Replacing Collections Software at a Bank or Lender: A 12-Step Cutover Plan
Most collections cutovers don't fail on technology. They fail on rules nobody wrote down. A 12-step plan for moving strategies, contact rules and in-flight accounts without losing cure rates or breaching conduct rules.
Steps 1–3: Inventory the old system's hidden logic before signing anything
If you're replacing debt collection software for banks or lenders, the risk isn't the new platform. It's what the old one does that nobody documented. Over ten years, a legacy collections system picks up hard-coded suppression rules, odd queue priorities, and workarounds that compliance once approved in an email. Migrate without capturing them and you'll find out through complaints, missed payments, or a regulator.
The first three steps happen before vendor selection is final, and they set the scope for everything after.
- Extract every live strategy and rule. Pull treatment paths, segmentation trees, queue logic, letter and SMS templates, contact-frequency caps, and every suppression flag: disputes, insolvency, deceased, hardship, military or public-service protections, vulnerability markers. Don't stop at the configuration screens. Interview team leads too, because real logic often lives in habits like "we never call the agricultural book during harvest."
- Baseline your KPIs for at least six months. Without a stable before-picture you can't prove the new system works. Capture roll rates by bucket, 30-day cure rate, right-party contact rate, promise-to-pay kept rate, cost to collect, and complaints per 1,000 accounts in collections. Split by product and region; a blended number hides regressions.
- Freeze strategy changes. Freeze the legacy system from the end of rule extraction until go-live. Otherwise the old system keeps drifting while you migrate a snapshot that's stale before launch.
On budget, typical 2026 market ranges for a mid-sized lender are roughly $150,000–$600,000 a year in subscription fees for an enterprise platform, plus $250,000–$1.5 million for implementation. Large multi-country banks often spend well beyond that. The drivers are the number of integrations (core banking, dialer, payments, bureaus, document generation), how many jurisdictions' rule sets you need, and whether you're also buying a digital self-service layer. Contract terms and local pricing vary a lot, so get line-item quotes.
One red flag: a vendor who says your rules will "map automatically" without asking to see them. Established enterprise vendors such as FICO, Experian (PowerCurve), C&R Software and EXUS, and digital-first players such as TrueAccord or InDebted, all need your rule inventory to scope the work. If a vendor doesn't ask, it hasn't scoped it.
Steps 4–5: Rebuild strategy, don't photocopy it
The temptation is to recreate old treatment paths one-for-one. Resist it, but don't redesign everything at once either. The better approach is a like-for-like baseline plus one deliberate improvement, usually in early-stage delinquency, where better tooling pays off fastest.
- Translate strategies into the new platform's decision model. Older systems often segment on days past due and balance alone. Newer loan collections software can segment on behavioral scores, channel preference, payment history and self-cure likelihood. Build version 1 to reproduce current outcomes, and document each rule with its source: policy, regulation, or historical habit. Rules with no traceable source are candidates to retire, but only after compliance agrees.
- Design a single champion-challenger change. Pick one hypothesis to test after go-live, for example: "Low-risk self-curers in 1–10 days past due get a digital reminder only, with no agent call until day 12." Define the cohort, success metric and stop condition in advance.
Here's a hypothetical. Say a lender has 500,000 active installment loans, and about 4% (20,000 accounts) enter early delinquency each month. If history shows 55% of them self-cure within ten days, calling everyone on day 3 burns agent capacity and creates contacts customers find intrusive. That's a Consumer Duty concern in the UK and a complaint driver everywhere. Moving that segment to digital-first treatment could free a meaningful share of early-stage agent hours. You only prove it if the step 2 baseline is clean and you change one thing at a time.
Keep the challenger switched off during the parallel run. You're testing the migration first and the strategy second. Mix the two and you can't tell which caused a change in results.
Steps 6–7: Map contact rules by jurisdiction, then test them as code
This is where cutovers get lenders into regulatory trouble. Contact rules differ by country, by state or province, by product, and by whether you collect your own debt or a third party does. Your new system must enforce them as hard constraints, not guidance agents are trusted to remember.
- Build a jurisdiction rule matrix. For every market you collect in, list permitted contact hours, frequency limits, required disclosures, channel consent rules, vulnerability handling, and third-party agent requirements. Have local counsel or compliance sign off each row.
- Turn every row into an automated test case. For example: "An account in jurisdiction X with a dispute flag gets zero outbound contact attempts." Run the tests on every release, not just at go-live.
The table shows the kind of differences you'll need to encode. It's a general summary, not legal advice. Rules change and have exceptions, so confirm current requirements with qualified counsel in each market.
| Jurisdiction / regime | Who it mainly applies to | Rules that typically need encoding |
|---|---|---|
| UK: FCA Consumer Duty and CONC 7 | FCA-regulated lenders and debt collectors | Forbearance options before escalation, identification and treatment of vulnerable customers, proportionate contact frequency, outcome monitoring by customer group, fair-value evidence for fees and charges |
| US: Regulation F (FDCPA) | Mainly third-party debt collectors; some states (for example California's Rosenthal Act) extend similar rules to first-party creditors | Call-frequency limits per debt, validation notice content, electronic-communication opt-outs, inconvenient-time defaults for calls |
| India: RBI rules on recovery agents and fair practices | Banks, NBFCs and their recovery agents; digital lending rules also cover lending service providers | Permitted calling hours, no intimidation or harassment, borrower informed of recovery agent details, agent due diligence and training, records of lender oversight |
| EU / EEA: GDPR plus national collection laws | All lenders processing personal data; national rules vary significantly | Lawful basis for each channel, data minimization in letters, retention limits, cross-border transfer controls for cloud hosting, national rules on collection fees |
Also check markets such as Australia, the UAE, Brazil and South Africa, all of which have their own conduct rules. A common mistake is treating US Regulation F as a global default because many vendors are US-built. A US bank collecting its own debt may not be directly subject to Regulation F, though many adopt it as policy to manage unfair-practices risk. A UK lender that applies Regulation F logic wouldn't satisfy Consumer Duty. Encode each regime separately, and apply the strictest applicable rule when an account touches more than one, for example a customer who has moved country.
Steps 8–10: Migrate in-flight accounts and run both systems in parallel
Static data is the easy part. The hard part is account state mid-treatment: promises to pay due next Tuesday, payment plans halfway through, disputes awaiting a response, hardship pauses with an end date. Lose any of these and customers get chased for things they've already arranged.
- Migrate account state, not just balances. Your specification should cover open promises, active arrangements, contact history (needed to calculate frequency caps from day one), consent and opt-out records, vulnerability flags, and legal or insolvency holds. Reconcile counts and values between systems, and check every account carrying a protection flag individually.
- Run in shadow mode for roughly four to eight weeks. The new system ingests live data and makes decisions, but the legacy system still drives customer contact. Each day, compare which accounts each system would have contacted, through which channel, with what treatment. Investigate every difference above an agreed tolerance, often set at a low single-digit percentage of daily actions. Treat any difference involving a protected account as a blocking defect.
- Cut over by cohort, not big-bang. Move one product or region first, often a lower-risk unsecured book, then expand in waves of a few weeks each.
| Phase | Typical duration (mid-sized lender) | Exit criterion |
|---|---|---|
| Rule inventory and baseline (steps 1–3) | 6–10 weeks | Signed rule register; six-month KPI baseline |
| Strategy build and rule matrix (steps 4–7) | 10–16 weeks | All jurisdiction test cases passing |
| Data migration and shadow run (steps 8–9) | 6–10 weeks | Decision variance within tolerance; zero protected-account defects |
| Cohort cutover and hypercare (steps 10–12) | 8–16 weeks | KPIs within agreed bands for two full cycles |
Overall, expect roughly 8–14 months for a single-country lender and longer for multi-country banks. Be wary of any vendor plan under six months that also includes dialer, payment and core banking integration.
Steps 11–12: The KPIs that prove the new system works, and when to switch off the old one
Going live proves nothing. Two complete delinquency cycles with stable or better outcomes do. Before cutover, agree the bands each KPI must stay within, measured against the step 2 baseline for the same products and season.
- Track a balanced KPI set during hypercare. Measure collection performance and customer outcomes together, so a gain in one isn't hiding damage in the other:
- Roll rate from early to mid-stage delinquency: no worse than baseline plus a small agreed margin.
- 30-day cure rate and promise-kept rate: within baseline bands.
- Right-party contact rate and digital self-service share: these should rise if the digital layer is doing its job.
- Contact-rule breaches: target zero, with root cause analysis on every one.
- Complaints per 1,000 accounts and vulnerable-customer outcomes: monitored separately, as Consumer Duty-style regimes expect.
- Cost to collect per unit of currency recovered: often flat at first, improving once challenger strategies run.
- Decommission with evidence. Keep the legacy system read-only for the record-retention period that applies in each market (commonly several years for contact records, but it varies), or archive its data in searchable form. Switch off the old platform only after the go/no-go board signs off two stable cycles and the archive has passed a retrieval test.
A practical decision rule: if roll rates degrade beyond your band for two consecutive weeks in a cohort, pause further waves and roll that cohort back to the legacy system. That only works if the change freeze and shadow-run data from earlier steps are still in place, which is why those steps can't be skipped.
Your next move: before the next vendor meeting, assign one owner to produce the step 1 rule inventory and the step 6 jurisdiction matrix. Any vendor whose proposal can't be checked against those two documents isn't ready for your business.
Quick answers
How long should a parallel run last when replacing collections software?
Most lenders run shadow mode for roughly four to eight weeks, enough to cover at least one full early-stage delinquency cycle and a month-end. Base the exit on decision variance between the systems, not the calendar. Differences in daily actions should sit within an agreed tolerance, and protected accounts should show zero defects. Seasonal books, such as agricultural or holiday-spend portfolios, may need longer.
Does Regulation F apply to a bank collecting its own loans?
Generally, Regulation F implements the FDCPA, which mainly covers third-party debt collectors rather than creditors collecting their own debts. Some US states extend similar rules to first-party creditors, and many banks adopt Regulation F-style limits as policy to manage unfair-practices risk. Your compliance team or outside counsel should confirm which rules apply to each portfolio and state.
What does it typically cost to replace a collections management system?
For a mid-sized lender, typical 2026 market ranges are roughly $150,000–$600,000 a year in subscription fees, plus $250,000–$1.5 million for implementation. Large multi-country banks often spend considerably more. The biggest drivers are the number of integrations, the number of jurisdictions' rule sets, and data migration complexity. Pricing varies by contract and region, so get itemized quotes from shortlisted vendors.
Written by the Smartnode Suite editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.