The two churn types and why each requires a different response
Prepayment churn: a borrower who is about to prepay and switch to a competitor is a borrower who has found a better deal. The institution's response options are: proactively offer a rate reduction (converting a departing borrower into a retained one), acknowledge the borrower's strong credit position and make a retention offer that is better than what the competitor is likely to offer, or accept the prepayment and focus on capturing the next loan cycle. None of these responses is possible if the institution does not know the prepayment is coming. The Lifecycle AI's 60-day warning gives the institution time to make a retention offer before the borrower has already committed to the competitor.
NPL churn: a borrower who is about to miss their first instalment has been showing stress signals for 60 to 90 days — declining bank credits, increasing outward transactions, the appearance of new instalment debits from other lenders, portal disengagement. The institution's response to these early signals is very different from its response to an actual missed payment — at 60 days before the first miss, the right response is a proactive conversation about financial health, not a collection call. The Lifecycle AI identifies the stress pattern and routes the borrower to the appropriate team for a supportive, pre-emptive engagement.
The churn prediction signal model: 10 signals across 2 churn types
| Signal | Churn type | Weight | What it indicates |
|---|---|---|---|
| Multiple visits to competitor bank/finance company websites (via browser fingerprint, where consent-based data is available) | Prepayment | High | Borrower is actively comparing rates — the strongest prepayment intent signal available |
| instalment calculator used for a loan amount equal to or close to current outstanding | Prepayment | High | Borrower may be modelling what the refinanced instalment would look like at a competitor's rate |
| credit score improvement of 50+ points — now significantly better credit risk than at origination | Prepayment | Medium | Borrower now qualifies for better rates at other institutions — prepayment incentive exists |
| Portal login frequency declining over 60 days (from regular to near-zero) | Both types | Medium | Decreasing engagement is a leading indicator for both disengagement-before-prepayment and disengagement-before-default |
| Bank statement: salary credits declining or irregular over last 3 months (AA data) | NPL risk | Very high | Income stress is the primary precursor to payment default — declining salary credits are the earliest warning available |
| Bank statement: new GIRO / direct debit debits appearing (new instalment obligations to other lenders) | NPL risk | High | Borrower has taken on additional debt — total instalment obligation is increasing relative to income |
| Bank statement: end-of-month balance declining trend over 3 months | NPL risk | High | Borrower's buffer is shrinking — a short-term income interruption will produce a missed payment |
| GIRO / direct debit bounce followed by late recovery (recovered Day 8–15, not Day 1–3) | NPL risk | High | Late recovery suggests the borrower is stretching to make the payment — the next month may not recover at all |
| SME: GST filing gaps or revenue declining YoY | NPL risk | Very high | Business distress for SME borrowers shows up in GST before it shows up in instalment payments — lead time of 60–90 days |
| External market: sector-specific stress event (commodity price shock, sector regulatory change) | NPL risk | Medium | Borrowers in affected sectors are portfolio-level NPL risk — used in combination with individual signals |
The churn prediction panel: two at-risk borrowers
The conversation at Day −60 costs a phone call. The conversation at Day +90 costs a lawyer.
Rajan Textiles' NPL risk score crossed 70 today. The RM call this week costs 20 minutes and may reveal that the business is facing a temporary setback that a payment moratorium would resolve — preserving the relationship and avoiding the NPL classification entirely. If that call does not happen and Rajan misses November's instalment, the Early Bucket Caller takes over the account, the DPD clock starts, and the relationship shifts from partnership to enforcement in 30 days. The Lifecycle Campaign Manager AI's churn prediction system is not a collections tool — it is the mechanism that keeps accounts out of collections by giving the institution 60 days of forewarning during which the right response is still a support conversation rather than a demand.