System architecture
How CanFund underwrites, prices, and funds a file
Five stages run between application and disbursement. Each stage produces a defined output that the next stage consumes, which is why the process is measurable end to end rather than estimated.
- Avg. decision
- 4m 12s
- Funding speed
- 1 business day
- Data points
- 1,900+
- Sales data connectedNext
- Cash flow reviewedNext
- Amount and terms setNext
- Offer ready to acceptNext
Approved for
$0
Example
Time to decision
0.0 min
Example
Your sales trend
+18.4%
Pipeline
Stage-by-stage walkthrough
- 01
Data connection
You upload the last three months of business bank statements as PDFs, and optionally card processing or accounting exports. We never ask for online banking credentials and never connect to your account, so nothing on our side can move funds.
- Categorized transactions
- Daily balance series
- Recurring debit map
- 02
Real-time cash flow model
Transactions are normalized and projected forward. The model reads deposit frequency and variance, negative-balance days, NSF events, existing financing positions, and revenue trend at daily rather than monthly resolution.
- Forward cash position
- Volatility index
- Existing position schedule
- 03
Risk pricing
Approved amount is constrained by a debt service coverage floor for fixed-payment products, or by a delivery-window target for revenue-based products. Cost is derived from measured volatility, tenure, and industry loss base rates.
- Max exposure
- Priced structure
- Total dollar cost
- 04
Funding deployment
The offer states amount, term, remittance mechanics, and total dollar obligation before signature. Signed files are verified and disbursed by ACH, with vendor-direct disbursement for equipment transactions.
- Signed agreement
- Verification pass
- ACH disbursement
- 05
Servicing
Remittance runs on the contracted cadence and is reconciled against settled revenue. Balance, delivery progress, and next debit date remain visible for the life of the position.
- Reconciliation ledger
- Delivery progress
- Renewal eligibility
Underwriting technology
What the model reads, and why it reaches a decision sooner
The short answer: the model reads the transaction record directly instead of reading a summary of it. Monthly statement totals conceal the two signals that matter most for short-duration credit — how consistent deposits are week to week, and how often the account runs negative.
Conventional underwriting is not slow because the analysis is hard. It is slow because collecting, transcribing, and reconciling documents takes days. Removing that phase is what compresses the timeline; the modeling itself was never the bottleneck.
No claim is made here about proprietary data sources. The inputs below are standard across the industry. The difference is resolution, automation, and consistency of application.
- Deposit consistencyVariance between weekly deposits, not the monthly total
- Negative daysFrequency and duration of sub-zero balances
- Existing positionsRecurring daily or weekly debits from other funders
- Revenue trendDirection across the trailing 6–12 months
- Processor behaviorRefund rate, chargeback rate, settlement cadence
- Receivable detailInvoice-level aging and payer concentration
- Industry base ratesVertical loss behavior as an adjustment, not a verdict
Questions
Common questions about the process
- How does CanFund decide faster than a bank?
- The time saved is in data handling, not analysis. A bank's timeline is dominated by requesting statements, waiting on the applicant, and re-keying them into a model by hand. We ask for three statements up front and parse them automatically the moment they arrive, which removes that phase entirely and leaves only the modeling step, which is fast in both processes.
- What data does the underwriting model use?
- Bank transaction history including deposits, withdrawals, daily balances, negative-balance days, and recurring debits; card processing settlement volume, refund rate, and chargeback behavior; accounting data at invoice level where connected; business registration and time in business; and a soft-pull credit profile. Industry-level loss base rates are applied as an adjustment.
- Is the decision fully automated?
- No. Automated underwriting handles the routine majority of files. Applications near the decision boundary, unusually large requests, and files with contradictory data route to a credit analyst for review. Automation concentrates human attention on files where judgment changes the outcome.
- Can I be declined and reapply later?
- Yes. Declines are usually a function of insufficient operating history, coverage headroom, or existing positions — all of which change over time. There is no fixed waiting period, and reapplying after a material change in deposit behavior frequently produces a different result.
Deeper detail is in how automated small business underwriting actually works , or compare structures across the product index.
Run your file through the pipeline
Connect an account and see the modeled structure, cost, and total obligation before you commit to anything.