group-expense-settler

Group Expense Settler 👥💰

You went on a trip with five friends. Everyone paid for something — hotels, gas, dinners, tickets. Now there's a tangle of "I paid for you there, you paid for me here" that nobody can untangle. This skill reduces the whole mess to the fewest possible payments that make everyone whole.

Overview

Settling shared expenses has two separate problems, and most people conflate them:

1. Fairness — how much should each person have contributed? (Split a dinner 5 ways; the grocery run only 3 ways; the rental car by days present.)
2. Settlement — given what everyone actually paid vs. their fair share, what's the cheapest set of transfers to square everything?

For #2 the naive approach ("everyone who underpaid pays the person who overpaid the most") generates O(n²) transfers. The min-cash-flow algorithm — greedy matching of the largest creditor against the largest debtor — provably needs at most n−1 transfers and typically far fewer. For a 6-person trip: 15 possible payment edges → usually 4 or 5 actual transfers.

scripts/settle_up.py implements:

• Ledger format: payer,amount,people... (+ optional --weights or per-item exclusions)
• Net-position math: for each person, paid − fair_share = net. Positive net = is owed money; negative = owes.
• Min-cash-flow settlement (max-creditor ↔ max-debtor greedy, exact to the cent, no rounding residue)
• Fairness audit table: paid / share / net per person
• Tolerance handling (± a cent) and "everyone's square within X" check
• JSON export for receipts and chat-pasteable settlement lines

When to Use

• Settling after a group trip / weekend cabin / festival
• Monthly roommate settle-up: rent is uneven, utilities split evenly, someone bought shared groceries
• Any "we really don't need another app for this" moment — this runs offline from a terminal with a 3-line ledger
• Recurring clubs: gaming group, sports team dues, communal equipment

Don't use for: continuous running balances you want synced with a bank (Splitwise et al.), or formal bookkeeping/double-entry accounting.

How It Works — Steps

1. Write the ledger — one line per expense: who paid, total, who shares it:

   Ana,450,Ana Ben Cho Dan      # hotel, 4-way
   Ben,120,Ben Cho               # taxi, 2-way
   Cho,80,Ana Ben Cho Dan        # groceries all
   

2. Run:

   python3 scripts/settle_up.py --ledger trip.txt
   

3. Read the net table (who's owed / owes what) and the settlement plan — minimum transfers, each printed as a chat-ready sentence.
4. Weighted splits (room sizes, income shares, days attended):

   python3 scripts/settle_up.py --ledger rent.txt --weights "Ana:2,Ben:1,Cho:1"
   

5. Audit with --show-items to see every person's share of every line — settle arguments with arithmetic, not vibes.

Algorithm (min-cash-flow, exact)

1. Compute each person's net = total paid − total fair share (integer cents).
2. Split into creditors (net > 0) and debtors (net < 0). Sum of both sides is equal — money is conserved.
3. Repeat: take max creditor and max debtor. Transfer min(creditor.net, −debtor.net). One of them reaches exactly 0 and leaves the pool. Append the transfer.
4. Stop when all nets are 0. Transfers ≤ n−1, and the greedy choice maximizes the chance that each transfer zeroes someone out — the well-known heuristic used in payment-netting systems.
5. Cent-exactness: everything is integer cents from input onward; the split of an expense among k people distributes the remainder cents deterministically (earlier-listed people get the extra cent), so nets always sum to zero with no leftover dust.

Worked Example

Ledger:

Ana,450,Ana Ben Cho Dan
Ben,120,Ben Cho
Cho,80,Ana Ben Cho Dan
Dan,60,Ben Cho Dan

Nets: Ana paid 450 / share 132.50 → +317.50 · Ben paid 120 / share 212.50 → −92.50 · Cho paid 80 / share 212.50 → −132.50 · Dan paid 60 / share 152.50 → −92.50. (Check: 317.50 − 92.50 − 132.50 − 92.50 = 0 ✓)

Settlement (3 transfers, not up to 6):

Ben  pays Ana  $92.50
Cho  pays Ana  $132.50
Dan  pays Ana  $92.50

Everyone else is done — no "you pay me and I pay her" chains.

Common Pitfalls

1. Including the payer in the split list. Ana,450,Ana Ben Cho Dan splits among all four including Ana — usually right. If Ana paid but isn't consuming, list only the consumers: Ana,450,Ben Cho Dan.
2. Currency commas vs CSV commas. Write 1200.50 not 1,200,50 — the ledger is comma-separated.
3. Rounding debt. The script is cent-exact; if you hand-round each person's share to whole dollars, nets won't sum to zero and someone will be a few cents short. Keep cents.
4. Mixing currencies in one ledger. All lines must be one currency. Convert first, note the rate.
5. Trust but audit. Run --show-items before announcing numbers; a typo in one amount silently shifts everyone's net.
6. Weights must cover everyone in the ledger. A person missing from --weights gets weight 1 silently — check the printed weight table.

Verification Checklist

• Net column sums to exactly 0 (script asserts this internally)
• Number of transfers ≤ people − 1
• Every person appears at most once as payer in settlements (a person should never both pay and receive — script avoids this by design)
• --show-items totals match your receipts
• Same currency across all ledger lines

One-Shot Recipes

Trip settle-up with per-line participants:

python3 scripts/settle_up.py --ledger trip.txt --show-items

Roommates — rent weighted by room, utilities even:

Ana,2400,Ana Ben Cho
utility,180,Ana Ben Cho

python3 scripts/settle_up.py --weights "Ana:2,Ben:1,Cho:1" --ledger rent.txt

Paste-ready plan for the group chat:

python3 scripts/settle_up.py --ledger trip.txt --json | python3 -c \
  "import json,sys; [print(t['line']) for t in json.load(sys.stdin)['transfers']]"