CFPB Complaint Intelligence Platform

Screen vendors, detect consumer risk, and defend your decisions — from 5M+ CFPB complaints

Most CFPB tools return complaints. This actor returns decisions you can defend.

• Pass / Review / Fail vendor screening
• Complaint Intelligence Score (0-100) + A-F grade
• Decision evidence — exactly why each verdict landed, so you can defend every call
• Executive summaries — paste-ready, no LLM
• Monitoring alerts — what changed since your last run
• Audit-ready due-diligence packs

Built for: vendor-risk teams · procurement · third-party-risk programs · compliance & AML · M&A due diligence · fintech competitive intelligence · journalists & researchers.

No API key, no LLM, no infrastructure — just deterministic, structured decisions.

---

Why not use the CFPB API directly?

The raw CFPB API gives you complaints:

{ "company": "ABC Bank", "issue": "Identity theft", "timely": "Yes" }

This actor gives you a decision:

{
  "company": "ABC Bank",
  "screeningResult": "review",
  "complaintIntelligenceScore": 74,
  "grade": "D",
  "primaryArchetype": "Complaint Accelerator",
  "decisionEvidence": {
    "primaryDrivers": ["accelerating-complaints", "high-dispute-rate"],
    "impactWeights": { "handling": 32, "severity": 18, "velocity": 8, "concentration": 16 }
  },
  "executiveSummary": { "headline": "Complaint risk increasing", "summary": "ABC Bank: Complaint Intelligence Score 74/100 (high risk, grade D), driven by accelerating-complaints and high-dispute-rate. 12% untimely and 19% disputed across 240 complaints." }
}

The CFPB API gives you data. This actor gives you risk signals you can act on — and the evidence to defend the call.

---

What decision are you trying to make?

Your question	Use	You get
Should I onboard this vendor?	workflow: "vendor-screening"	Pass/Review/Fail + Complaint Intelligence Score + decision evidence, per company
Has this vendor's risk changed?	workflow: "vendor-monitoring"	What changed since last run, risk events, complaint spikes
Deep due diligence on one company?	workflow: "due-diligence"	A board-ready pack: exec summary, score, evidence, trend, top issues
I need evidence for a review committee	workflow: "audit-evidence"	Due-diligence pack + the underlying complaint records as evidence
What's happening across an industry?	workflow: "competitive-monitoring"	Cohort leaderboard, cohort averages, market-level regulatory-risk signals

Pick a workflow and the actor configures the right mode for you. Power users can still drive outputMode directly (it always wins).

---

Vendor screening in 30 seconds

Input:

{
  "workflow": "vendor-screening",
  "companies": ["Company A", "Company B", "Company C"],
  "maxResults": 200
}

Output (one companySummary row per company, ranked by risk):

[
  { "company": "Company A", "screeningResult": "pass",   "complaintIntelligenceScore": 18, "grade": "B" },
  { "company": "Company B", "screeningResult": "review", "complaintIntelligenceScore": 52, "grade": "C" },
  { "company": "Company C", "screeningResult": "fail",   "complaintIntelligenceScore": 81, "grade": "F" }
]

Plus a run summary naming the lowest-risk and highest-risk vendor, and riskEvent records for anything that spiked.

---

CFPB API vs this actor

Capability	Raw CFPB API	This actor
Complaint search & filtering	✅	✅
Aggregation / leaderboards	❌	✅
Complaint Intelligence Score + A-F grade	❌	✅
Pass / Review / Fail vendor screening	❌	✅
Decision evidence (why)	❌	✅
Cross-run monitoring + "what changed"	❌	✅
Cohort benchmarking + real product baseline	❌	✅
Market-level regulatory-risk signals	❌	✅
Executive summaries	❌	✅
Due-diligence + audit packs	❌	✅

---

Built on the official CFPB consumer complaint database — the largest public source of US consumer financial grievances (5M+ complaints since 2011). This actor handles pagination, rate limiting, and normalization for you, then layers the intelligence on top.

---

Defend every decision

Decision evidence is the difference between a score and a recommendation a compliance team can stand behind. Every screening verdict ships with the exact drivers and their weighted contribution:

{
  "company": "ABC Bank",
  "screeningResult": "fail",
  "complaintIntelligenceScore": 81,
  "grade": "F",
  "recommendedAction": "escalate",
  "decisionEvidence": {
    "primaryDrivers": ["high-dispute-rate", "accelerating-complaints", "high-severity-mix"],
    "impactWeights": { "handling": 38, "severity": 22, "concentration": 13, "velocity": 8 },
    "topComponent": "handling"
  }
}

When someone asks "why did you flag this vendor?", you have the answer in the record — no black box, no LLM, fully reproducible.

---

What changed since last run

Run with a watchlistName on a schedule and the actor tells you what moved, not just the current state:

{
  "changes": [
    { "company": "Equifax", "change": "rank-up", "previousRank": 7, "currentRank": 2 },
    { "company": "Acme Lending", "change": "complaints-increased", "newComplaints": 31 }
  ],
  "portfolio": { "portfolioRiskScore": 64, "newCriticalAlerts": 2, "portfolioTrend": "worsening" }
}

Plus riskEvent records for spikes, rapid deterioration, and market-level regulatory-risk themes.

---

Typical workflows

Vendor onboarding — screen 20 vendors in one run, identify the highest-risk supplier, generate an audit trail:

{ "workflow": "vendor-screening", "companies": ["Vendor A", "Vendor B", "..."] }

Quarterly vendor review — monitor a saved portfolio, detect deteriorating vendors, escalate complaint spikes:

{ "workflow": "vendor-monitoring", "portfolioName": "vendors", "watchlistName": "q-review" }

M&A due diligence — build a board-ready pack for a target with evidence and trend:

{ "workflow": "due-diligence", "company": "Acquisition Target Inc" }

---

Outputs designed for automation

Every decision is returned as stable, structured fields — no parsing, no prompt engineering, deterministic across runs:

{ "screeningResult": "pass", "recommendedAction": "monitor", "riskLevel": "low", "grade": "B" }

Branch on screeningResult, recommendedAction, riskLevel, or eventType directly in Zapier, Make, n8n, Dify, or an agent tool call. The enums are additive within a major version, so your rules don't break.

---

Why use this Complaint Intelligence Platform?

Built on a CFPB consumer complaint search, but it returns decisions, not just rows:

• No API key required -- The CFPB API is a free US government data source. No authentication, no signup, no usage limits.
• Decisions, not data -- Every company comes back with a Complaint Intelligence Score, an A-F grade, a pass/review/fail screening verdict, decision evidence, and a plain-English executive summary.
• Eleven output modes -- from raw records and leaderboards to monitoring, cohort benchmarking, due-diligence packs, audit packs, industry profiles, and time-series. One actor, the mode for the question you're asking.
• Up to 10,000 records per run -- Auto-pagination handles the entire matching result set within the API's 10,000-row single-query limit; time-series and trend modes use exact per-period counts to sidestep the cap entirely.
• Cloud scheduling -- Run on a recurring schedule with a watchlist to build company memory, detect deterioration, and surface what changed since the last run.
• No infrastructure needed -- Runs entirely on Apify's cloud. No servers, no Docker setup, no dependencies to install.
• Direct complaint links -- Every record-mode result includes a generated URL linking directly to the complaint detail page on the CFPB website.

---

Data foundation (the table stakes)

The retrieval layer everything is built on — search by keyword, company, product, US state, and date range; sort by recency or relevance; up to 10,000 complaints per run with built-in pagination and auto-pagination; consumer narratives where disclosed; a direct CFPB URL per complaint; and a polite delay between API pages. The intelligence layer above (scoring, screening, monitoring, evidence) is what you actually run it for.

---

How to use

Apify Console

1. Go to the CFPB Consumer Complaint Search actor page on Apify.
2. Click Start to open the input configuration.
3. Enter a search term, company name, or select a product type. You can combine multiple filters.
4. Optionally set a date range and state to narrow results.
5. Choose your sort order and maximum number of results.
6. Click Start to run the actor.
7. When finished, download results from the Dataset tab in JSON, CSV, Excel, or other formats.

Python

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")

run = client.actor("ryanclinton/cfpb-consumer-complaints").call(run_input={
    "searchTerm": "unauthorized charges",
    "company": "Wells Fargo",
    "product": "Checking or savings account",
    "state": "CA",
    "dateFrom": "2024-01-01",
    "dateTo": "2024-12-31",
    "sortBy": "created_date_desc",
    "maxResults": 100,
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(f"{item['complaintId']} | {item['company']} | {item['issue']}")

JavaScript

import { ApifyClient } from "apify-client";

const client = new ApifyClient({ token: "YOUR_API_TOKEN" });

const run = await client.actor("ryanclinton/cfpb-consumer-complaints").call({
    searchTerm: "identity theft",
    company: "EQUIFAX, INC.",
    sortBy: "created_date_desc",
    maxResults: 50,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(`Found ${items.length} complaints`);
items.forEach((item) => console.log(`${item.complaintId}: ${item.issue}`));

---

Input parameters

Parameter	Type	Required	Default	Description
workflow	Select	No	--	Job-named entry point: vendor-screening / vendor-monitoring / due-diligence / audit-evidence / competitive-monitoring. Configures the right outputMode for you. An explicit outputMode always wins.
searchTerm	String	No	--	Keyword search across complaint narratives and fields (e.g., "credit report", "late fee", "identity theft")
company	String	No	--	Filter by a single company name (e.g., "Bank of America", "Wells Fargo", "Equifax")
companies	String[]	No	--	Screen multiple companies in one run (records mode). One filtered query per company (maxResults applies per company), capped at 25. Overrides company when set.
product	Select	No	Any	Financial product type: credit reporting, debt collection, mortgage, credit card, checking/savings, student loan, vehicle loan, money transfer, payday loan, personal loan
state	String	No	--	US state code, automatically uppercased (e.g., "CA", "NY", "TX")
dateFrom	String	No	--	Start date for complaints received (YYYY-MM-DD format)
dateTo	String	No	--	End date for complaints received (YYYY-MM-DD format)
sortBy	Select	No	created_date_desc	Sort order: Newest first, Oldest first, Most relevant, Least relevant. Records mode only.
maxResults	Integer	No	50	Maximum complaints to return (1--10000). In aggregation mode, caps how many buckets are emitted.
outputMode	Select	No	records	records / aggregation / compare / monitor / cohort / dueDiligence / auditPack (pack + evidence + records) / industryProfile / timeseries / trend / stateRisk
watchlistName	String	No	--	Track companies across scheduled runs — each run reports newComplaintsSinceLastRun, a change flag, a persistent companyProfile (first seen / runs / all-time-high score / score history / trend) and rankChange.
portfolioName	String	No	--	Save a reusable company list under a name (pass with companies to save; pass alone later to reload). Pair with watchlistName + a schedule for hands-off portfolio monitoring.
trendIntervalMonths	Integer	No	12 / 3	timeseries: trailing months to chart (default 12). trend: current-period length in months (default 3); the previous equal window is compared.
aggregateBy	Select	No	company	When outputMode=aggregation: company, product, sub_product, issue, sub_issue, state, company_response, submitted_via, tags, or timely
autoPaginate	Boolean	No	false	When true and total matching > maxResults, fetch every matching complaint up to the 10,000 single-query cap. Records mode only.
enableCompanySummary	Boolean	No	false	Adds one company-risk-summary row per company in the result set plus a run summary, ranked by complaint-handling risk. Records mode only; not charged.
outputProfile	Select	No	standard	minimal (headline decision fields only), standard, or full. Records mode only.

Example input — records mode (JSON)

{
    "outputMode": "records",
    "searchTerm": "credit report",
    "company": "EQUIFAX, INC.",
    "product": "Credit reporting, credit repair services, or other personal consumer reports",
    "state": "FL",
    "dateFrom": "2024-06-01",
    "dateTo": "2024-12-31",
    "sortBy": "created_date_desc",
    "maxResults": 100
}

Example input — aggregation mode (top companies in 2024)

{
    "outputMode": "aggregation",
    "aggregateBy": "company",
    "dateFrom": "2024-01-01",
    "dateTo": "2024-12-31",
    "maxResults": 25
}

Tips: Combine multiple filters for precise results. Use date ranges for trend analysis. When using a keyword search in records mode, try sorting by "Most relevant first" for best matches. For "top N" leaderboard questions over millions of complaints, use aggregation mode — one HTTP call instead of thousands. If no filters are set, the actor returns the most recent complaints (records mode) or the global top buckets (aggregation mode).

---

Output

Every row carries a kind discriminator: "record" for individual complaints, "aggregation" for count buckets. Filter kind === "aggregation" downstream to isolate leaderboard rows.

Example output — records mode (JSON)

{
    "kind": "record",
    "complaintId": "8234561",
    "dateReceived": "2024-11-15",
    "dateSentToCompany": "2024-11-15",
    "product": "Credit reporting, credit repair services, or other personal consumer reports",
    "subProduct": "Credit reporting",
    "issue": "Incorrect information on your report",
    "subIssue": "Information belongs to someone else",
    "company": "EQUIFAX, INC.",
    "companyResponse": "Closed with explanation",
    "companyPublicResponse": "Company has responded to the consumer and the CFPB and chooses not to provide a public response",
    "state": "FL",
    "zipCode": "331XX",
    "narrative": "I discovered several accounts on my Equifax credit report that do not belong to me. I have filed disputes multiple times but the inaccurate information remains.",
    "consumerDisputed": "N/A",
    "timely": "Yes",
    "submittedVia": "Web",
    "cfpbUrl": "https://www.consumerfinance.gov/data-research/consumer-complaints/search/detail/8234561",
    "extractedAt": "2025-06-10T14:32:08.123Z"
}

Example output — aggregation mode (JSON)

Top companies by complaint count for 2024 (aggregateBy: "company", dateFrom: "2024-01-01", dateTo: "2024-12-31"):

[
    { "kind": "aggregation", "aggregateBy": "company", "rank": 1, "term": "TRANSUNION INTERMEDIATE HOLDINGS, INC.", "count": 790668, "share": 0.2891, "totalMatching": 2734293, "queryFilters": { "searchTerm": null, "company": null, "product": null, "state": null, "dateFrom": "2024-01-01", "dateTo": "2024-12-31" }, "extractedAt": "2026-05-08T10:00:00.000Z" },
    { "kind": "aggregation", "aggregateBy": "company", "rank": 2, "term": "EXPERIAN INFORMATION SOLUTIONS INC.", "count": 754851, "share": 0.2761, "totalMatching": 2734293, "queryFilters": { "searchTerm": null, "company": null, "product": null, "state": null, "dateFrom": "2024-01-01", "dateTo": "2024-12-31" }, "extractedAt": "2026-05-08T10:00:00.000Z" },
    { "kind": "aggregation", "aggregateBy": "company", "rank": 3, "term": "EQUIFAX, INC.", "count": 752722, "share": 0.2753, "totalMatching": 2734293, "queryFilters": { "searchTerm": null, "company": null, "product": null, "state": null, "dateFrom": "2024-01-01", "dateTo": "2024-12-31" }, "extractedAt": "2026-05-08T10:00:00.000Z" }
]

Output fields — records mode

Field	Type	Description
kind	String	Always "record" in records mode
complaintId	String	Unique CFPB complaint identifier
dateReceived	String	Date the CFPB received the complaint (YYYY-MM-DD)
dateSentToCompany	String	Date the complaint was forwarded to the company
product	String	Financial product or service category
subProduct	String	More specific product subcategory
issue	String	Primary issue described in the complaint
subIssue	String	More specific issue detail
company	String	Name of the company the complaint is against
companyResponse	String	How the company responded (e.g., "Closed with explanation", "Closed with monetary relief")
companyPublicResponse	String	The company's optional public-facing response text
state	String	Two-letter US state code where the consumer is located
zipCode	String	Consumer's ZIP code (may be partially redacted)
narrative	String/null	Consumer's written description of the complaint, or null if not provided or consent not given
consumerDisputed	String	Whether the consumer disputed the company's response
timely	String	Whether the company responded in a timely manner ("Yes" or "No")
submittedVia	String	Submission channel (Web, Phone, Referral, Postal mail, Fax, Email)
cfpbUrl	String	Direct link to the complaint on the CFPB website
extractedAt	String	ISO 8601 timestamp of when the data was extracted
productCategory	String	Product mapped to a stable category enum (credit-reporting, debt-collection, mortgage, credit-card, bank-account, student-loan, auto-loan, money-transfer, payday-loan, personal-loan, other)
responseType	String	Normalized company response (closed-with-monetary-relief, closed-with-explanation, untimely, …)
reliefProvided	Boolean	Whether the company provided monetary or non-monetary relief
timelyResponse	Boolean	Whether the company responded on time
hasNarrative	Boolean	Whether the complaint carries a consumer narrative
severityFlags	String[]	Stable flags: untimely-response, consumer-disputed, relief-provided, no-relief
dataCompleteness	Integer	0-100 share of key fields populated on this record
severityScore	Integer	0-100 weighted severity of the complaint type (identity theft, foreclosure, fraud, harassment, …) — distinct from how the company handled it
severityClass	String	critical / high / moderate / low
severityFactors	String[]	Which severity rules fired: identity-theft, fraud, foreclosure, repossession, debt-harassment, legal-action, …

Output fields — company summary (when enableCompanySummary is on)

One companySummary row per company in the result set, ranked by complaint-handling risk, plus one runSummary row. All scores are within-result-set signals. Complaint volume is heavily confounded by company size (large credit bureaus and banks naturally generate more complaints), so the risk score deliberately leans on response-quality rates and velocity, not raw count — and is never a cross-universe "this company is risky overall" rating.

Field	Type	Description
kind / recordType	String	"companySummary"
company	String	Company display name
companyKey	String	Stable hash of the normalized name for cross-run joins
complaintCount	Integer	Complaints for this company in the result set
complaintIntelligenceScore	Integer	Signature 0-100 holistic score (higher = more concerning): handling 50% + complaint-type severity 25% + issue concentration 15% + velocity 10%. The leaderboard ranks on this.
grade	String	A-F letter grade from the intelligence score
complaintRiskScore	Integer	Complaint-handling sub-score (untimely + dispute + low-relief + velocity) — a transparent driver of the intelligence score
signals	String[]	Company-level signal tokens: accelerating-complaints, high-dispute-rate, high-severity-mix, issue-concentration, no-monetary-relief, …
decisionEvidence	Object	Why the score is what it is — primaryDrivers[] + impactWeights (the actual point contribution of handling / severity / concentration / velocity) + topComponent
executiveSummary	Object	Deterministic, paste-ready { headline, summary } narrative — no LLM
themeExposure	Object	Share of the company's complaints per regulatory theme (fraud, identity-theft, mortgage-servicing, credit-reporting, …)
primaryArchetype	String	Single memorable label: Complaint Accelerator / Dispute Magnet / Slow Responder / High-Severity Risk / Single-Issue Concentrated / Consumer-Friendly Resolver / Standard Responder
companyProfile	Object	(watchlist mode) persistent memory — firstSeen, runsSeen, highestIntelligenceScore, scoreHistory[], intelligenceTrend
intelligenceTrend	String	(watchlist mode) improving / worsening / stable / new vs the last run
vsIndustryBaseline	Object	(product filter set) this company's untimely/relief rate ÷ the real product-wide rate
monetaryReliefRate	Number	Share of complaints closed with monetary relief (distinct from total relief)
relativeRisk / cohortComparison	Number / Object	(cohort mode) intelligence score ÷ cohort average, and a worse/near/better-than-cohort position
previousRank / rankChange	Integer	(watchlist mode) prior-run risk rank and the change since (positive = moved up the leaderboard)
riskLevel	String	minimal / low / medium / high / critical
complianceRating	String	A-F letter grade (alias of grade)
screeningResult	String	Vendor-screening verdict: pass / review / fail
recommendedAction	String	monitor / review / escalate
confidence	String	How much to trust the verdict given sample size: high (≥20 complaints) / medium (≥5) / low
verdict	String	One-sentence plain-English risk verdict, ready for a report or Slack message
riskFactors	String[]	Codes that drove the score (high-untimely-rate, dispute-heavy, low-relief-rate, accelerating-complaints)
archetype	String[]	Labels: Poor Responder, Dispute-Heavy, Consumer-Friendly Resolver, Rising Complaints, …
untimelyRate / reliefRate / disputeRate	Number	Response-quality rates (0..1)
severityScore / maxSeverityScore / criticalComplaintCount	Integer	Average & worst complaint-type severity, and count of critical-severity complaints
topIssueConcentration / concentrationRisk	Number / String	Share of the company's complaints from its single biggest issue, and the high/medium/low band — high concentration reads as a systemic weakness
trend	String	12-month velocity: accelerating / steady / declining / insufficient-data
totalMatchingForCompany	Integer	Exact CFPB total matching the company's query (compare / companies[] mode); null on broad searches
watchlistDelta	Object	When watchlistName is set: { changeFlag, newComplaintsSinceLastRun, riskScoreDelta, … } vs the last run
timeline	Object[]	Per-year complaint counts
topProduct / topIssue	String	Most common product category / issue for this company
rankInResults / percentileInResults	Integer	Risk rank and percentile within the result set

The runSummary row carries a headline + oneLine takeaway, riskDistribution, screeningDistribution, highestRiskCompany, and an exceptions[] list of the fail-screened companies with reasons. It is also mirrored to the SUMMARY key-value store key.

Output fields — aggregation mode

Field	Type	Description
kind	String	Always "aggregation" in aggregation mode
aggregateBy	String	Field used for grouping (e.g., "company", "product", "state")
rank	Integer	1-based rank within the result set (1 = largest count)
term	String	The bucket label (e.g., "EQUIFAX, INC.", "Mortgage", "CA")
count	Integer	Number of complaints in this bucket
share	Number	count / totalMatching — fraction of all matching complaints in this bucket (0..1)
totalMatching	Integer	Total complaints matching the query filters across all buckets
queryFilters	Object	Echo of the filters that produced this aggregation, for traceability
extractedAt	String	ISO 8601 timestamp of when the aggregation was computed

Output — monitoring & analytical modes

timeseries (one timeseries row): series[] of { period: "YYYY-MM", count } (exact monthly counts, computed via one count query per month so the 10,000-row cap never applies), plus total, velocityScore (0-100, 50 = on-average), velocityClass (spike / rising / steady / dropping), and changePercent / trend.

trend (one trend row): currentPeriodComplaints vs previousPeriodComplaints (exact counts for two equal-length windows), changePercent, trend, severity, and emergingIssues[] — issues that are new or grew ≥50% vs the previous period ({ issue, currentCount, previousCount, growthRate, trend }).

stateRisk (one stateRisk row per state): complaints, population (bundled US Census 2020), complaintsPer100k (normalized rate — far more comparable than raw counts), and rank.

compare = records mode with the company risk leaderboard forced on (see the company-summary fields above), ranked by the Complaint Intelligence Score with exact per-company totals.

monitor (needs watchlistName): emits only the companies whose complaint total changed since the last run, plus riskEvent records. Each riskEvent: { eventType, severity, company, detail, changePercent } where eventType is complaint-spike / risk-escalation / rapid-deterioration / high-severity-company / critical-screening (company-level) or regulatory-risk (market-level — a regulatory theme like identity-theft dominant across many companies, with theme + warningLevel + affectedCompanies). Paste-ready for Slack/Zapier/PagerDuty.

cohort: the company leaderboard plus a cohortStats block on the run summary (cohortAverageDisputeRate, cohortAverageIntelligenceScore, cohortLeader/cohortLaggard) and a per-company relativeRisk + cohortComparison. Averages are across the companies in your run, never a fabricated cross-industry figure.

dueDiligence (one dueDiligence row): a board-ready pack for a single company — executiveSummary (paste-ready {headline, summary}), complaintIntelligenceScore + grade + screeningResult, decisionEvidence, trendAnalysis, topIssues[], watchItems[], supportingMetrics, recommendedAction, vsIndustryBaseline, and the persistent companyProfile.

auditPack (one auditPack row + the company's complaint records): the due-diligence pack with the underlying complaint records kept in the dataset as evidence — built for "download → attach to the vendor review file."

When you run with a watchlistName, the runSummary adds a portfolio block (portfolioRiskScore, criticalAlerts, newCriticalAlerts, portfolioTrend) and a changes[] feed — the "what changed since last run" diff (new / rank-up / rank-down / complaints-increased per company) that monitoring workflows read first.

industryProfile (one industryProfile row): for a product/category — topIssues[], fastestGrowingIssues[] (new or +50% vs the previous period), earlyWarnings[] (named surge signals: identity-theft-surge, fraud-surge, … + a warningLevel), highestSeverityIssues[] (ranked by complaint-type severity), a real product-wide baseline (exact untimely + relief rates from CFPB aggregation across the whole product — not a fabricated cross-industry figure), and riskTrend.

Cohort mode also surfaces cohortStats.industrySignals[] — signals shared by enough of the run's companies to read as an industry-wide pattern (e.g. an identity-theft wave across the credit bureaus).

Honest scope: every per-company score is a within-result-set signal. The industryProfile baseline is the one true cross-company figure — and it's exact (a complaint-weighted product-wide rate from CFPB's own aggregation), labelled "product-wide", never an invented percentile. Complaint volume tracks company size, so the risk score deliberately leans on response-quality rates (untimely / dispute / relief) and velocity — never a cross-universe "this company is risky overall" or industry-percentile rating, which would need per-complaint data for every company in the industry.

---

Use cases

Enterprise risk & compliance (budget-owning teams)

• Third-party / vendor risk management -- Screen vendors before onboarding (compare mode + a companies list → a pass/review/fail verdict each), then monitor them after onboarding (monitor mode on a schedule with a watchlistName), and auto-escalate any vendor whose complaint handling deteriorates (riskEvent records + rankChange).
• Procurement due diligence -- Compare competing suppliers head-to-head, benchmark their response quality against the cohort (cohort mode), and detect worsening complaint trends before signing.
• M&A screening -- Surface hidden consumer-risk signals before an acquisition: complaint velocity, severity mix, issue concentration, and the Complaint Intelligence Score for the target and its peers.

Analysis, monitoring & research

• Most-complained-about leaderboards (aggregation mode) -- "Top 25 companies by complaint count in 2024", "Top 10 issues for credit cards in CA", "State-by-state complaint distribution for mortgages". One call. No raw-record paging.
• Compliance monitoring -- Track complaints filed against your own institution or competitors on a daily or weekly schedule to catch emerging issues early.
• Due diligence -- Before partnering with, acquiring, or investing in a financial company, review their complaint history for red flags.
• Investigative journalism -- Analyze complaint trends by product type, geography, or time period to uncover systemic consumer harm patterns.
• Legal research -- Find complaints related to specific issues like identity theft, unauthorized charges, predatory lending, or discriminatory practices.
• Market intelligence -- Identify where consumers are most dissatisfied to spot product improvement or market entry opportunities.
• Regulatory risk assessment -- Monitor complaint volumes and company response patterns as leading indicators of potential regulatory action.
• Academic research -- Build datasets for studying consumer finance behavior, complaint resolution effectiveness, or financial inclusion topics.
• Brand reputation tracking -- Track how a company's complaint volume and resolution quality changes over time relative to industry peers.
• Geographic analysis -- Compare complaint patterns across states to identify regional concentrations of specific financial product issues. Aggregation mode with aggregateBy: "state" returns the full distribution in one call.
• Customer experience benchmarking -- Compare timeliness and resolution types across companies in the same product category. Use aggregateBy: "company_response" or aggregateBy: "timely" for cohort-level metrics.

---

API & integration

Python

from apify_client import ApifyClient

client = ApifyClient("YOUR_API_TOKEN")

run = client.actor("OuMDkYU3IROUS0AEj").call(run_input={
    "company": "Bank of America",
    "product": "Mortgage",
    "dateFrom": "2024-01-01",
    "maxResults": 200,
})

dataset = client.dataset(run["defaultDatasetId"])
for item in dataset.iterate_items():
    print(f"{item['dateReceived']} - {item['issue']} ({item['companyResponse']})")

JavaScript

import { ApifyClient } from "apify-client";

const client = new ApifyClient({ token: "YOUR_API_TOKEN" });

const run = await client.actor("OuMDkYU3IROUS0AEj").call({
    searchTerm: "late fee",
    product: "Credit card or prepaid card",
    state: "NY",
    maxResults: 75,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(JSON.stringify(items, null, 2));

cURL

# Start the actor run
curl -X POST "https://api.apify.com/v2/acts/OuMDkYU3IROUS0AEj/runs?token=YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "searchTerm": "debt collection harassment",
    "state": "TX",
    "sortBy": "created_date_desc",
    "maxResults": 50
  }'

# Fetch results from the dataset (use defaultDatasetId from the run response)
curl "https://api.apify.com/v2/datasets/DATASET_ID/items?token=YOUR_API_TOKEN&format=json"

Platform integrations

• Zapier -- Trigger workflows when new complaint data is collected. Push to Google Sheets, Slack, email, or CRMs.
• Make (Integromat) -- Build multi-step automations combining CFPB data with other sources.
• Google Sheets -- Export results directly to a spreadsheet for collaborative analysis.
• Webhooks -- Receive HTTP callbacks when actor runs complete to trigger downstream processing.
• Apify Dataset API -- Access results in JSON, CSV, XML, Excel, HTML, or RSS formats.

---

Use in Dify

Drop this actor into Dify workflows via the Apify plugin's Run Actor node. With enableCompanySummary on, each company comes back scored, classified, and screened as structured JSON — pass / review / fail plus the riskLevel and recommendedAction enums your downstream node branches on. A raw CFPB API call returns Elasticsearch hits you still have to interpret; this returns a vendor-screening decision.

• Actor ID: ryanclinton/cfpb-consumer-complaints
• Sample input (screen a list of banks before onboarding a vendor):

{
    "outputMode": "records",
    "companies": ["Wells Fargo", "Bank of America", "Equifax"],
    "maxResults": 200,
    "enableCompanySummary": true,
    "outputProfile": "standard"
}

The companySummary records carry the routing primitives. A Dify if/else node branches on them without parsing prose:

• screeningResult == "fail" → route to a manual-review / escalation branch (the verdict + riskFactors[] are paste-ready into a ticket or Slack message).
• screeningResult == "review" → flag for a human; surface untimelyRate / disputeRate.
• screeningResult == "pass" → continue the onboarding flow automatically.

You can also branch on recommendedAction (monitor / review / escalate) or gate automation on confidence == "high" so thin-sample verdicts route to a human instead of firing a downstream action. The runSummary record's headline / oneLine and exceptions[] are ready to drop straight into an exec email or alert. Filter the dataset by recordType (companySummary vs record vs runSummary) so each Dify branch reads only the rows it needs.

---

How it works

Records mode

1. The actor reads your input configuration (search term, company, product, state, date range, sort order, max results, autoPaginate flag).
2. It constructs a query URL for the CFPB Consumer Complaint API with pagination parameters (size=100, frm=0, no_aggs=true).
3. The actor sends an HTTP GET request to the CFPB API endpoint.
4. The Elasticsearch-style response (hits.hits) is parsed and each complaint record is transformed from the raw _source format into a clean flat JSON object.
5. Transformed records are pushed to the Apify dataset.
6. If more results are needed, the actor increments the offset and fetches the next page after a 200ms delay.
7. When autoPaginate=true and the total matching count exceeds maxResults, the actor extends the run to fetch every matching record up to the 10,000 single-query cap.
8. Steps 3--6 repeat until the requested count is collected or no more results are available.

Aggregation mode

1. The actor reads your filters and the aggregateBy field selection.
2. It constructs a single query URL with size=0&no_aggs=false&field=all.
3. CFPB returns the bucket counts for every aggregatable field in one response.
4. The actor extracts the buckets for the chosen aggregateBy field, ranks them by count (largest first), computes share = count / totalMatching, and emits one row per bucket up to maxResults.
5. The result is a ready-to-graph leaderboard — no client-side counting required.

CFPB Consumer Complaint Search Pipeline

  Input Config         CFPB API                Transform              Apify Dataset
 +------------+    +-----------------+    +------------------+    +---------------+
 | outputMode |    | records mode:   |    | hits -> flat JSON |    | kind=record   |
 | aggregateBy|--->| size=100&no_aggs|--->| + cfpbUrl         |--->| complaintId   |
 | searchTerm |    | =true&frm=offset|    | + extractedAt     |    | company       |
 | company    |    |                 |    |                   |    | issue,...     |
 | product    |    +-----------------+    +------------------+    +---------------+
 | state      |
 | dateFrom   |    +-----------------+    +------------------+    +---------------+
 | dateTo     |    | aggregation     |    | aggregations[X]  |    | kind=aggreg.  |
 | sortBy     |--->| mode:           |--->| .buckets ->       |--->| rank, term,   |
 | maxResults |    | size=0&no_aggs= |    | rank+share+count  |    | count, share, |
 | autoPagin. |    | false&field=all |    |                   |    | totalMatching |
 +------------+    +-----------------+    +------------------+    +---------------+

---

Performance & cost

Scenario	Mode	Results	Approx. Run Time	Memory
Quick lookup	records	25	3--5 seconds	256 MB
Default run	records	50	5--10 seconds	256 MB
Medium batch	records	200	15--20 seconds	256 MB
Large batch	records	1,000	30--60 seconds	256 MB
Full single-query	records	10,000	4--6 minutes	256 MB
Top-25 leaderboard	aggregation	25 buckets	2--4 seconds	256 MB
Full distribution	aggregation	up to 10,000 buckets	2--4 seconds	256 MB

Pricing: $0.002 per row emitted (records or aggregation buckets), pay-per-event. A 25-bucket leaderboard run costs ~$0.05; a 10,000-record full-corpus pull costs ~$20. There are no external API costs — the CFPB database is a free US government resource.

---

Limitations

• 10,000 records per run (records mode) -- The CFPB API enforces an Elasticsearch single-query limit of 10,000 hits. For full-corpus analysis beyond that, switch to aggregation mode (returns counts over the entire matching set, not capped at 10,000) or split records-mode queries by date range.
• Aggregation mode is per-query -- Each aggregation call returns the buckets for one chosen field across all matching complaints. To group by multiple fields (e.g., company × product), run aggregation once per dimension and join the results downstream.
• No ZIP code filtering -- The CFPB API does not support filtering by ZIP code directly. Filter by state, then post-process using the zipCode field in the output.
• Narrative availability -- The narrative field is only populated when the consumer opted in to public disclosure. Many complaints will have null for this field.
• 15-day publication delay -- Complaints typically appear in the CFPB database approximately 15 days after being sent to the company, to allow time for the company to respond.
• Company name matching -- Company names must match the CFPB's canonical name format (e.g., "EQUIFAX, INC." not just "Equifax"). Partial matches may not return expected results.
• No real-time data -- The database is updated daily by the CFPB, not in real time. There may be a short lag between updates.
• US complaints only -- The CFPB database covers only complaints filed by US consumers against financial companies operating in the United States.

---

Responsible use

• Public data only -- This actor accesses a publicly available US government database. All complaint records are published by the CFPB with consumer consent considerations already applied.
• Respectful access -- The actor includes a 200ms delay between API pages and uses no_aggs=true to minimize server load on the government endpoint.
• No personal identification -- While complaints include state and partial ZIP codes, they do not contain personally identifiable information about the consumers who filed them.
• Fair interpretation -- A high volume of complaints does not necessarily indicate wrongdoing. Larger companies naturally receive more complaints. Always normalize by company size or customer base when drawing conclusions.
• Comply with terms -- Users should review the CFPB data terms and ensure their use aligns with applicable regulations and ethical standards.

---

FAQ

Do I need a CFPB API key to use this actor? No. The CFPB consumer complaint database is a free, public US government resource. This actor accesses it without any authentication or API key.

How current is the complaint data? The CFPB updates its database daily. Complaints typically appear approximately 15 days after being sent to the company, giving the company time to respond before publication.

Can I get more than 10,000 results per run? The CFPB API enforces a 10,000-record cap on any single query (an Elasticsearch default). To go beyond, you have two options: (1) split records-mode runs by date range and combine the datasets, or (2) use aggregation mode, which returns count buckets computed over the entire matching set without the 10,000-record cap.

What's the difference between records mode and aggregation mode? Records mode returns one row per complaint (the full detail — company, issue, narrative, etc.). Aggregation mode returns top-N count buckets — e.g. "TransUnion: 790,668 complaints, 28.9% of all 2024 complaints." Use records when you need individual complaint details; use aggregation when you need leaderboards, distributions, or share-of-complaints metrics.

What does the autoPaginate flag do? By default the actor stops at maxResults. With autoPaginate=true, when the total matching count exceeds maxResults, the actor extends the run to fetch every matching complaint up to the 10,000 cap. Useful when you want "all matches for this query" without knowing the total in advance.

What does the narrative field contain? When consumers submit complaints, they can include a written description of what happened. If the consumer consents to public disclosure, this text appears in the narrative field. Otherwise it will be null.

Can I search for complaints from a specific ZIP code? The CFPB API does not support direct ZIP code filtering. Filter by state using the state parameter, then filter the output by zipCode in your post-processing.

How far back does the data go? The CFPB complaint database includes records from 2011 to the present, covering over 5 million complaints across all financial product categories.

What product categories are available? The actor supports 10 product filters: Credit reporting/repair, Debt collection, Mortgage, Credit card/prepaid card, Checking/savings account, Student loan, Vehicle loan/lease, Money transfer/virtual currency, Payday/title/personal loan, and Personal loan.

How are company names matched? The company parameter filters by the CFPB's canonical company name. For best results, use the official name as it appears in CFPB records (e.g., "EQUIFAX, INC." rather than "Equifax"). You can also try the searchTerm field for more flexible matching.

Can I schedule this actor to run automatically? Yes. Use Apify's built-in scheduler to run the actor on any cadence -- daily, weekly, or custom cron expressions. Combine with webhooks or integrations to get notified when new data is available.

What does the timely field mean? The timely field indicates whether the company responded to the complaint within the CFPB's expected timeframe (typically 15 days). A value of "Yes" means the response was on time.

Can I export results as CSV or Excel? Yes. Apify datasets support export in JSON, CSV, Excel (XLSX), XML, HTML, and RSS formats. Use the Dataset tab in the console or the Dataset API endpoint with a format parameter.

Is this actor suitable for large-scale research? For datasets up to 10,000 records, a single records-mode run is sufficient. For full-corpus research questions ("what's the complaint distribution across the 2.7 million 2024 complaints?"), use aggregation mode — one call returns counts across the entire matching set. For raw-record needs above 10,000, segment by date range and merge.

---

Related actors

Actor	Description
FDIC Bank Data Search	Look up detailed information about FDIC-insured banks and financial institutions. Correlate complaint volumes with bank size, assets, or regulatory status.
SEC EDGAR Filing Analyzer	Analyze SEC filings for publicly traded financial companies. Cross-reference CFPB complaint trends with financial disclosures and risk factors.
OFAC Sanctions Search	Check whether companies or individuals appearing in complaints are on the OFAC sanctions list for enhanced due diligence.
SEC Insider Trading	Track insider trading activity at financial companies. Compare insider selling patterns with spikes in consumer complaints.
EDGAR Filing Search	Search SEC EDGAR filings by company, form type, or date range. Build a complete regulatory picture alongside CFPB complaint data.
Congressional Stock Trade Tracker	Monitor stock trades by members of Congress. Investigate whether legislative activity correlates with changes in complaint patterns for regulated industries.