Five stages take a raw bid tape to an investor-ready bid summary — every loan priced through the same Program & Pricing engines your originators use, in a workspace completely isolated from your production pipeline.
Bulk loans live only in per-client staging tables and a short-lived in-memory cache —
never the production loan schema. The engine runs in test-only mode: no
loan-table writes, no logging side-effects. A 5,000-loan batch can be priced, re-priced, and discarded
without touching — or risking — a single real loan in your pipeline.
Bring the loans in — a bid tape from a counterparty, a portfolio extract, or a what-if scenario set.
The File workspace manages source spreadsheets per client: upload, pick the sheet to import, and remove old files. The source file is stored against the client, so a recurring vendor format is always at hand.
The accuracy lever. Point each spreadsheet column at the loan field it represents — once.
Mapping ties each column (and, where needed, its values) to a loan field, then saves it as a reusable map for that vendor format — the same tape maps the same way every time. A mis-mapped column feeds wrong data to a correct engine, producing wrong-but-plausible results, so the validated, reusable map is what keeps a batch trustworthy. A Scan then loads the mapped rows into the isolated staging area, ready to price.
Every staged loan runs through the same Program & Pricing engines as a single scenario — just isolated and optimized for volume.
A query builder filters and selects which staged loans to run. The engine is invoked in test-only mode (no production writes, no debug-log rows) and prices in single-rate mode — each loan prices its note rate only, not the full rate matrix, a fraction of the work per loan. Batches run in parallel with a configurable degree of concurrency, and one bad row never kills the run: a missing field or unmappable value is captured against its source line, reported, and skipped.
testOnly: true — in-memory pipeline, no loan-table writesParallel.ForEachAsync, configurable parallelismA rollup of the whole batch — and a drill-in to any single loan.
The dashboard breaks the run down by outcome — priced, failed, counter-offer, and no-pricing — so an analyst sees the shape of the tape at a glance and can open any loan's full priced result. Portfolio-level weighted-average statistics — coupon, price, FICO, and LTV — summarize the whole tape, the numbers a desk needs to size a bid. Need to re-run against current pricing? Reprocess is one click — and safely cancellable, restoring the prior results if you stop it mid-run.
Package the priced results into an investor-ready bid — built the way a secondary desk works.
Assemble the response with a drag-and-drop bid-tape builder, then export a Bid Summary PDF organized into sections by plan type — Conforming, Jumbo, and more — each loan with its rate, LTV, FICO, and best-execution result, and each section totaled with its weighted-average coupon, price, FICO, and LTV: the figures an investor actually bids on. From raw tape to a deliverable a desk can hand to investors or management, without leaving the platform.
From a single eligibility check to a full parallel batch — the same engine throughout.
Run a single row through eligibility only — a quick pass/fail badge before committing a full batch.
Price one row end-to-end (Program + Pricing) to inspect a specific loan's result in detail.
Price the whole set as a tracked, parallel job — with live progress and a processing history.
Re-run a job against current pricing — safely cancellable, restoring the prior results if you stop it.
A guided walkthrough and the architecture detail are available to qualified parties in the private data room.