Choosing Between an On-Site ATM and a White-Label Tie-Up
Your rural branch has to keep cash accessible but the premises are rent-free and the regional office wants a cost decision before the financial year closes. Two options are on the table, both offering the same service to the customer.
Option A - own on-site ATM: fixed monthly cost of ₹21,000 (depreciation plus annual maintenance contract) and ₹3 per transaction for cash replenishment and reconciliation.
Option B - white-label ATM run by a vendor: no capital outlay; the bank pays the vendor ₹7 per transaction. The contract carries a minimum guarantee of 3,500 transactions a month, below which the bank pays for 3,500 anyway.
Current footfall is 4,200 transactions a month. The branch's projections are flat for the next two years. Your cash officer would also have to spend 20 minutes a day on reconciliation under Option A, which is 22 working days at a loaded cost of ₹240 an hour. A colleague argues that "owning the machine always works out cheaper in the long run".
1. Compute the break-even volume and decide which option is cheaper at 4,200 transactions.
Read the worked answer
Set the cost functions.
- Option A: C(A) = 21,000 + 3x
- Option B: C(B) = 7x (the minimum guarantee is not binding above 3,500)
Break-even: 21,000 + 3x = 7x → 21,000 = 4x → x = 5,250 transactions a month, i.e. 175 a day.
At 4,200:
- Option A = 21,000 + (3 × 4,200) = 21,000 + 12,600 = ₹33,600
- Option B = 7 × 4,200 = ₹29,400
Decision: Option B is cheaper by ₹4,200 a month (₹50,400 a year). The colleague's instinct - that owning beats renting over the long run - is only true beyond 5,250 transactions, which is 25% above current footfall. With flat projections, that threshold is not in sight, so the "long run" argument does not apply on these numbers.
2. Bring in the reconciliation time and the minimum guarantee, and re-test the decision at current volume and at a 20% fall in footfall.
Read the worked answer
Add the hidden cost to Option A. 20 minutes a day × 22 days = 7.33 hours × ₹240 = ₹1,760 a month. So C(A) = 22,760 + 3x and C(B) = max(x, 3,500) × 7.
At 4,200 transactions:
- A = 22,760 + 12,600 = ₹35,360
- B = 7 × 4,200 = ₹29,400 → B still cheaper by ₹5,960.
At a 20% fall (3,360):
- A = 22,760 + (3 × 3,360) = 22,760 + 10,080 = ₹32,840
- B: volume is below the guarantee, so the bank pays for 3,500 → 7 × 3,500 = ₹24,500 → B still cheaper by ₹8,340.
Reading it correctly: the minimum guarantee hurts Option B exactly when footfall falls, yet it still wins, because A's problem is its fixed base, not its variable rate. Note also that A's cost per transaction rises as volume falls, which is the opposite of the usual assumption.
Revised break-even with the officer's time: 22,760 = 4x → x = 5,690, about 35% above current volume.
3. State the transaction band in which the recommendation flips, and list two factors outside the arithmetic that could still justify Option A.
Read the worked answer
Where it flips. Option A becomes the cheaper choice once sustained volume exceeds 5,250 (or 5,690 counting the officer's time). In monthly terms that is a band of roughly 5,250–5,700 transactions; below it, B wins on cost. The recommendation is therefore volume-contingent, not permanent, and the contract should be reviewed annually against actual footfall.
Two non-cost factors that can justify A:
- Control over cash and uptime. Under a white-label arrangement the vendor's replenishment failures and downtime become the branch's customer complaints, but not the branch's levers. In a rural area with a single cash point, control can outweigh ₹5,000 a month.
- Data and cross-sell. Transactions at an on-site machine stay within the bank's own channel and can be used for customer profiling and account acquisition; a vendor-run machine keeps that data at arm's length.
Also weigh: exit costs and lock-in in the vendor contract, the reputational cost of a minimum-guarantee invoice in a month when footfall dips, and whether the branch's obligation is genuinely customer service rather than a pure cost-minimisation exercise. A defensible answer states the arithmetic, then names the factor it is overriding and why.