Methodology
How Trovavo's Financial Operating System works.
No magic. A 23-signal forensic read of your finances, a 4-agent AI pipeline that gates every answer, and a public scorecard you can hold us to. If a draft can't pass the critic, it doesn't ship — full stop.
In plain English
What this means for you.
What it does
Reads your financial data the way a forensic analyst would — checking 23 separate signals across five risk areas — and translates the findings into clear, actionable language powered by your AI guide.
Why it helps you
You stop guessing what's hurting your financial health, which obligation is draining you, and what to do next. Every recommendation is checked by a separate critic agent before you ever see it.
What you get
Your Financial Health Score, Resolution Opportunities, AI-guided communications, and a step-by-step roadmap — explained so you always know the next move.
The 23 parameters
23 signals. Five groups. One verdict.
Most "free credit checks" return a number. A Forensic Audit returns a verdict on each of the twenty-three signals below, plus the exact dispute or restructure action you can take next.
Group 01 · 4 checks
Identity & profile integrity
PAN ↔ CIBIL name match
Exact-string + fuzzy match against PAN database to catch typos that block lender pulls.
DOB / gender consistency
Cross-checked across TransUnion, Experian and Equifax records.
Address history vs. KYC trail
Stale or contradicting addresses flagged as verification drag for new loans.
Phone & email linkage age
How long each contact has been on file — recent changes trigger lender re-verification.
Group 02 · 5 checks
Account-level health
Active vs. closed split
Total open lines, closed lines, and any 'closed' still incorrectly reporting.
Secured vs. unsecured ratio
Exposure mix — too heavy on unsecured hurts pricing across the board.
Oldest tradeline age
Depth of credit history. Closing your oldest card silently lowers the score.
Average account age
Mean age across active lines — a quick proxy for credit maturity.
Enquiry density (90d)
Hard pulls clustered in 90 days flagged as loan-shopping behaviour.
Group 03 · 4 checks
Repayment behaviour
DPD heat-map (36 months)
Every late payment, with lender, amount, and bucket (1–30, 31–60, 60+).
Settled / WO / Suit-filed
Auto-reject triggers at most lenders — surfaced first with exact wording.
Restructured / OTS markers
Post-COVID and current-cycle restructures flagged for narrative repair.
Revolver vs. transactor
Card behaviour pattern — revolvers price worse on new unsecured products.
Group 04 · 4 checks
Utilisation & exposure
Aggregate utilisation
Total balance vs. total limit across all cards — 30% is the safe ceiling.
Per-card outliers (>80%)
Individual cards near limit even when aggregate looks fine.
EMI-to-take-home (FOIR)
Banks reject above 50–55%. We compute against your real net pay.
Unsecured leverage
Unsecured-to-secured ratio — the single biggest driver of personal-loan APR.
Group 05 · 6 checks
Bureau-data hygiene · dispute candidates
Duplicate tradelines
Same loan reported under two lender codes — common after portfolio sales.
Stale 'closed' accounts
Loans you closed years ago still reporting as active or running.
Wrong DPD vs. statements
Bureau says you were late; statements prove you weren't. Letter-ready.
Mis-tagged ownership
Joint loans showing as sole, or vice-versa — affects your borrowing headroom.
Enquiries you didn't authorise
Unrecognised hard pulls flagged for an RBI dispute and bureau takedown.
Pre-CIBIL-2.0 legacy flags
Old defaults that should have aged out but didn't — disputable under current norms.
The pipeline
Four agents. One answer. Critic has veto power.
Step 01
Router
Intent classification
Audit · dispute draft · settlement · restructure — picked by a cheap model.
Track selection
Routes to the right persona and depth tier before any user-facing token is spent.
No output to user
Silent step. Logged for analytics so we can tune intent classes monthly.
Step 02
Specialist
The actual draft
Expert persona — CIBIL auditor, RBI-dispute drafter, settlement strategist, or restructure analyst.
Cited reasoning
Every claim ships with the rule, circular, or statement line it leans on.
Token-budgeted
Bounded output length — no rambling, no fabricated padding.
Step 03
Compliance critic
Hard gate
If a draft can't pass, it doesn't ship. Full stop. The user sees a retry, not a bad answer.
Citation verifier
Refuses fabricated RBI / CIBIL references — every cite is matched against our circular index.
Outcome-promise blocker
Anything implying a guaranteed score bump or settlement is killed at this step.
Step 04
Synthesiser
Plain-language pass
Tightens hedging, removes jargon, formats for human reading.
Only post-critic
Never runs on a draft the critic flagged — prevents polishing a bad answer into believability.
Final format
PDF-ready letters, markdown for the app, structured JSON for the API.
The critic is not cosmetic. Every model run is logged to signals with critic_passed and critic_flags. A provider that fails too often is demoted automatically from the router — visible in our internal provider scorecard.
What we won't do