A scattered application funnel
Applications submitted, approved, rejected and disbursed live in different systems, so it is hard to tell which agents bring in good loans.
Solutions by industry
Consumer finance companies need to see the application funnel, disbursements, collections and portfolio quality at the same time, by branch, agent and online channel. The Financial services pack ships with approval rate, average loan balance, bad-debt ratio, PAR30 and collection rate.
Free plan with no time limit · No card required · Sample workspace ready in minutes
Applications submitted, approved, rejected and disbursed live in different systems, so it is hard to tell which agents bring in good loans.
PAR30 and bad-debt ratios worked out by hand each month show problems late; threshold alerts surface them sooner.
A gap metric between booked revenue and interest actually collected gets accounting and risk talking about the same number.
Personal-data columns are hashed on load and masked before they reach any AI model; no API, Si answer or MCP call returns borrower details.
Each team has its own metrics and dashboards on the same industry data.
Si answers with approved metrics, their source and confirmation level.
Tickets linked to the branch, product and dealer of the loan mentioned, against outstanding balance, without exposing borrowers.
Ticket volume, response and resolution speed, and customer satisfaction by channel and category.
Productivity, speed and quality of handling for each customer service agent and team.
Company-wide disbursements, approval rate, outstanding balance and bad debt by branch and product.
Definitions come from the analytics packs; Si, dashboards and reports use exactly these.
| Metric | Definition | Unit |
|---|---|---|
| Loan applications | Number of loan applications submitted in the period (by submission date), across all decisions: approved, rejected, pending review and withdrawn by the customer. | Count |
| Approval rate | Approved applications divided by decided applications (approved plus rejected), by submission date. Applications pending review or withdrawn by the customer are excluded.= Approved applications / (Approved applications + Rejected applications) | Ratio (%) |
| Disbursed amount | Total principal disbursed for loans in the period (by disbursement date). | Currency |
| Average outstanding balance | Outstanding principal averaged over the month-end snapshots in the period = total outstanding across snapshots / number of snapshots.= fs_outstanding_snapshot_total / fs_snapshot_count | Currency |
| NPL ratio | Outstanding balance in debt groups 3-5 (more than 90 days past due) divided by total outstanding, on the month-end snapshots in the period.= fs_npl_outstanding_total / fs_outstanding_snapshot_total | Ratio (%) |
| Portfolio at risk 30+ days (PAR30) | Outstanding principal of loans more than 30 days past due divided by total outstanding, on the month-end snapshots in the period.= fs_par30_outstanding_total / fs_outstanding_snapshot_total | Ratio (%) |
| Collection rate | Amount collected on installments falling due in the period divided by the amount due on those installments (by due date, with collections counted up to the data date).= fs_collected_for_due / Amount due | Ratio (%) |
| Interest income collected | The interest portion of customer payments in the period (by payment date). Payments are applied first to the interest of earlier installments, the remainder to principal. | Currency |
| Net income on average outstanding | Ledger gross profit (booked revenue minus funding costs recorded in cost of goods sold) divided by the average of month-end outstanding balance snapshots in the period. A rate for the period asked, not annualized.= Gross profit / Average outstanding balance | Ratio (%) |
Not in its original form. Columns flagged as personal data are hashed on load and masked before they enter Si's prompt.
The workspace defines its own metric (Approved level), versioned and checked before it goes live; the standard pack is a starting point.
The warehouse and backups are hosted in Vietnam, and the Enterprise plan offers an AI model that runs in-country. See the Data in Vietnam page.
Free plan with no time limit · No card required · Sample workspace ready in minutes