Blog · Forecast and pipeline · Commercial banking
How a commercial bank finds the relationships whose operating balances have fallen against their own baseline while the accounts remain open, from the daily or month-end balance file: the balance baseline per customer, the decline signal, the split between seasonal, business-driven and moved-elsewhere, and the list per relationship manager ranked by the balances that have gone.
A commercial bank measures attrition when an account closes. The balance file shows the money leaving a year earlier, at customers whose accounts stay open, in transfers to another bank that the transaction data can see. This guide sets out the balance baseline, the decline signal, the three-way split, and the list.
Per customer, per month:
Baseline = median month-end operating balance, trailing twelve months Prior-year ratio = this month's balance ÷ same month last year Decline if balance < share × baseline for n consecutive months, and prior-year ratio < the same share
Split, from transactions in the decline months:
| Pattern | Reading |
|---|---|
| Outbound transfers to other institutions rose | Moved elsewhere |
| Receipts fell, transfers flat | Business shrinking |
| Neither; prior-year ratio near 1 | Seasonal |
Customer identifiers only; counterparty type, not name.
closing balance = opening + receipts − payments, per customer per month
A month where it fails by more than a tolerance has a missing transaction file, and the split for that month is marked unavailable.
Share 60 percent; three months.
| Customer | Baseline | Last 3 months | Prior-year ratio | Split | Balances gone | RM |
|---|---|---|---|---|---|---|
| 4471 | $2.1m | $1.1m, $0.9m, $0.7m | 0.34 | Moved: transfers to another bank up 4× | $1.4m | RM-04 |
| 2210 | $0.8m | $0.4m, $0.4m, $0.3m | 0.41 | Business shrinking: receipts down 55% | $0.5m | RM-11 |
| 9034 | $1.4m | $0.7m, $0.6m, $0.8m | 0.96 | Seasonal | RM-04 |
Customer 4471's operating balances have dropped by two thirds and the transaction data shows four times the usual outbound transfers to another institution. The account is open. The relationship manager has a conversation to have this month that will be a closure notice next year.
Per relationship manager and per segment: customers in decline by split, and the balances gone. A segment where moved-elsewhere is rising is a pricing or service problem the bank is losing to a named type of competitor.
Closure as the measure. A year late.
Balances without transactions. The shrinking business and the moving customer get the same call.
Seasonality ignored. Every retailer in February.
Counterparty names required. They are not; type suffices, and the export stays pseudonymised.
Mapped once, the balance file, the transaction summary and the customer master produce the baselines, the declines, the split and the list per relationship manager every month. Covirage builds this from the exports as they are. The commercial banking page describes the setup, and the lending-only guide covers the customers whose balances were never here to leave.
Balances say the money left. Transactions say where: regular transfers to another institution are a customer moving its banking; a fall in receipts is the customer's business shrinking; both look the same on a balance chart. The split is what makes the list actionable, and it needs only counterparty type, not counterparty names.
The same month last year is on every line. A retailer whose balances fall every February is not declining in February. A decline is measured against both the baseline and the prior year's pattern, and the report shows both.
Not yet, which is the point. The account closes six to eighteen months after the balances move, and the closure is the first thing most banks measure. The balance decline is the window in which the relationship manager can still ask why.