01 · FREE SCOPING
Freeze the question
Choose one book, one economic buyer and one baseline process. Confirm which historical outcomes can be used without moving customer data.
For banks, NBFCs and credit teams · fixed-scope pilot
Prove the warning on your own book before you buy.
Run LiquiLens in parallel with your existing DPD, review or counterparty process for six weeks. The work runs inside your environment, against success criteria agreed before the first score.
Reply from the founder · no sales handoff · no customer rows required for the first scoping call
6 weeksshort enough to finish, long enough to observe the workflow
4 measuresdetection, lead time, false positives and alerts per catch
0 live actionsparallel run only; humans retain every decision
1 decisionconvert only if the agreed bar is met
The work
A proof, not an open-ended implementation.
A paid pilot protects both sides from the usual failure mode: months of custom work with no agreed definition of success. Scope, data boundary and decision date are written down before deployment.
02 · PARALLEL RUN
Score without acting
LiquiLens runs beside the current workflow. No alert changes a limit, classification or collection action during the proof.
03 · DECISION PACK
Keep the evidence
You receive the backtest, misses, alert-budget trade-off, model card, controls and a production recommendation whether you continue or not.
Success criteria
The bar is agreed before the result is known.
The pilot does not win by producing the most alerts. It wins only if it finds useful warning earlier than the current process at a workload the team can absorb.
Early-detection rateHow many later adverse outcomes were flagged before the reactive baseline?
Median lead timeHow many usable days of warning did the first valid flag create?
False-positive rateHow many performing cases consumed review capacity?
Alerts per catchHow much operating work was required for each genuinely early detection?
Commercial ladder
The public proof stays free. Private-book work is paid.
Public
Evidence layer
Free
Live public boards, historical replays, REST, MCP and the terminal CLI. Use it to audit the method before a sales conversation.
Connect a public surface →Proof
Parallel pilot
₹2.5 lakh
Six weeks, one defined book, self-hosted by default. The full fee is credited toward the first annual production contract.
Scope the pilot →Production
Annual deployment
From ₹12 lakh/yr
One production book, governed updates and operating support. New connectors, custom integration and additional books are scoped separately.
Discuss production →Fit
This pilot is useful when three things are true.
A real owner
A risk, credit, finance or treasury leader owns both the current workflow and the go/no-go decision.
A measurable history
The team can provide a controlled historical extract with outcomes, timestamps and the existing baseline.
A deployable boundary
There is a sandbox or supervised environment where the code can run without copying the book to LiquiLens.
Why is the pilot paid?
It includes deployment, a frozen study design, an honest comparison and a filable governance pack. The fee keeps the work bounded and is credited in full if the product earns a production contract.
Does LiquiLens make credit decisions during the pilot?
No. This is a parallel screen. It does not approve, reject, downgrade, classify or execute. A named human retains every decision.
Does our book leave our environment?
Self-hosted is the default. The first scoping call requires no customer rows, and the pilot agreement defines the exact data boundary before access.
What if the agreed bar is missed?
The pilot stops at week six. You keep the result, including the misses and the reason the bar was not met. There is no production obligation.
Next step
Bring one book and one baseline. We will define the proof together.
Send the institution, your role, approximate book or AUM size, current early-warning process and preferred deployment boundary. You will get a scoped response from the founder.