# Evaluate one USD funding research workflow

Start at https://beepboop2025.github.io/financial-evidence-skills/funding/.
Inspect the dated inputs, save a packet and import the CSV into your existing
research process. Public access is free and requires no account.

The intended job is a funding research review: rates, rate distribution,
volumes, reserve balances, Treasury cash and facility usage with evidence dates.
There is no trading execution or promise of investment performance.

## Define a useful evaluation

Send the following to mrinal@liquilens.in when you want to evaluate the workflow
with us. Do not include positions, client names, credentials or confidential
data in a public issue.

- Team and the role reviewing the evidence.
- Existing tool and the current manual process.
- Required series, units and acceptable reporting lags.
- Review deadline, timezone and holiday expectations.
- What would count as a useful improvement, such as time saved or a caught discrepancy.
- Any access, retention, source-rights or procurement requirements.

Use the existing [private scorecard](../workflows/pilot-scorecard.csv) on your own
machine. Record the packet identity, exceptions you reviewed, task completion,
baseline time and time spent with this workflow. Agree the criteria before
evaluating the result. Daily internal use is different from permission to
redistribute data to clients or in another product; retain publisher terms.

## Evidence to look for

1. Source dates, cadence and units fit the research question. Reserves are a
   weekly average; TGA is an opening balance. A common SOFR/IORB date does not
   make the remaining observations simultaneous.
2. Packet hashes verify, and the JSON and CSV refer to the same capture.
3. Missing and contradictory data stays visible. Reopening an older packet
   does not make it a current successful review.
4. The reviewer uses the output on three distinct research days within fourteen,
   records any errors and decides whether it helps the existing workflow.
5. Review the 30-day reliability record once it exists. Separate source delays,
   service failures, missed probes and verified recovery.

The three-day and thirty-day criteria are evaluation targets, not a claim that
a fund has completed them. Anonymous download measurement is optional and does
not identify or verify an organization. A pilot starts only when a participant
agrees its scope; no automatic enrollment, paid contract or SLA is created.

## Access and support

The public workflow receives no portfolio or customer data and needs no API key.
Private datasets, organization entitlements, dedicated capacity, response-time
commitments and warehouse delivery are scoped only when required by a real
participant. Current support is best-effort email to mrinal@liquilens.in;
urgent operational dependence requires a separately agreed support arrangement.