When the cheaper order is barely cheaper

Knowing which order is cheaper says nothing about by how much, and by how much is what the decision actually turns on. This study asks the more decidable question: for portfolios shaped a given way, what is the choice worth?

It reads the same corpus as the order-comparison study, so both pages describe the same population rather than separate draws of it, and it segments that corpus by the spread between the dearest and the cheapest rate in each portfolio.

What the data shows

• Narrow rate spreads make the choice nearly free: a median $6.72 over the 1,209 portfolios in the two narrowest bands combined, with 70.31% costing under fifty dollars more. The table below reports each band on its own.
• Wide rate spreads are where the choice costs real money: a median $368.54 over the 6,230 portfolios in the two widest bands combined, and only 22.63% under fifty dollars.
• When the smallest balance is also the dearest, both orders open on the same debt — and in 58.39% of those 3,307 portfolios the two orders cost exactly the same to the cent. They can still diverge on the debts that follow, and in the rest they did.
• Everywhere else the median cost was $330.83, against $0.00 in that group.

The dataset

Showing 5 of 5 rows. The complete dataset — every row and every column — is in the CSV.

Download the full dataset (CSV) 5 rows · 8 columns

The table shows the dataset's own rows, formatted as this site formats money and terms. The download carries the same rows unformatted, which is what a spreadsheet wants.

What was varied

Portfolios drawn

10,000

Debts per portfolio

2 to 5, drawn uniformly

Balance per debt

$300.00 to $25,000.00, drawn log-uniform

Rate per debt

no interest (0%), weight 1 · instalment-rate (4.5% to 11.99%), weight 2 · lower-rate card (12.99% to 18.99%), weight 3 · ordinary card (19.99% to 25.99%), weight 4 · high-rate card (26% to 29.99%), weight 2 — weights are relative, and are a modelling choice rather than a measurement of the market

Generator

mulberry32, seeded once per study and drawn in a fixed order: debt count, then for each debt its balance, its rate band, and — only when that band spans a range — its rate within it, then the extra-payment share. ⛔ The no-interest band is a single value, so it consumes no draw; a reader who always draws a rate will diverge from this corpus. Inserting or removing a draw anywhere changes every portfolio after it, which is why the order is part of the method and not an implementation detail.

Extra each month

5% to 60% of the portfolio's total minimums

Segmentation

Bands over the gap between each portfolio's highest and lowest rate, upper-inclusive: every debt at the same rate (exactly 0%) · rates within a few points (0% to 5%) · a moderate spread of rates (5% to 12%) · a wide spread of rates (12% to 20%) · a very wide spread of rates (20% to above)

Cheap threshold

$50.00, decided at the precision a reader sees

Seed

20,260,917

What was held constant

Minimum payment (held constant)

Every debt's minimum is this month's interest plus 1% of the balance, with a $25.00 floor — a shape that always covers the interest, so every debt clears on its own minimum and the result is about the payoff ORDER rather than about the minimum. It is computed on the opening balance and held there for the term, which is the engine's model of a minimum payment. A flat percentage-of-balance minimum would not amortize above a certain rate and would make the ordering choice look far more expensive than it is.

Behaviour

Every payment lands on time, no new balance is added, no rate moves, no fee is charged and no promotional rate expires.

Same corpus

The same portfolios, the same seed and the same engine as the order-comparison study — re-read rather than redrawn, so both pages describe one population.

Plan anchor

January 2026 — pinned, so a rerun on any day produces identical output. These studies report terms and interest, never dates, so the anchor changes no published figure.

Data

Synthetic throughout. No customer data was used, approximated or fitted, and nothing here is a claim about what borrowers actually owe.

Engine version

payoff-25e86113b543

What was excluded, and how much

• 0 excluded — the payment does not clear the balance under one or both orders, so there is no pair of terms to compare.

How this was computed

The corpus, the seed and the engine are identical to the order-comparison study — the same portfolios, re-read rather than redrawn. Each portfolio was placed in a band by the gap between its highest and lowest rate, and the cost of the smallest-balance order was summarized within each band.

Whether a portfolio cost nothing extra is decided at the precision a reader sees, on rounded dollars and cents, not on a difference below the smallest unit anyone is shown.

What the model assumes

Everything the order-comparison study assumes, because it is the same run: constant payments, no new balances, no rate movement, no fees, and minimums held at their opening amount.

The bands are a presentation choice. The underlying spread is a continuous number and it is in the dataset, so a reader who prefers different cut points can make them.

What this does not show

It does not tell you which order to choose. It tells you how much the choice is worth for portfolios shaped like each band, which is the input to that decision rather than the decision.

A share within a band is a share of that band, not of the corpus. The band sizes are printed beside the shares for exactly that reason.

Synthetic data, and why

No customer data was used. This page reads the same published, regenerable corpus as the order-comparison study; the dataset below is the band summary, and the portfolio-level rows are in that study's own download.

Related

• What the snowball order actually costs
• Debt snowball vs avalanche
• Debt snowball calculator
• Methodology