buckets. A small lookup of keyword patterns does the job in one pass.

In Excel, the IFS() function is the direct analog of case_when()/when().then().

  • If the statement is laid out with Date in column A, Description in column B, and Amount in column C. Type Category in cell D1, then paste the formula below into D2:
=IFS(
  ISNUMBER(SEARCH("Payroll", B2)), "Income",
  ISNUMBER(SEARCH("Rent", B2)), "Housing",
  OR(ISNUMBER(SEARCH("Comcast",B2)), ISNUMBER(SEARCH("PG&E",B2)), ISNUMBER(SEARCH("AT&T",B2))), "Utilities",
  OR(ISNUMBER(SEARCH("Trader Joe",B2)), ISNUMBER(SEARCH("Whole Foods",B2)), ISNUMBER(SEARCH("Costco",B2))), "Groceries",
  ISNUMBER(SEARCH("Shell", B2)), "Transportation",
  OR(ISNUMBER(SEARCH("Starbucks",B2)), ISNUMBER(SEARCH("Chipotle",B2))), "Dining",
  OR(ISNUMBER(SEARCH("Netflix",B2)), ISNUMBER(SEARCH("Spotify",B2)),
     ISNUMBER(SEARCH("NYTimes",B2)), ISNUMBER(SEARCH("24 Hour Fitness",B2))), "Subscriptions",
  ISNUMBER(SEARCH("Geico", B2)), "Insurance",
  TRUE, "Other"
)
  • Then copy D2 down through every transaction row (select D2, double-click the small square in the bottom-right corner, or use Ctrl+D after selecting D2:D{last_row}).

IFS() in Excel

IFS() in Excel

To add a new merchant when one shows up: edit the formula in D2, then re-fill down. Keep the new merchant near the bottom of its category group so the existing patterns still resolve first.

polars closely mirrors the dplyr code:

show/hide py categorize transactions 
py_bank_statement = py_bank_statement.with_columns(
    Category = pl.when(pl.col("Description").str.contains("Payroll")).then(pl.lit("Income"))
        .when(pl.col("Description").str.contains("Rent")).then(pl.lit("Housing"))
        .when(pl.col("Description").str.contains("Comcast|PG&E|AT&T")).then(pl.lit("Utilities"))
        .when(pl.col("Description").str.contains("Trader Joe|Whole Foods|Costco")).then(pl.lit("Groceries"))
        .when(pl.col("Description").str.contains("Shell")).then(pl.lit("Transportation"))
        .when(pl.col("Description").str.contains("Starbucks|Chipotle")).then(pl.lit("Dining"))
        .when(pl.col("Description").str.contains("Netflix|Spotify|NYTimes|24 Hour Fitness")).then(pl.lit("Subscriptions"))
        .when(pl.col("Description").str.contains("Geico")).then(pl.lit("Insurance"))
        .otherwise(pl.lit("Other"))
)

py_bank_statement
shape: (17, 4)
Date Description Amount Category
date str f64 str
2026-01-02 "Employer Payroll" 2400.0 "Income"
2026-01-02 "Rent" -1500.0 "Housing"
2026-01-03 "Comcast Internet" -75.0 "Utilities"
2026-01-03 "Trader Joe's" -82.45 "Groceries"
2026-01-04 "Shell Gas" -45.2 "Transportation"
2026-01-04 "Netflix" -15.99 "Subscriptions"
2026-01-05 "Starbucks" -6.75 "Dining"
2026-01-06 "PG&E Electric" -142.3 "Utilities"
2026-01-07 "AT&T Wireless" -85.0 "Utilities"
2026-01-08 "Spotify" -10.99 "Subscriptions"
2026-01-09 "Whole Foods" -64.12 "Groceries"
2026-01-10 "Chipotle" -14.5 "Dining"
2026-01-11 "Geico Auto Insurance" -125.0 "Insurance"
2026-01-12 "NYTimes Digital" -17.0 "Subscriptions"
2026-01-13 "Shell Gas" -42.8 "Transportation"
2026-01-14 "Costco" -156.78 "Groceries"
2026-01-15 "24 Hour Fitness" -39.99 "Subscriptions"

For use in a Shiny app, the when().then() chain above can be wrapped into a reusable function that accepts a rules table, making it easy to let users add or edit keyword patterns at runtime:

  • df = data.frame with a Description column
  • rules = data.frame with Pattern and Category columns
show/hide Python apply_rules() function 
# Apply keyword rules to categorize transactions.

def apply_rules(df, rules):
    df = df.with_columns(pl.lit("Other").alias("Category"))
    for row in rules.iter_rows(named=True):
        df = df.with_columns(
            pl.when(
                pl.col("Description").str.contains(row["Pattern"])
                & (pl.col("Category") == "Other")
            )
            .then(pl.lit(row["Category"]))
            .otherwise(pl.col("Category"))
            .alias("Category")
        )
    return df

Pass the same keyword lookup as rules and the function returns the data frame with a Category column appended:

show/hide Python apply_rules() demo 
py_rules = pl.DataFrame({
    "Pattern":  ["Payroll", "Rent", "Comcast|PG&E|AT&T",
                 "Trader Joe|Whole Foods|Costco", "Shell",
                 "Starbucks|Chipotle",
                 "Netflix|Spotify|NYTimes|24 Hour Fitness", "Geico"],
    "Category": ["Income", "Housing", "Utilities", "Groceries",
                 "Transportation", "Dining", "Subscriptions", "Insurance"]
})

apply_rules(py_bank_statement, py_rules)
shape: (17, 4)
Date Description Amount Category
date str f64 str
2026-01-02 "Employer Payroll" 2400.0 "Income"
2026-01-02 "Rent" -1500.0 "Housing"
2026-01-03 "Comcast Internet" -75.0 "Utilities"
2026-01-03 "Trader Joe's" -82.45 "Groceries"
2026-01-04 "Shell Gas" -45.2 "Transportation"
2026-01-04 "Netflix" -15.99 "Subscriptions"
2026-01-05 "Starbucks" -6.75 "Dining"
2026-01-06 "PG&E Electric" -142.3 "Utilities"
2026-01-07 "AT&T Wireless" -85.0 "Utilities"
2026-01-08 "Spotify" -10.99 "Subscriptions"
2026-01-09 "Whole Foods" -64.12 "Groceries"
2026-01-10 "Chipotle" -14.5 "Dining"
2026-01-11 "Geico Auto Insurance" -125.0 "Insurance"
2026-01-12 "NYTimes Digital" -17.0 "Subscriptions"
2026-01-13 "Shell Gas" -42.8 "Transportation"
2026-01-14 "Costco" -156.78 "Groceries"
2026-01-15 "24 Hour Fitness" -39.99 "Subscriptions"

In pandas, numpy.select() takes a list of conditions and a matching list of categories, and returns the first category whose condition is true:

show/hide pandas categorize transactions 
import numpy as np

desc = pd_bank_statement["Description"]

pd_bank_statement["Category"] = np.select(
    [
        desc.str.contains("Payroll"),
        desc.str.contains("Rent"),
        desc.str.contains("Comcast|PG&E|AT&T"),
        desc.str.contains("Trader Joe|Whole Foods|Costco"),
        desc.str.contains("Shell"),
        desc.str.contains("Starbucks|Chipotle"),
        desc.str.contains("Netflix|Spotify|NYTimes|24 Hour Fitness"),
        desc.str.contains("Geico"),
    ],
    ["Income", "Housing", "Utilities", "Groceries",
     "Transportation", "Dining", "Subscriptions", "Insurance"],
    default="Other",
)

pd_bank_statement
#>          Date           Description   Amount        Category
#> 0  2026-01-02      Employer Payroll  2400.00          Income
#> 1  2026-01-02                  Rent -1500.00         Housing
#> 2  2026-01-03      Comcast Internet   -75.00       Utilities
#> 3  2026-01-03          Trader Joe's   -82.45       Groceries
#> 4  2026-01-04             Shell Gas   -45.20  Transportation
#> 5  2026-01-04               Netflix   -15.99   Subscriptions
#> 6  2026-01-05             Starbucks    -6.75          Dining
#> 7  2026-01-06         PG&E Electric  -142.30       Utilities
#> 8  2026-01-07         AT&T Wireless   -85.00       Utilities
#> 9  2026-01-08               Spotify   -10.99   Subscriptions
#> 10 2026-01-09           Whole Foods   -64.12       Groceries
#> 11 2026-01-10              Chipotle   -14.50          Dining
#> 12 2026-01-11  Geico Auto Insurance  -125.00       Insurance
#> 13 2026-01-12       NYTimes Digital   -17.00   Subscriptions
#> 14 2026-01-13             Shell Gas   -42.80  Transportation
#> 15 2026-01-14                Costco  -156.78       Groceries
#> 16 2026-01-15       24 Hour Fitness   -39.99   Subscriptions

The same rules table approach works in pandas:

  • df = DataFrame with a Description column
  • rules = DataFrame with Pattern and Category columns
show/hide pandas apply_rules_pd() function 
# Apply keyword rules to categorize transactions.

def apply_rules_pd(df, rules):
    df = df.assign(Category="Other")
    for row in rules.itertuples():
        hits = df["Description"].str.contains(row.Pattern) & (df["Category"] == "Other")
        df.loc[hits, "Category"] = row.Category
    return df

Pass the same keyword lookup as rules and the function returns the data frame with a Category column appended:

show/hide pandas apply_rules_pd() demo 
pd_rules = pd.DataFrame({
    "Pattern":  ["Payroll", "Rent", "Comcast|PG&E|AT&T",
                 "Trader Joe|Whole Foods|Costco", "Shell",
                 "Starbucks|Chipotle",
                 "Netflix|Spotify|NYTimes|24 Hour Fitness", "Geico"],
    "Category": ["Income", "Housing", "Utilities", "Groceries",
                 "Transportation", "Dining", "Subscriptions", "Insurance"]
})

apply_rules_pd(pd_bank_statement, pd_rules)
#>          Date           Description   Amount        Category
#> 0  2026-01-02      Employer Payroll  2400.00          Income
#> 1  2026-01-02                  Rent -1500.00         Housing
#> 2  2026-01-03      Comcast Internet   -75.00       Utilities
#> 3  2026-01-03          Trader Joe's   -82.45       Groceries
#> 4  2026-01-04             Shell Gas   -45.20  Transportation
#> 5  2026-01-04               Netflix   -15.99   Subscriptions
#> 6  2026-01-05             Starbucks    -6.75          Dining
#> 7  2026-01-06         PG&E Electric  -142.30       Utilities
#> 8  2026-01-07         AT&T Wireless   -85.00       Utilities
#> 9  2026-01-08               Spotify   -10.99   Subscriptions
#> 10 2026-01-09           Whole Foods   -64.12       Groceries
#> 11 2026-01-10              Chipotle   -14.50          Dining
#> 12 2026-01-11  Geico Auto Insurance  -125.00       Insurance
#> 13 2026-01-12       NYTimes Digital   -17.00   Subscriptions
#> 14 2026-01-13             Shell Gas   -42.80  Transportation
#> 15 2026-01-14                Costco  -156.78       Groceries
#> 16 2026-01-15       24 Hour Fitness   -39.99   Subscriptions

In R, a lookup table can be created using dplyr:

show/hide R categorize transactions 
r_bank_statement <- r_bank_statement |>
  dplyr::mutate(
    Category = dplyr::case_when(
      grepl("Payroll",                                   Description) ~ "Income",
      grepl("Rent",                                      Description) ~ "Housing",
      grepl("Comcast|PG&E|AT&T",                         Description) ~ "Utilities",
      grepl("Trader Joe|Whole Foods|Costco",             Description) ~ "Groceries",
      grepl("Shell",                                     Description) ~ "Transportation",
      grepl("Starbucks|Chipotle",                        Description) ~ "Dining",
      grepl("Netflix|Spotify|NYTimes|24 Hour Fitness",   Description) ~ "Subscriptions",
      grepl("Geico",                                     Description) ~ "Insurance",
      TRUE                                                            ~ "Other"
    )
  )

r_bank_statement |> knitr::kable()
Date Description Amount Category
2026-01-02 Employer Payroll 2400.00 Income
2026-01-02 Rent -1500.00 Housing
2026-01-03 Comcast Internet -75.00 Utilities
2026-01-03 Trader Joe’s -82.45 Groceries
2026-01-04 Shell Gas -45.20 Transportation
2026-01-04 Netflix -15.99 Subscriptions
2026-01-05 Starbucks -6.75 Dining
2026-01-06 PG&E Electric -142.30 Utilities
2026-01-07 AT&T Wireless -85.00 Utilities
2026-01-08 Spotify -10.99 Subscriptions
2026-01-09 Whole Foods -64.12 Groceries
2026-01-10 Chipotle -14.50 Dining
2026-01-11 Geico Auto Insurance -125.00 Insurance
2026-01-12 NYTimes Digital -17.00 Subscriptions
2026-01-13 Shell Gas -42.80 Transportation
2026-01-14 Costco -156.78 Groceries
2026-01-15 24 Hour Fitness -39.99 Subscriptions

For repeated use, the case_when() pipeline above can be wrapped into a reusable function that accepts a lookup table, making it easy for users to add or edit keyword patterns:

  • df = data.frame with a Description column
  • rules = data.frame with Pattern and Category columns
show/hide R apply_rules() function 
# Apply keyword rules to categorize transactions
apply_rules <- function(df, rules) {
  df$Category <- "Other"
  for (i in seq_len(nrow(rules))) {
    unmatched <- df$Category == "Other"
    hits <- grepl(rules$Pattern[i], df$Description, ignore.case = TRUE) & unmatched
    df$Category[hits] <- rules$Category[i]
  }
  df
}

Create a data.frame with Pattern and Category:

show/hide R apply_rules() demo 
r_rules <- data.frame(
  Pattern  = c("Payroll", "Rent", "Comcast|PG&E|AT&T",
               "Trader Joe|Whole Foods|Costco", "Shell",
               "Starbucks|Chipotle",
               "Netflix|Spotify|NYTimes|24 Hour Fitness", "Geico"),
  Category = c("Income", "Housing", "Utilities", "Groceries",
               "Transportation", "Dining", "Subscriptions", "Insurance")
)

r_rules |> 
  knitr::kable()
Pattern Category
Payroll Income
Rent Housing
Comcast|PG&E|AT&T Utilities
Trader Joe|Whole Foods|Costco Groceries
Shell Transportation
Starbucks|Chipotle Dining
Netflix|Spotify|NYTimes|24 Hour Fitness Subscriptions
Geico Insurance

Pass the same keyword lookup as rules and the function returns the data frame with a Category column appended:

show/hide R apply_rules() demo 
apply_rules(r_bank_statement, r_rules) |>
  knitr::kable()
Date Description Amount Category
2026-01-02 Employer Payroll 2400.00 Income
2026-01-02 Rent -1500.00 Housing
2026-01-03 Comcast Internet -75.00 Utilities
2026-01-03 Trader Joe’s -82.45 Groceries
2026-01-04 Shell Gas -45.20 Transportation
2026-01-04 Netflix -15.99 Subscriptions
2026-01-05 Starbucks -6.75 Dining
2026-01-06 PG&E Electric -142.30 Utilities
2026-01-07 AT&T Wireless -85.00 Utilities
2026-01-08 Spotify -10.99 Subscriptions
2026-01-09 Whole Foods -64.12 Groceries
2026-01-10 Chipotle -14.50 Dining
2026-01-11 Geico Auto Insurance -125.00 Insurance
2026-01-12 NYTimes Digital -17.00 Subscriptions
2026-01-13 Shell Gas -42.80 Transportation
2026-01-14 Costco -156.78 Groceries
2026-01-15 24 Hour Fitness -39.99 Subscriptions

In data.table, fcase() is the analog of case_when(), and := adds the column in place:

show/hide data.table categorize transactions 
dt_bank_statement[, Category := data.table::fcase(
  grepl("Payroll",                                 Description), "Income",
  grepl("Rent",                                    Description), "Housing",
  grepl("Comcast|PG&E|AT&T",                       Description), "Utilities",
  grepl("Trader Joe|Whole Foods|Costco",           Description), "Groceries",
  grepl("Shell",                                   Description), "Transportation",
  grepl("Starbucks|Chipotle",                      Description), "Dining",
  grepl("Netflix|Spotify|NYTimes|24 Hour Fitness", Description), "Subscriptions",
  grepl("Geico",                                   Description), "Insurance",
  default = "Other"
)]

dt_bank_statement |> knitr::kable()
Date Description Amount Category
2026-01-02 Employer Payroll 2400.00 Income
2026-01-02 Rent -1500.00 Housing
2026-01-03 Comcast Internet -75.00 Utilities
2026-01-03 Trader Joe’s -82.45 Groceries
2026-01-04 Shell Gas -45.20 Transportation
2026-01-04 Netflix -15.99 Subscriptions
2026-01-05 Starbucks -6.75 Dining
2026-01-06 PG&E Electric -142.30 Utilities
2026-01-07 AT&T Wireless -85.00 Utilities
2026-01-08 Spotify -10.99 Subscriptions
2026-01-09 Whole Foods -64.12 Groceries
2026-01-10 Chipotle -14.50 Dining
2026-01-11 Geico Auto Insurance -125.00 Insurance
2026-01-12 NYTimes Digital -17.00 Subscriptions
2026-01-13 Shell Gas -42.80 Transportation
2026-01-14 Costco -156.78 Groceries
2026-01-15 24 Hour Fitness -39.99 Subscriptions

The same rules table approach works in data.table:

  • dt = data.table with a Description column
  • rules = data.table with Pattern and Category columns
show/hide data.table apply_rules_dt() function 
# Apply keyword rules to categorize transactions
apply_rules_dt <- function(dt, rules) {
  dt <- data.table::copy(dt)
  dt[, Category := "Other"]
  for (i in seq_len(nrow(rules))) {
    dt[Category == "Other" & grepl(rules$Pattern[i], Description, ignore.case = TRUE),
       Category := rules$Category[i]]
  }
  dt[]
}

Create a data.table with Pattern and Category:

show/hide data.table rules 
dt_rules <- data.table::data.table(
  Pattern  = c("Payroll", "Rent", "Comcast|PG&E|AT&T",
               "Trader Joe|Whole Foods|Costco", "Shell",
               "Starbucks|Chipotle",
               "Netflix|Spotify|NYTimes|24 Hour Fitness", "Geico"),
  Category = c("Income", "Housing", "Utilities", "Groceries",
               "Transportation", "Dining", "Subscriptions", "Insurance")
)

dt_rules |>
  knitr::kable()
Pattern Category
Payroll Income
Rent Housing
Comcast|PG&E|AT&T Utilities
Trader Joe|Whole Foods|Costco Groceries
Shell Transportation
Starbucks|Chipotle Dining
Netflix|Spotify|NYTimes|24 Hour Fitness Subscriptions
Geico Insurance

Pass the same keyword lookup as rules and the function returns a copy with a Category column appended:

show/hide data.table apply_rules_dt() demo 
apply_rules_dt(dt_bank_statement, dt_rules) |>
  knitr::kable()
Date Description Amount Category
2026-01-02 Employer Payroll 2400.00 Income
2026-01-02 Rent -1500.00 Housing
2026-01-03 Comcast Internet -75.00 Utilities
2026-01-03 Trader Joe’s -82.45 Groceries
2026-01-04 Shell Gas -45.20 Transportation
2026-01-04 Netflix -15.99 Subscriptions
2026-01-05 Starbucks -6.75 Dining
2026-01-06 PG&E Electric -142.30 Utilities
2026-01-07 AT&T Wireless -85.00 Utilities
2026-01-08 Spotify -10.99 Subscriptions
2026-01-09 Whole Foods -64.12 Groceries
2026-01-10 Chipotle -14.50 Dining
2026-01-11 Geico Auto Insurance -125.00 Insurance
2026-01-12 NYTimes Digital -17.00 Subscriptions
2026-01-13 Shell Gas -42.80 Transportation
2026-01-14 Costco -156.78 Groceries
2026-01-15 24 Hour Fitness -39.99 Subscriptions

Each month, add a new pattern when you see a real transaction that didn’t match. The goal isn’t to classify every edge case up front; the lookup handles 90%+ of your monthly transactions automatically, leaving only the unusual ones to sort out by hand.

The categorization code above uses all three of these ideas at once to create the keyword test:

  1. We use grepl() / str.contains() to return a true/false result for every row at once

  2. Then logical-OR alternation chooses between merchants

  3. The final fall-through catches anything that didn’t match

Below is a quick summary of these topics in R and Python.

Boolean values

R writes them TRUE / FALSE (sometimes abbreviated T/F).

Python writes them True / False, capitalized, no abbreviation.

flag <- TRUE
flag = True

Logical operators

R uses &, |, ! for vectorized element-wise logic (and &&, || for scalar logic).

c(TRUE, FALSE) & c(TRUE, TRUE) # vectorized
TRUE && FALSE  # scalar

Python uses and, or, not for scalar logic, and &, |, ~ for vectorized logic in numpy/polars expressions.

True and False # scalar

# vectorized:
df.filter((pl.col("a") > 0) & (pl.col("b") < 10))

Vectorization

In R, vectors are first-class, so most operations apply element-wise by default.

In Python, native list is not vectorized; for that you reach for numpy arrays or polars Series / expressions.

c(1, 2, 3) * 2
# returns c(2, 4, 6)
import polars as pl
pl.Series([1, 2, 3]) * 2
# returns Series [2, 4, 6]

2.4 The Monthly Review

Tracking is only valuable if you actually look at the totals. A monthly review is a thirty-minute habit, scheduled on the same day each month (i.e., the day after your statements close is a natural anchor).

The ritual has four steps:

  1. Pull the bank and credit-card statements for the month.
  2. Categorize every line into your buckets, using whichever capture method you’ve chosen.
  3. Compare category totals to your budget. Where did you go over budget? Where did you come in under budget?
  4. Pick one change for next month. One. The point is sustainable adjustment, not perfection.

Over a year, twelve small adjustments compound into a budget that actually fits your life. That’s the real return on tracking: the habit of checking and adjusting, not the spreadsheet itself.

2.5 Online Subscriptions

This is the most overlooked category in personal finance. Streaming, apps, software, memberships, deliveries; they’re individually small but collectively enormous and can silently drain your accounts.

%%{init: {'theme': 'neutral', 'themeVariables': { 'fontFamily': 'monospace', "darkMode":false, "fontSize":"16px"}}}%%

flowchart TD
    Subscriptions(["<strong>Monthly<br>Subscriptions</strong>"]) --> Audit("Quarterly<br>Audit")
    Audit --> Use{"Used in<br>the last 30<br>days?"}
    Use -->|"Yes"| Worth{"Worth the<br>cost?"}
    Use -->|"No"| Cancel(["Cancel<br>Today"])
    Worth -->|"Yes"| Keep(["Keep"])
    Worth -->|"No"| Cancel
    Cancel --"<em>Redirect to</em>"--> Redirect("<strong>Investments</strong>")
    
    style Subscriptions fill:#48a56a,color:#F5F2E8,stroke:#1C1C1E,stroke-width:2px
    style Audit fill:#0e9aa7,color:#F5F2E8,stroke:#1C1C1E,stroke-width:2px
    style Use fill:#f0cfcf,color:#1C1C1E,stroke:#1C1C1E,stroke-width:2px
    style Worth fill:#f0cfcf,color:#1C1C1E,stroke:#1C1C1E,stroke-width:2px
    style Cancel fill:#f44242,color:#F5F2E8,stroke:#1C1C1E,stroke-width:2px
    style Keep fill:#48a56a,color:#F5F2E8,stroke:#1C1C1E,stroke-width:2px
    style Redirect fill:#86ddcd,color:#1C1C1E,stroke:#1C1C1E,stroke-width:2px

Subscription Audit

The Subscription Reality Check

A 2022 survey by C+R Research found Americans spend about $219/month on subscriptions, and guessed $133 less than that when asked.3 At $219/month invested at a 7% real return (the return after inflation) over 30 years, that’s roughly $267,000 in lost wealth.

Action steps:

  1. Pull your last 3 months of bank/card statements
  2. List every recurring charge
  3. Cancel anything you haven’t actively used in 30 days
  4. Set a calendar reminder every 90 days to repeat this
  5. Use one card exclusively for subscriptions to track them easily

This isn’t deprivation; it’s the Sethi principle in action: cut mercilessly on what you don’t love so you can spend freely on what you do.

2.6 Math for tracking spending

The math in this chapter answers one question: what does a subscription really cost? The answer is its opportunity cost, the money you give up by spending instead of investing.

The Subscription Audit Math

A $20/month subscription feels trivial. Imagine canceling it and investing the $240 a year instead. The chart below shows what that money grows into at each checkpoint, split into the money you put in and the growth it earned.

$240 a year invested at 7% instead of a $20/month subscription

$240 a year invested at 7% instead of a $20/month subscription

$240 a year invested at 7% instead of a $20/month subscription

$240 a year invested at 7% instead of a $20/month subscription

Each bar adds another year’s $240 to everything saved so far, and the whole pile grows by 7%. Adding up every year’s $240, each with its own growth, is what the formula does in one step.

Formula: Annual cost × ((1.07)^n − 1) ÷ 0.07 = Opportunity cost in n years

  • Where: Annual cost = the monthly cost × 12, 0.07 = a 7% real yearly return, and n = the number of years.

Example: A $20/month subscription you don’t really use, over 30 years:

  • Annual cost: $240
  • Invested at a 7% real return for 30 years: $240 × ((1.07)^30 − 1) ÷ 0.07 = ~$22,700

Check: you put in $240 × 30 = $7,200, so growth added about $15,500

That “small” $20/month is a $22,700 decision. Audit accordingly.

In Excel, create a new sheet with the inputs:

Monthly cost Number of Years Annual Return (as decimal) Opportunity Cost
$20 30 0.07 Formula below

The formula (given this setup) is below:

=(A2*12)*((1+C2)^B2-1)/C2

A2*12 converts the monthly cost to annual; the rest is the future value formula for regular deposits, using the rate in C2 and the time horizon in B2.

Opportunity Cost in Excel

Opportunity Cost in Excel

Python is similar:

show/hide 
def subscription_opportunity_cost(monthly_cost, years, rate=0.07):
    annual_cost = monthly_cost * 12
    return annual_cost * ((1 + rate) ** years - 1) / rate

The same $20/month subscription for 30 years at 7%:

show/hide 
subscription_opportunity_cost(monthly_cost=20, years=30)
#> 22670.58871769843

In R, this is a simple function definition with the formula.

show/hide 
subscription_opportunity_cost <- function(monthly_cost, years, rate = 0.07) {
  annual_cost <- monthly_cost * 12
  annual_cost * ((1 + rate)^years - 1) / rate
}

Now let’s consider a $20/month subscription for 30 years at 7%:

show/hide 
subscription_opportunity_cost(monthly_cost = 20, years = 30)
#> [1] 22670.59

2.7 Recap

Tracking is the recording side of personal finance; without it, your budget is just a guess. Five rules keep it sustainable:

  1. Pick one capture method and run it for a month. Spending journal, checkbook register, cash envelopes, or pulling bank statements: any of them work. Switching every week works for nobody.

  2. Keep your categories small and stable. Six to ten buckets are plenty. Stability is what makes month-to-month comparisons meaningful; rotating “smart” categories destroys the very signal you’re trying to build.

  3. Schedule a monthly review and limit yourself to one change. Pull the statements, categorize them, compare them against your budget, and pick the single biggest leak to plug. Twelve small adjustments per year compound into a budget that actually fits your life.

  4. Audit subscriptions every quarter. They’re individually small, collectively enormous, and trivially canceled. A $20/month service you don’t really use is roughly a $22,700 decision over thirty years (see the audit math above), so Sethi’s “cut mercilessly on what you don’t love” lands hardest here.

  5. Simplicity wins. As Bogle would have it for investing, the same applies to tracking: a spreadsheet you’ll actually open beats any app you won’t.

The point of tracking isn’t a perfect ledger; it’s the feedback loop. When you can see where your money actually went, next month’s allocation in Budgeting becomes a decision rather than a guess.

  1. Inheritance hide-and-seek. Chicago Tribune. 2005. Stashing cash around the house was actually quite common for people who lived through the depression.↩︎

  2. Read the Overview of Excel tables on the Microsoft documentation.↩︎

  3. Survey Finds Americans Spend $133 More Than They Realize on Monthly Subscription Fees. The Motley Fool. 2022. The survey of 1,000 US adults was run by C+R Research. Similar numbers appear on Next Gen Personal Finance and CBS News.↩︎