The Modeling screen is where Vintage shows how a group of your loans has actually behaved, and what that behavior means for the price of a new loan. You choose the group, a cohort (loans you expect to behave alike), and the screen shows, by loan age (months since origination), how much of the cohort’s original balance was lost to charge-offs and how quickly it paid down early. It then prices a representative new loan on top of those two behaviors.
Every number on the screen is meant to be re-derivable by hand. The method is historical averaging by loan age, weighted by the balances behind each age. There is no machine learning, no fitted curve, and no hidden parameter. The methods are explained in Vintage analysis.
Modeling starts from a segment
Section titled “Modeling starts from a segment”The screen models a segment: a slice of your portfolio you define in the Segment Builder, the filter panel beside the analysis. It is the same segment the Portfolio screen uses, so narrowing it on either screen narrows it on both. Vintage remembers your segment for you, per organization, between visits and across devices. The Segments page covers the filters themselves.
Modeling does not analyze your whole book at once, by design. A curve is only meaningful when the loans behind it behave alike. Averaging auto loans with mortgages produces a curve that describes neither. In practice a useful cohort holds at least a single product type, and usually something narrower: a FICO band, a loan-size band, or an origination window.
So when you arrive with no segment defined, the screen does not show empty charts. It shows Define a segment to see results, with three examples of how analysts narrow toward a cohort: New auto loans, FICO 660 to 719, Originated 2021 to 2023. These are examples to read, not buttons. Build your segment in the Segment Builder and the analysis appears.
What you see once a segment is modeled
Section titled “What you see once a segment is modeled”The top of the analysis carries the cohort’s headline figures:
| Figure | What it is |
|---|---|
| Loans | How many loans in the segment Vintage could model: loans it can age and that have a term. A loan that was already closed on the first row Vintage ever saw for it is not counted here. |
| Total Balance | The current balance of those same loans. The count and the balance always describe the same population. |
| Expected Lifetime Loss | The cohort’s cumulative loss at its contractual term (the balance-weighted average term of the modeled loans), as a share of original balance. See Reading the credit-loss section. |
| Avg Annualized Prepayment | The cohort’s prepayment speed as an annualized rate (CPR, the conditional prepayment rate): one annualization of the cohort’s pooled monthly speed over every measurable loan-month. See Reading the prepayment section. |
Either of the last two can show a dash in place of a number. A dash means the figure could not be measured, and the line under it says why. Vintage never shows a confident 0.00% for something it could not measure.
Under the headline figures sits a row of tabs: Credit loss, Prepayment, Pricing, Results. The analysis is long, so the tabs stay docked at the top of the screen once you scroll past them. Clicking one jumps to that section, and the highlighted tab follows the section you are reading as you scroll. Results is the last part of the Pricing section, where the priced loan’s figures live.
The three sections are:
- Credit Loss: the cumulative loss curve by loan age. See Reading the credit-loss section.
- Prepayment: the prepayment speed by loan age. See Reading the prepayment section.
- Pricing: your assumptions, and the required rate and profitability for a representative new loan. See Pricing a loan.
The screen’s five states
Section titled “The screen’s five states”The Modeling screen always shows exactly one of five things, and each one means something different.
| State | When you see it | What it shows |
|---|---|---|
| No portfolio yet | Your organization has no loans. | An empty state. Admins and editors also see an Upload loan data button; viewers see the message alone. |
| Your first upload is still processing | Your organization has uploaded, but its first portfolio is not built yet. | A reassurance that the portfolio is being built, with progress when Vintage can count it. |
| Define a segment to see results | No segment is defined. | The starting prompt described above, with the Segment Builder beside it. |
| Segment is too large | The segment holds more loans than your organization’s maximum. | The loan count, your maximum, and what to do. See below. |
| The analysis | Everything else. | The headline figures, tabs, and three sections. |
Your first upload is still processing is worth a closer look. It appears only before your organization has ever had a finished portfolio. It shows N of M loans processed whenever those counts are reliable, and it says You can leave this page. We will email you when it is ready. only when an email really is on its way to you. Otherwise it says This view will unlock automatically. It quotes no duration, because a first upload can be a very large book. This is the one state shown without the Segment Builder beside it: until the first portfolio exists, there are no ranges or values to build filters from. Once the portfolio is ready the screen opens on its own, and you never see this state again. Later uploads are handled as described in When your portfolio is updating.
Other messages you might see
Section titled “Other messages you might see”A few situations stop the analysis for a reason specific to the segment:
- We could not apply this filter: one of the segment’s filters could not be read, so nothing was analyzed. Remove or re-enter the filter you changed most recently.
- This segment is too complex to analyze: the analysis ran past its time limit. Narrow the segment, for example with fewer text filters or a shorter date range, and run it again. There is no retry button, because the same segment run again unchanged would not fit either.
- Nothing in the segment can be measured. The screen names the reason from the segment’s own counts, with what would bring the loans in: No loans can be aged yet (no origination date, and seen only in your earliest snapshot month), No loan terms yet (no term or maturity date), No snapshots for these loans yet, These loans closed before your data begins, or Nothing here can be measured yet (a mix of the last two). The rules behind these are in Origination and term and Missing and unreported data.
- We could not load the modeling analytics. Please try again. This appears when the analysis itself failed to load, as opposed to any of the reasons above.
The segment size maximum
Section titled “The segment size maximum”Vintage models a segment up to a maximum number of loans, and that maximum belongs to your organization. It is not one number shared by every customer.
What sets it is how many months of history your data spans. The analysis reads every month each loan was observed, not only the loans themselves, so the same number of loans is more work on a book carrying years of monthly snapshots than on a book with a single snapshot. A deeper history therefore gets a lower loan maximum. This holds every customer to the same amount of analysis rather than the same loan count, so no organization waits far longer than another for a segment of the same size. Your maximum moves only when your own history genuinely gets deeper, so it is a number you can plan around.
Past the maximum, the screen does not attempt the analysis. It says so and names your organization’s own number:
Segment is too large There are X loans in this segment. The maximum segment size for modeling is Y. Update the segment definition to be more narrow in order to see results.
X is the segment’s loan count and Y is your organization’s maximum, shown as a round figure (rounded down, so the number you read is always one Vintage will accept). The Segment Builder’s live loan count stays visible while you narrow, so you can steer toward the limit rather than hit it blind. A cohort large enough to exceed the maximum is usually also too mixed for its averaged curves to describe any real group of loans, so narrowing tends to improve the analysis as well.
When your portfolio is updating
Section titled “When your portfolio is updating”After an upload is accepted, an upload or file is deleted, a field is retyped, or you change your value vocabulary or reporting-basis declarations on the Fields screen, Vintage re-reads the affected loans before the analytics fully reflect the change. Vintage also does this when it improves its own calculations. The screen does not go blank while that runs.
For a segment Vintage has analyzed before, the screen keeps showing the last complete analysis, with a notice above it:
Updating your analytics These numbers reflect your last complete run. A refresh is in progress (8,200 of 10,000 loans updated). Updated as of 3/14/2026, 9:42 AM.
(Figures illustrative. The date is when the analysis on screen was computed, shown in your own date format.)
The fraction appears only while loan work is actually outstanding. When the refresh involves no loan work Vintage can count, the notice says a refresh is in progress and quotes no fraction, no bar, and no time estimate. The notice stays up until every loan has been brought up to date. Vintage never declares the numbers current while some loans carry the new interpretation and some the old. It raises the notice whenever a refresh is in flight anywhere in your organization, even if this segment’s numbers turn out not to change, because a stale number read as current is the worse mistake.
For a segment Vintage has never analyzed (one you define for the first time while the portfolio is still updating), there is nothing complete to fall back on. The screen shows the same processing state the Portfolio screen shows, and unlocks on its own when the portfolio is current. A segment defined against an up-to-date portfolio computes right away.
In every case the screen never shows preliminary curves computed from a half-updated portfolio. A tab you left open updates itself: the notice appears, then clears with fresh numbers, without a reload. A tab in the background stops checking while it is hidden and catches up when you return to it.
If a processing problem on Vintage’s side temporarily prevents some loans from being read, those loans drop out of the curves until the issue is resolved, rather than holding the whole screen in a waiting state. The Portfolio screen discloses how many loans are affected.
When you change the segment
Section titled “When you change the segment”A segment change that comes back quickly replaces the results in place, with no waiting card. When a change takes longer, the analysis dims and a small card appears over it: first Processing Segment with a short progress bar, then Running Analysis with a spinner. If the analysis is still running after several seconds, the card adds Still running. Large segments take a few seconds. The card fades out when the new results arrive.
This card belongs to segment changes only. It never appears for the screen’s own background checks, and never on top of the processing state, so you only ever see one waiting indicator at a time.
Downloading what you see
Section titled “Downloading what you see”Every chart and table carries a Download menu with two formats: CSV (data) and Excel (data + chart). Downloads carry the same assumption sentences the screen shows beside the figures. See Exporting results.
Who can use Modeling
Section titled “Who can use Modeling”Every member of your organization can open Modeling, view-only members included, and it is the screen you land on when you sign in. (An admin or editor whose organization has no loans yet starts on the Uploads page instead.) Your pricing assumptions are saved for you personally, within each organization.
While the Vintage team is setting up a new organization’s first portfolio, Modeling is not in the navigation and the screen explains the wait instead. See Your first upload.