Blog · Forecast and pipeline · Commercial banking
The complete deposit flight calculation on five commercial banking relationships, small enough to check by hand: the trailing twelve month-end balances and the median as baseline, the last three months against the baseline and against the same months last year, the decline rule, the transaction split that says whether the money moved to another bank, the business shrank or the season turned, the balances gone, and the assertion that balances reconcile to receipts and payments, so a reader can reproduce every figure and then run it on their own balance file.
Deposit flight is a balance against its own history and a transaction split, and on five relationships it can be checked by hand. This page works the baseline, the decline rule, the prior-year test, the split, the balances gone, and the assertion. The last three months are July, August and September.
| Relationship | Median of trailing 12 (baseline) | Jul | Aug | Sep | Same months last year: Jul, Aug, Sep |
|---|---|---|---|---|---|
| A | $2,100 | $1,100 | $900 | $700 | $2,000, $2,050, $2,150 |
| B | $800 | $400 | $400 | $300 | $780, $820, $790 |
| C | $1,400 | $700 | $600 | $800 | $720, $650, $760 |
| D | $500 | $520 | $480 | $510 | $490, $500, $520 |
| E | $3,000 | $2,900 | $1,600 | $2,950 | $2,950, $3,100, $2,900 |
Decline if balance < 60% of baseline for three consecutive months, and balance < 60% of the same month last year
| Relationship | Sep ÷ baseline | Three months under 60%? | Sep ÷ same month last year | Decline? |
|---|---|---|---|---|
| A | 33% | Yes: 52, 43, 33 | 33% | Yes |
| B | 38% | Yes: 50, 50, 38 | 38% | Yes |
| C | 57% | Yes: 50, 43, 57 | 105% | No: seasonal; same every year |
| D | 102% | No | 98% | No |
| E | 98% | No: August alone | 102% | No: one month; a large payment |
| Relationship | Receipts, 3 months vs prior year | Payments to other institutions, vs prior year | Reading |
|---|---|---|---|
| A | $4.1m vs $4.0m: flat | $3.9m vs $0.9m: ×4.3 | Moved elsewhere |
| B | $1.1m vs $2.4m: −54% | $0.3m vs $0.3m: flat | Business shrinking |
Gone = baseline − current, for decliners
| Relationship | Baseline | Current | Gone | Relationship manager | Conversation |
|---|---|---|---|---|---|
| A | $2,100 | $700 | $1,400 | RM-04 | Where the money went, and why |
| B | $800 | $300 | $500 | RM-11 | The business, and the facility |
For each relationship and month: closing = opening + receipts − payments. Relationship A, August: opening $1,100, receipts $1,350, payments $1,550: closing $900. Holds. A month that fails by more than the tolerance has a missing transaction file, and that month's split is marked unavailable.
Baseline as a mean. E's one large receipt month raises the mean; E looks in decline the following month.
No prior-year test. C is on the list every summer.
No transaction split. A and B get the same call; one needs a treasury conversation and one a credit one.
Closures as the measure. A's account is open. It closes next year.
The same baseline, rule, prior-year test and split per relationship. Covirage runs it on the balance file and the transaction summary every month. The deposit flight guide covers the measure, and the baseline and norm guide covers why the reference is the relationship's own history.
Because one month with a large receipt in it, a customer's sale proceeds, would set a mean baseline the customer never normally holds. The median is the balance the relationship typically carries.
By the prior-year ratio: this month's balance over the same month last year. A relationship at 40 percent of baseline and 95 percent of last year's same month is doing what it does every year. Both tests must indicate decline.
Receipts, and payments split by counterparty type: to the customer's own accounts at other institutions, to suppliers, to payroll. No counterparty names. Transfers to other institutions rising while receipts hold is the moved-elsewhere signature.