See what Pave would do for your portfolio
A backtest applies Pave's cashflow models to your historical data. No integration required. Real results from your own borrowers in 1 to 3 weeks.
WHAT IS A BACKTEST
Think of it as a time machine for your underwriting decisions.
A backtest is a simulation that applies Pave's cashflow attributes and models to your historical loan data. We re-score your past applications and outcomes to measure exactly how Pave would have performed.The result is proof before commitment.
You see the impact on your actual portfolio before changing a single production decision. No guesswork, no pilots on live traffic. Just concrete projections based on your own data.
How many additional good borrowers you could have approved safely.
How delinquency rates and ACH returns would have improved across your portfolio.
Quick wins you can implement immediately. Plus a staged path from shadow test to full rollout.
Results broken out by cohort, amount bucket, and risk tier so you know where to act first.
From conversation to concrete results
Most lenders are surprised by how quickly they move from first conversation to quantified impact. Here is the process.
We learn about your lending product, portfolio, and what success looks like. Together we outline what the backtest will measure and which scores or attributes to evaluate.
Send us transaction history, outcomes, and a loan tape via SFTP, warehouse export, or aggregator feed. We adapt to your existing infrastructure. NDA in place before any data moves.
Our data science team ingests your data, generates 10,000+ cashflow attributes, and scores every borrower. We compare off-the-shelf scores and build a custom model tuned to your portfolio.
We walk you through a readout covering approval lift, delinquency rates, ACH returns, expected loss, and segmentation. Every number comes from your own borrowers.
We learn about your lending product, portfolio, and what success looks like. Together we outline what the backtest will measure and which scores or attributes to evaluate.
Three inputs. That's it.
We work with your reality. Whether you pull data from SFTP exports, warehouse tables, or aggregator feeds, we adapt to your infrastructure.
6 to 12 months is ideal to capture seasonality and pay cycles. Plaid, MX, or raw bank data all work.
Even 90 days can produce results. We document confidence limits clearly.
Approvals and denials with performance labels. We use this to measure how Pave's scoring maps to real repayment behavior.
Approvals-only works too. We incorporate reject inference and flag the trade-offs.
Application dates, amounts, terms, and status. This ties transactions to lending decisions so we can model impact precisely.
Standard format. We handle transformation into Pave's schema on our side.
Backtests that became production wins
Before you start
No, but they help. Many teams start with approvals-only and add reject inference later. We document the trade-offs and run sensitivity checks so you understand how denied data would strengthen calibration.
We handle that. We publish coverage dashboards showing the share of traffic with high-confidence tags, merchants, and MCCs. You will see exactly how data quality affects lift, so you can prioritize the highest-ROI fixes.
Yes. Whether you are evaluating personal loans, earned wage access, credit cards, or SMB credit, we configure product-specific attributes and score settings for each segment.
We accept Plaid format, raw bank data, CSV,Parquet, and database exports. Data arrives via SFTP or direct warehouse connection. Wetransform everything into Pave's schema on our side so you do not have to reformat anything.
A backtest uses historical data sothere is zero risk to your current operations. No integration required and no changes toproduction decisions. Once results look good, we move to shadow testing and then a controlledA/B test before full production.
Ready to see your numbers?
Most lenders get from first conversation to backtest results in under three weeks. The faster you start, the sooner you stop leaving good borrowers on the table.