Learn · the people, seat by seat

What does a credit risk analyst do when a vintage goes bad?

What a credit risk analyst does, in one bad vintage: assembles the files, sorts fraud from credit misses from onboarding slips, and gets each fix to the team that owns it — with a fraud strategy manager, a credit policy manager, a head of servicing and a risk director alongside. Early Payment Default Review changes what each of the five seats confirms, not the seats themselves.

Updated · 13 min read

What does a credit risk analyst do, in the week a vintage turns? In one sentence: they find out why loans failed and route each answer to the team that owns the fix. It is not one job but a relay across five seats — the analyst who confirms causes, the fraud strategy manager who closes rings, the credit policy manager who amends rules, the head of loan servicing who repairs the borrowers the bank itself made late, and the credit risk director who signs the report the committee reads. Early Payment Default Review does not replace any of them; it changes what each spends the week confirming instead of re-reading.

The five people on this page are the solution's fictional characters — Priya Raman, Daniel Okoro, Lena Hoffmann, Marcus Bell and Renee Alvarez, at the invented Lanternfield Bank — and their week is sample data. The wages and task descriptions come from the US Department of Labor's O*NET database for the real occupations, linked where they appear.

What the federal data says the two core seats earn

$83,510

median annual wage for credit analysts — $40.15 an hour, 2025 data

$81,100

median annual wage for fraud examiners, investigators and analysts — $38.99 an hour, 2025 data

5seats

one bad vintage runs through — confirm, fix fraud, fix the rule, repair onboarding, sign

O*NET OnLine, sponsored by the US Department of Labor, Employment and Training Administration: Credit Analysts (13-2041.00) and Fraud Examiners, Investigators and Analysts (13-2099.04), 2025 wage data.Source: O*NET OnLine, Credit Analysts 13-2041.00

What does a credit risk analyst do, in one bad vintage?

The occupational definition is dry and accurate: credit analysts "analyze credit data and financial statements of individuals or firms to determine the degree of risk involved in extending credit or lending money" (O*NET, Credit Analysts 13-2041.00). The vintage version of that job is anything but dry. August's book of 4,812 loans has produced 87 early defaults — $1,268,450 at 30 days past due, 1.81% against 1.06% for the six vintages before — and causes are due Friday. The analyst's week is to turn 87 unexplained loans into one explained vintage: every file re-read, every cause proposed with its evidence, every fix drafted for its owner, and nothing sent that they have not confirmed themselves. The senior version of the role carries the same load with the authority to confirm in bulk: 79 causes at the configured confidence threshold together, the 8 below it read one by one.

The list an analyst actually works: 87 proposed causes, 8 left to read

The August early defaults page: 87 early defaults, 87 of 87 causes proposed, 0 confirmed, a review clock with 4 days left. A run log shows how the 87 were explained — the early default monitor assembled the files in 14 seconds, the file re-reader re-read every application in 27, the link analyst linked the 87 across 31,640 loans and 6,214 applications in 22, the credit rule tracer replayed all 87 decisions in 19, the onboarding checker read payment set-up in 18, and the cause and fix writer proposed a cause for each loan in 14. A patterns panel lists ring R-0926 (owner Lena Hoffmann), rule CP-26-07 (owner Marcus Bell) and autopay never saved on Android app 6.14 (owner Renee Alvarez), each with its loans and dollar amount. A button reads Confirm 79 at 90% or more.
Fictional sample run in Early Payment Default Review. The analyst's job on this screen is confirmation: accept the 79 above threshold together, read the 8 below one by one, change any cause the evidence disagrees with.

The five seats: who confirms, who fixes, who signs

Every early default touches all five seats, and the handoffs are the job. The analyst confirms causes and sends fixes. The fraud strategy manager receives the ring — link map, sub-source, open applications to hold, BSA referral. The credit policy manager receives the rule — which change admitted the credit misses, and a what-if replay under any proposed cap. The head of loan servicing receives the borrowers the bank made late, with fees to waive and bureau marks to correct. The consumer credit risk director receives the month: a report that splits the rise into its causes and names each fix and its owner. Nobody signs what they have not read; nothing leaves until the analyst sends it.

The five seats at Lanternfield Bank, and what each still owns

Priya Raman · Senior credit risk analyst

Today

Re-reads a sample of files by hand, weeks later; the rest of the early defaults stay a number in the vintage chart.

With agents

Starts from 87 proposed causes with their evidence, confirms 79 together and reads the 8 below threshold one by one.

Still theirs: Every confirmation, every send — nothing counts as explained on an agent’s proposal alone.

Lena Hoffmann · Fraud strategy manager

Today

Learns about a ring in the quarterly review, after it has moved on.

With agents

Gets the ring the week it is found — link map, sub-source, open applications to hold, and the BSA referral.

Still theirs: The hold, the partner decision and the referral call.

Marcus Bell · Credit policy manager

Today

Is asked to tighten the score cut-off without knowing which rule let the defaults in.

With agents

Sees the rule, the band, and a replay of the vintage under any cap and residual-income floor before he proposes it.

Still theirs: The proposal and the fair-lending review that precedes it.

Renee Alvarez · Head of loan servicing

Today

Learns that autopay was never set up only once the borrower is already 30 days late.

With agents

Gets the 16 borrowers the bank made late, with fees to waive and late marks to correct with the bureaus.

Still theirs: The corrections and the borrower conversation.

Daniel Okoro · Consumer Credit Risk Director

Today

Hears that early defaults went up, not which team fixes it.

With agents

Signs a report that splits the rise into fraud, credit and onboarding, with each fix and its owner.

Still theirs: The signature, and the committee’s question time.

Fictional characters in the sample bank. The pattern they act out is the general one: agents read the files first, and the five seats keep the decisions.

What a credit risk analyst's week looks like when 87 files need an answer

The week has a shape even when it goes well. Monday: the vintage crosses the line and cases open on the servicing feed; the files assemble themselves — application, documents, verification results, decision record, fraud checks, channel, every payment attempt — for each of the 87. Tuesday morning: six agents re-read everything and propose one cause per loan with its evidence and its runner-up; on the sample run that takes a minute and 52 seconds, a sample-run figure to demonstrate the mechanism, not a processing promise. Tuesday afternoon through Thursday: the analyst confirms — 79 together at the configured threshold, the 8 below it one by one, changing any cause the evidence disagrees with. Friday: fixes go to their owners with what each costs, and the report drafts itself from the confirmed record. The whole cycle ran inside the 5-business-day review clock the bank set; the July report, done the same way, took 4 days; the sampled review it replaced took about six weeks to reach a cause — our estimate, not a measured figure.

Where the review days went, sampled by hand against agents-first

Stage of the reviewSampled review by handAgents re-read first
Pulling each file together5.0 days0.5 days
Re-reading applications and verification15.0 days1.0 days
Linking loans that share devices or accounts5.0 days0.5 days
Tracing which rule let the loan in5.0 days0.5 days
Agreeing causes and fixes with owners7.0 days2.0 days
Writing and signing the report5.0 days0.5 days
Total for one monthly vintage42 working days ≈ six weeks (estimated)5 working days
Our estimated split for one monthly vintage: the sampled-review baseline of about six weeks is an estimate, not a measured benchmark, and the agents-first column is a target on the sample vintage.

The same review as shares: re-reading shrinks, deciding stays

Sampled review by hand · 42 working days (estimated)

12%
36%
12%
12%
17%
12%
  • Assembling files · 12%
  • Re-reading · 36%
  • Linking · 12%
  • Tracing rules · 12%
  • Agreeing with owners · 17%
  • Writing the report · 12%

Agents re-read first · 5 working days on the sample vintage

10%
20%
10%
10%
40%
10%
  • Assembling files · 10%
  • Re-reading · 20%
  • Linking · 10%
  • Tracing rules · 10%
  • Agreeing with owners · 40%
  • Writing the report · 10%
The same estimate as the table above, drawn as shares. The shift is not that the week vanishes — it is that the largest share moves from re-reading to agreeing causes and fixes with the people who own them.

What a fraud strategy manager actually owns

The fraud seat is the easiest to misunderstand, because the title sounds analytic and the job is operational. The federal task description runs: fraud examiners "obtain evidence, take statements, produce reports, and testify to findings regarding resolution of fraud allegations", and "analyze financial data to detect irregularities in areas such as billing trends, financial relationships, and regulatory compliance procedures" (O*NET, Fraud Examiners, Investigators and Analysts 13-2099.04). On the sample vintage that translates to: accept the ring referral, pause the partner sub-source FW-2207, hold the two open applications, commission the new device rule (three or more applications in 30 days goes to review), and take the referral decision to the BSA team, whose clock started the day the ring was detected. What the seat does not own is the score cut-off — the ring's applications passed the old scores, and tightening them would have caught nothing while the door stayed open.

What a credit policy manager checks before a rule changes

The policy seat answers a different question: not "who did this" but "what did we change". Before CP-26-07 — the July rule that lifted the debt-to-income cap for scores 640–679 from 40% to 45% — the credit rule tracer replays every decision under the version in force at funding and the one before, and names the change that made the difference: 21 of August's early defaults were approved only under the new cap. Before any amendment, the manager replays the whole vintage under the proposed line: a 42% cap with a $1,800 monthly residual-income floor declines 268 of the 412 loans booked above the old cap and avoids 17 of the 21 early defaults, while the 144 kept default at 2.8%. The OCC expects this discipline to be visible to management, not just exercised: "Policy exception volumes and analysis should be a normal part of board and senior management reporting. Policy compliance is a useful risk management tool, and examiners should view adherence to the exceptions process as an important internal control" (Comptroller's Handbook, Retail Lending). Every proposal from this seat passes fair-lending review before the committee sees it.

Fixes with owners: what each one costs, sent only by the analyst

The Fixes page: four fixes drafted, four sent to their owners, one accepted so far, covering 53 of the 87 early defaults. The first fix, close the ring and the door it came through, is owned by Lena Hoffmann and shows its evidence and cost — $338,400 in the August ring, $41,000 of pipeline applications not funded, 0.6% more applications to review — and is marked Accepted at 02:29. Below it, amend CP-26-07 and end the income waiver goes to Marcus Bell with 17 of 21 early defaults avoided, $3,210,640 of volume given up and a 2.8% rate on the loans kept, and fix the borrowers the bank made late goes to Renee Alvarez with 16 borrowers, $400 of fees waived.
Fictional sample run. Each fix carries its evidence and its cost; owners accept or decline, and nothing leaves until the analyst sends it.

What does a senior credit risk analyst get paid? What federal wage data shows

The federal numbers put the role's midpoint at a $83,510 median annual wage — $40.15 an hour — for credit analysts, and $81,100 for fraud examiners, investigators and analysts, in Bureau of Labor Statistics 2025 wage data as O*NET reports it (O*NET 13-2041.00; O*NET 13-2099.04). A median is the middle of the whole occupation; O*NET also publishes state and local wages for each, which is the comparison a senior analyst or a hiring manager should make. The premium the market pays for is judgment under deadline — the version of the job this page describes, where 87 files need an answer by Friday and the answer has to survive an examiner.

Is credit risk analyst a good career, and what does it take?

It is a good career for a particular temperament: people who like documents more than drama, and who want their work to be checked rather than applauded. A common entry path runs through a bachelor's degree in finance, economics or accounting plus a few years in underwriting or analytics; the recognized certificates — the CFA charter for the markets side, FRM for the risk side, CFE for the fraud side — mark specialisms rather than requirements. The work's direction of travel is visible in this series itself: the analyst who once spent weeks assembling and re-reading files now spends the week confirming what software proposes and owning what it cannot. The skills that appreciate in that world are the ones O*NET's description quietly assumes — reading a file skeptically, writing down why, and standing behind the number.

What changes for these five roles once agents read the files first

The seats stay; the week changes. Where the sampled review spent its effort reading a quarter of the files and extrapolating, the agents-first week starts from all of them proposed and cited, and the people spend their hours on the decisions only people can make: confirming causes, holding applications, amending rules, correcting bureau marks, signing the report. Fannie Mae's quality-control language states the supervision-side logic in one sentence — a statistically valid random sample, "augmented with targeted, discretionary sampling, which aids in the measurement of the overall quality of mortgage loan deliveries" (Servicing Guide, A2-4-01). The discretionary part is the point: when the review population itself is chosen by risk, the human seats spend their attention where it matters, and the agents' job is to make sure that is where the evidence already is.

One vintage across the five seats: who hands off to whom

  1. Agent

    Six agents

    Re-read all 87 files and propose one cause each, with evidence, runner-up and confidence.

    Never: confirm a cause or send a fix

  2. A person decides

    Priya Raman · analyst

    Confirms 79 together at the configured threshold, reads the 8 below one by one, sends each fix to its owner.

  3. A person decides

    Lena · Marcus · Renee

    Accept or decline their fixes: pause the sub-source, amend the rule after fair-lending review, correct the borrowers the bank made late.

  4. A person decides

    Daniel Okoro · director

    Signs the monthly report: the rate, the causes, each fix and its owner, for the credit risk committee.

The handoffs on the fictional August vintage. Every step is recorded — who proposed, who confirmed, who accepted, who signed — on the audit trail.

The month, signed: what the committee actually reads

The October early default report for the credit risk committee, drafted by the cause and fix writer and confirmed by the analyst. Its summary reads: August's early default rate was 1.81% against a 1.06% average and a 1.25% appetite, and the rise is not one problem but three — a fraud ring through one partner source, a July credit rule and an app defect that made 13 borrowers late. A cause table lists fraud 34, credit 29, onboarding 16, hardship 8 with loans, dollars, share and detail. A section removes the ring and the app defect to show the underlying 1.08%. The sign-off panel shows Priya Raman confirmed all 87 causes and sent the report, and Daniel Okoro signed it, with the line Explained in 2 days · July report took 4 days · a sampled review used to take about 6 weeks.
Fictional sample report. What the director signs is the decomposition — causes, fixes, owners — not just the rate.

The signature is where the relay ends. In the sample run the October report went to the committee on Tuesday 10 November: the rate, the four causes, the fixes and their owners, and the thread that ran through all three problems — income verification waived on the instant-decision path, 41% of the early defaults. What the other five guides in this series cover is the work behind each seat — the term itself, the ring, and the six agents doing the reading. To see the five seats on a working vintage, start from Early Payment Default Review.

Questions people ask

What does a credit risk analyst do?

They analyze credit data and financial statements to judge the risk of extending credit — and, when a vintage turns, they find out why loans failed and route each cause to the team that owns the fix. In an early-default review that means confirming a proposed cause for every loan and sending each fix to its owner.

What does a senior credit risk analyst do differently?

The same work with confirmation authority: reading the files below the confidence threshold one by one, changing causes the evidence disagrees with, and owning what is sent. On the sample vintage the senior analyst confirms 79 causes together and personally reads the remaining 8.

What is a fraud analyst's job description?

In the federal occupation's words: obtaining evidence, taking statements, producing reports and testifying to findings on fraud allegations, and analyzing financial data for irregularities in billing trends, financial relationships and compliance procedures. In a lending operation, add the operational calls — pausing a sub-source, holding pipeline applications, referring a ring to the BSA team.

What does a credit policy manager check before changing a rule?

Which rule admitted the losses: a replay of each decision under the policy version in force and the one before, the band the change touched, and a what-if replay of the whole vintage under any proposed cap — volume given up, early defaults avoided, rate on the loans kept. Every proposal passes fair-lending review before the committee sees it.

Is credit risk analyst a good job?

For people who like documents, deadlines and accountability, yes: median pay of $83,510 nationally in 2025, a clear seniority ladder, and skills — skeptical file-reading, defensible writing — that appreciate as automation takes the mechanical share of the work.

Does the analyst still matter if agents read the files first?

More, not less: the agents propose, and every proposal waits on a person's confirmation. The analyst's week moves from re-reading a sample of files to deciding all of them — and the audit trail records whose decision each cause was.

Early Payment Default Review.
See it on your own files.

An agentic solution from Botminds. Every loan that misses an early payment is re-read end to end, and agents say whether it was fraud, a credit miss or an onboarding slip. Book a demo, or explore how it works.