A cash recovery forecast estimates when eligible collections may settle from existing and future cohorts. The useful output is not a single confident number. It is a versioned set of scenarios that connects portfolio composition, cohort maturity, payment behavior, operational capacity, and known uncertainty to a rolling cash view.

This is educational planning guidance, not an investment, valuation, accounting, or financial recommendation. Forecasts are uncertain and should be reviewed by qualified business, finance, compliance, legal, and data owners.

Separate the forecast questions

Distinguish short-term cash operations from long-term portfolio performance. A 13-week treasury view may use scheduled payments, settlement timing, returns, and pipeline events. A portfolio forecast may extend by cohort age and rely more heavily on historical liquidation patterns.

Do not mix gross collections, settled cash, client remittances, and net contribution in one line. Reconcile the definitions to the recovery-rate framework.

Build a controlled source layer

• fixed account and balance cohort snapshots;
• settled transaction and reversal events;
• active payment schedules and promise status;
• portfolio, product, vintage, and workflow segment;
• legal, dispute, complaint, and communication holds;
• recalls, returns, sales, and placement changes;
• calendar, banking-day, and cutoff rules;
• operational capacity and known system constraints.

Use cohort maturity explicitly

Map each cohort to its current age and historical incremental recovery pattern. Forecast the remaining curve rather than applying a mature lifetime rate to a new placement. Keep the observed portion separate from the modeled portion.

Use the liquidation curve guide to construct equal-age histories. Reweight or exclude older history only through a documented policy.

Create baseline, downside, and upside scenarios

The baseline should represent the approved central assumptions. The downside should pressure-test weaker conversion, slower settlement, higher returns, missing placements, or capacity limits. The upside should use supportable improvements rather than a desired budget plug.

Record every assumption with its source, owner, effective period, and sensitivity. Show which assumptions explain most of the range.

Model payment timing carefully

Separate promise creation from successful settlement. Account for authorization dates, scheduled dates, processing lag, weekends, holidays, pending status, returns, reversals, refunds, and forwarded payments. Forecasting on promises alone can pull cash into the wrong period.

Reconcile observed settlements to Kaizen’s payment reconciliation workflow before using them to recalibrate the model.

Backtest and monitor drift

At each refresh, compare the prior forecast with actual settled cash by cohort, segment, and horizon. Decompose error into volume, mix, timing, conversion, return rate, and data corrections. Retain the original forecast so the backtest cannot be rewritten.

For a material statistical or machine-learning model, establish independent validation, change controls, and ongoing monitoring. Federal Reserve SR 11-7 provides a useful model-risk reference for covered banking organizations and a practical set of governance questions for others.

Publish uncertainty with the number

Show the forecast date, horizon, scenario range, observed-versus-modeled split, confidence limitations, cohort coverage, data freshness, and major assumptions. Name events that the model does not capture. A clean chart without these disclosures encourages false precision.

Restrict access to account-level data and publish aggregated planning views where possible. Ensure forecast development does not bypass approved consumer-treatment controls.

Conclusion

A rolling cash recovery model should connect reconciled transactions, stable cohorts, realistic timing, and explicit scenarios. Backtest every version, explain forecast error, and publish uncertainty beside the estimate. Kaizen’s Recovery Suite can support transaction and account visibility while the organization retains responsibility for its assumptions and decisions.

Frequently asked questions

How often should a recovery forecast be refreshed?

Match the cadence to the decision and data. Short-term cash planning may need weekly refreshes, while long-range cohort forecasts may be monthly. Material events can trigger an out-of-cycle update.

Should scheduled payments equal forecast cash?

No. Scheduled payments are an input. Settlement timing, failures, returns, reversals, and account changes should be modeled and reconciled separately.