Radio Ad Pipeline
Demo Data
Internal Executive Brief · Not for public distribution

Radio Ad Intelligence Pipeline

Automated monitoring system for discovering loan-related radio advertisers, keyword opportunities, and direct-response ad signals from U.S. talk/news radio stations.

View Live Harvest DashboardReview Station Coverage →
System Status
Ready / Monitoring
Stations Reachable
9 / 9
Primary Vertical
Loan & Financial Services
Current Mode
Overnight Harvest
Dashboard Route
/radio-harvest
01

Executive Summary

This system automatically monitors selected U.S. radio stations and captures advertising signals related to loans and financial services. Instead of manually listening to hours of radio, the pipeline records chunks, transcribes audio, detects keyword matches, identifies likely advertisers, and sends detections into a review queue.

The result is a repeatable intelligence system that helps us discover:

  • New advertiser names
  • Brand keywords
  • Loan-related phrases
  • Compliance wording patterns
  • Markets with active financial advertisers
  • Timing patterns for when loan ads air
  • Opportunities for Google Ads / Bing Ads expansion
Business Value

Convert offline radio advertising signals into online search advertising opportunities.

02

The Problem

Loan advertisers often test offers on radio before, during, or alongside search campaigns. These ads contain brand names, phone numbers, URLs, offer language, disclaimers, and intent-heavy phrases that are valuable for keyword research.

Manual research is slow because:

  • Radio ads are scattered across many stations and markets
  • Ads may air overnight or in narrow time windows
  • Advertiser names may be mentioned only briefly
  • Offers change frequently
  • Humans cannot monitor many stations continuously
  • Useful phrases are buried inside long programming blocks
03

What The System Does

1
Select target stations

Operator picks U.S. stations by market, format, and likelihood of direct-response financial ads.

2
Probe streams

System verifies each station stream is reachable before a harvest starts.

3
Record audio

Captures short audio chunks from live streams on a rolling basis.

4
Transcribe

Audio chunks are converted to text so ads become searchable.

5
Detect keywords

Looks for loan and financial-service phrases across transcripts.

6
Score & store

Each detection saved with station, market, timestamp, transcript, match, and confidence.

7
Surface in dashboard

Detections, queue, and station health visible to the operator.

8
Review opportunities

Human review validates advertiser and keyword candidates.

9
Export intelligence

Approved keywords flow to paid search and affiliate research.

04

Current Build Status

Radio Harvest DashboardCompletedInternal route at /radio-harvest
Station ProbeCompleted9/9 stations reachable in latest test
Overnight Harvest ProfileCompletedCan start overnight keyword harvest
Stop Harvest ControlCompletedOperator can stop active harvest
Detection ViewCompletedDetections visible from dashboard
Queue ViewCompletedQueue status visible
Station ViewCompletedStation availability visible
Automated TestsCompletedHarvest dashboard covered by tests
Loan Keyword DetectionActiveDetects loan & financial-service phrases
Advertiser Review WorkflowIn ProgressDetections need human review before campaign use
Export / Manager ReportingNeededThis page supports manager visibility
05

Latest Station Probe Result

9 of 9 stations reachableLast probe: most recent run
StationMarketFormatStatus
KLIF 570 AMDallas, TXNews/TalkReachable
WBAP 820 AMDallas/Fort Worth, TXNews/TalkReachable
KTRH 740 AMHouston, TXNews/TalkReachable
WOAI 1200 AMSan Antonio, TXNews/TalkReachable
WHBO 1040Tampa, FLNews/TalkReachable
WSB 750 AMAtlanta, GANews/TalkReachable
WTAM 1100 AMCleveland, OHNews/TalkReachable
WIBC 93.1 FMIndianapolis, INNews/TalkReachable
WWTN 99.7 FMNashville, TNNews/TalkReachable

These stations are high-priority because talk/news radio tends to carry direct-response financial ads — loan, debt, tax, insurance, and cash-related offers.

06

Operator Dashboard Features

Probe Stations

Checks whether configured station streams are reachable.

$ harvest_control.py probe --limit 20
Start Overnight Harvest

Starts the overnight keyword monitoring profile.

$ harvest_control.py start --profile overnight_keyword_harvest
Stop Harvest

Stops the currently running harvest process.

$ harvest_control.py stop
Status View

Shows whether the system is running, idle, stopped, or errored.

Detections View

Captured keyword detections with station, timestamp, confidence, transcript snippet.

Queue View

Pending, processed, dropped, or failed audio chunks.

Stations View

Configured stations and stream reachability.

07

Example Detection Signals

Illustrative examples — not yet connected to live backend data.

Personal Loan Signal
Medium
Station
KLIF 570 AM
Market
Dallas, TX
Category
Personal Loan
Phrase
“get the cash you need”
Action: Review transcript and identify advertiser
Tax Relief Signal
High
Station
WSB 750 AM
Market
Atlanta, GA
Category
Tax Relief
Phrase
“back taxes / unfiled tax returns”
Action: Add to tax relief keyword review queue
Debt Relief Signal
Medium
Station
WTAM 1100 AM
Market
Cleveland, OH
Category
Debt Relief
Phrase
“stop collections”
Action: Review for advertiser name and landing page
08

Confirmed Intelligence Example

Advertiser
Billshappen.com

The system found loan-related ad detections connected to this advertiser across multiple target stations. Repeated cross-market coverage is exactly the signal we want — it suggests active media buying and direct-response testing.

All detections require human review before being used for campaign decisions.
StationMarketConfidence
KLIF 570 AMDallas, TX0.92
WSB 750 AMAtlanta, GA0.92
KTRH 740 AMHouston, TX0.95
WOAI 1200 AMSan Antonio, TX0.95
09

From Radio Detection To Keyword Opportunity

  1. 1
    Radio Ad Mention
  2. 2
    Transcript Match
  3. 3
    Advertiser / Brand Extraction
  4. 4
    Human Review
  5. 5
    Keyword Candidate
  6. 6
    Landing Page / Campaign Research
  7. 7
    Google Ads / Bing Ads Testing

Example keyword opportunities

billshappenbills happen loanbills happen cashback taxes helpunfiled tax returns helpstop collections helptimeshare exit companyemergency cash loanget money for billsloan help today

Keyword types we capture

  • Brand keywords
  • Competitor keywords
  • Offer phrase keywords
  • Problem-aware keywords
  • Radio slogan keywords
  • URL / domain keywords
  • Phone-number advertiser lookup
  • Local-market financial keywords
Keyword usage must follow platform policy and internal risk rules. Trademark keywords are reviewed separately before campaign use.
10

Why This Matters For The Business

Finds advertisers before they become obvious in search tools
Discovers real-world ad language from live radio
Identifies markets where loan advertisers are active
Helps build new keyword lists faster
Supports competitor intelligence
Reduces manual listening work
Creates a repeatable research process
Connects offline ad spend to online search opportunity
Supports landing page and ad copy research
Helps prioritize which verticals to test next

Instead of guessing which loan keywords or advertisers to research, we use live radio ads as a signal source.

11

Data Collected By The System

Data TypePurpose
Station nameKnow where the ad appeared
Market / stateUnderstand geographic opportunity
TimestampIdentify airing time patterns
Audio chunkPreserve source evidence
TranscriptSearchable ad text
Keyword matchesDetect relevant vertical
Confidence scorePrioritize review
Advertiser candidateIdentify possible brand / domain
Review statusSeparate raw detections from approved intelligence
Queue statusMonitor pipeline health

The system separates raw machine detections from reviewed business intelligence.

12

Human Review Workflow

The system does not automatically assume every detection is usable. A human reviewer verifies important detections before they inform campaign planning.

  1. 1Open detection
  2. 2Read transcript snippet
  3. 3Listen to audio chunk if needed
  4. 4Identify advertiser name / domain / phone number
  5. 5Classify vertical
  6. 6Mark confidence
  7. 7Approve, reject, or keep for later
  8. 8Export approved keyword candidates
Review Statuses
NewNeeds ReviewApprovedRejectedWatchlistExported
13

Risk & Compliance Notes

This system is for research and intelligence. It does not automatically create ads, claim affiliation with advertisers, or approve trademark keyword usage.

  • Radio detection does not mean a keyword is automatically safe to advertise on
  • Trademark / brand terms should be reviewed before use
  • Ad copy must not falsely imply partnership
  • Landing pages must avoid misleading representation
  • Loan-related offers require compliance review
  • Raw transcripts may contain transcription errors
  • Human review is required before campaign execution

The system discovers opportunities. It does not replace compliance review.

14

Current Limitations

  • Transcription can be imperfect
  • Station streams may occasionally fail
  • Some ads may be missed if audio quality is poor
  • Advertiser names may require manual confirmation
  • Phone numbers and URLs may need special extraction logic
  • Scaling beyond 20 stations requires CPU/GPU, storage, network, and queue monitoring
  • Not every detected phrase is commercially useful
  • Dashboard still needs stronger manager-level reporting and export views
15

Scaling Plan

Phase 1
Stabilize 9 Stations
  • •Confirm streams are stable
  • •Confirm detection quality
  • •Review first batch of detections
  • •Identify strongest verticals
  • •Build export workflow
Phase 2
Expand To 20 Stations
  • •Add more target markets
  • •Monitor queue pressure
  • •Track dropped chunks
  • •Tune chunk size and transcription throughput
  • •Add station priority levels
Phase 3
Improve Intelligence Layer
  • •Extract advertiser names
  • •Extract URLs
  • •Extract phone numbers
  • •Cluster repeated advertisers
  • •Score keyword opportunities
  • •Add manager reports
Phase 4
Campaign Research Integration
  • •Export approved keywords
  • •Group by vertical
  • •Group by market
  • •Connect with Google Ads / Bing Ads testing workflow
  • •Track which radio-sourced keywords perform
16

Manager KPIs

Stations Monitored
—
Number of radio streams being tracked
Stream Reachability
—
Whether stations are accessible
Chunks Processed
—
Amount of audio successfully analyzed
Detection Count
—
Number of keyword / ad signals found
High Confidence Hits
—
Detections likely worth review
Advertisers Identified
—
Possible brands / domains discovered
Approved Keywords
—
Keywords ready for campaign research
Exported Opportunities
—
Reviewed intelligence sent to marketing workflow

KPI values populate once the dashboard is connected to live backend endpoints.

17

Recommended Next Steps

  1. 1Run overnight harvest across the current 9 reachable stations
  2. 2Review detections the next morning
  3. 3Identify repeated advertisers and domains
  4. 4Add approved advertisers to keyword candidate list
  5. 5Expand station list toward 20 stations
  6. 6Build export view for Manager / Marketing
  7. 7Add advertiser clustering and confidence scoring
  8. 8Connect approved keywords to campaign planning workflow

The immediate goal is to turn raw radio detections into reviewed keyword opportunities that can support paid search testing.