Spying Agent

Key results

• 2 VA/mo — Equivalent manual roles removed from the monthly workload
• ~2 hr/day — Ideation time saved per media buyer and scriptwriter
• ~200/day — Videos in the winning-creative feed powering ideation
• $3M→$11M — Annual revenue range after scaling on this workflow

TL;DR

• Spying Agent is a live ads-intelligence SaaS at spyingagent.com: marketing site, keyword Wall for discovering winning creatives, and magic-link account access.
• The Wall surfaces competitor ads by keyword with status, days running, and detail modals (video, stats, transcript).
• Scripter turns those winner patterns into brand-specific scripts; Slack delivery puts curated packs in the buyer’s workflow.
• Screenshots on the case study page are from the production product—not mockups.

Project Overview

Spying Agent turns winning-creative intelligence into a daily Slack and Wall workflow—plus Scripter for brand-ready scripts.

The live product at spyingagent.com sells the loop clearly: find winning ads, analyze winners, prepare data, deliver to Slack—then use Scripter for copy and script variants.

The Keyword Wall is the in-product discovery surface: add keywords, browse active creatives, open ad detail with transcript and source metadata. Sign-in is magic-link for the account dashboard and bots.

• Marketing narrative: Slack delivery, time savings, and agency-ready pricing
• Keyword Wall: discover grid with days-running and ad volume signals
• Ad detail: video, landing page, stats, and transcript tabs
• Scripter: LLM script generation from winner patterns
• Magic-link auth into the account / bots dashboard

Business Challenge / Problem Statement

Manual competitor monitoring was collapsing creative energy.

Scriptwriters, editors, and media buyers were spending at least five hours daily on content scouting.

Teams reviewed mostly average-quality videos because they never had the capacity to evaluate everything.

Different people used different methods, so results were inconsistent and hard to scale.

Important creative directions were discovered too late, creating a real opportunity cost in spend optimization.

The process consumed cognitive bandwidth that should have been spent on strategy and execution.

• Massive Time Sink — 3 to 5 people repeatedly reviewed 30-40 videos each day, consuming 5+ hours in total.
• Low Signal Ratio — Most reviewed content was low-performing noise; winning ideas were buried.
• No Reliable Ranking — The team had no repeatable scorecard for what was actually winning.
• Geographic Blind Spots — Market-specific ad trends were missed without targeted location filters.
• Scaling Barrier — As budget and campaigns grew, manual spying stopped being sustainable.
• Execution Delays — Late insight handoffs reduced the ability to move fast on winning ad themes.

Objectives & Goals

What the automation had to solve from day one.

Reduce manual scouting time by automating competitor ad discovery and ranking.

Standardize creative intelligence so every person receives the same signal framework.

Identify winning videos from objective metrics including volume and trend indicators.

Deliver categorized daily bundles for immediate creative production decisions.

Support scalable operation as market coverage increases across keywords and geographies.

• Automated Source Collection — Continuously ingest competitor content from multiple sources, including Facebook Ads Library and Spy Hero.
• Winning-Video Scoring — Score videos using upload activity, impressions trend, and momentum signals to reduce false positives.
• Signal Filtering — Keep the pipeline output lean and action-focused by dropping low-confidence candidates.
• Topic + Market Routing — Group outputs by keyword and geography to match actual buying decisions.
• Team-Ready Distribution — Push hand-picked, review-ready sets to Slack channels that media buyers already use daily.
• Revenue Expansion Through Scale — Enable adding more people, campaigns, and markets without increasing manual scouting workload.

Key Features & Innovation

A practical intelligence engine for scaling paid growth teams.

Multi-Source Scraping Engine

A custom Node.js data layer scrapes multiple competitor-ad ecosystems to remove dependency on single-platform windows or manual checks.

Geo + Keyword Controls

• Keyword targeting: Teams define exact search terms and markets for focused competitive monitoring.
• Geographic filtering: Each bundle is segmented so teams can compare regional trends quickly.
• Category grouping: Videos are grouped by campaign context for easier script and creative planning.

Winning-Videos Scoring Model

• Upload volume: Measures how many creative versions or updates indicate market momentum.
• Impression trend: Tracks growth velocity and momentum rather than static snapshot counts.
• Trend correlation: Filters duplicate or stale videos and keeps the stream aligned with current winner patterns.

Delivery Intelligence

The output is formatted for team speed, not for dashboards: daily Slack drops with the most relevant winning videos and enough context to act immediately.

Category-Driven Slack Routing

The daily output is not random. Teams receive a fixed number of videos per keyword so review bandwidth is always protected.

Daily Scaling Loop

• Steady volume: About 300 curated videos sent every day.
• Balanced delivery: Each keyword can receive 5, 10, or 20 videos depending on market demand.
• Brainpower preserved: People focus on ideation, scriptwriting, and iteration instead of hunting.

Scale-Ready Architecture

• Modular adapters: Each data source is isolated so new platforms can be added without breaking the core ranking logic.
• Normalization layer: Inconsistent source payloads are standardized before scoring and distribution.
• Team extensibility: More buyers or markets can be added without retraining the whole process.

Operational Reliability

Scheduler controls and parsing checks keep the flow running day-to-day with resilience against source fluctuation.

Results & Impact

Operationally, we eliminated about two VA roles’ worth of work every month. Media buyers and scriptwriters each reclaimed roughly two hours per day on ideation—time that had been burned manually chasing references instead of producing.

Ideation intensity went through the roof in a good way: the pipeline maintained a steady stream of about 200 videos per day, always surfacing the latest content that was winning in the industry, so the team could riff on real market signal—not stale guesses.

That combination let the business scale from about $3 million to about $11 million in annual revenue while running on the same intelligence loop. The live product at spyingagent.com is the customer-facing expression of that system.

What our clients say

We stopped spending half our day digging through losing content. The automated spy feed gave our media buyers and creators a daily, ranked shortlist of winning videos, and that changed our speed across every campaign.

Campaign Intelligence Lead — Marketing Team

Technologies & Tools Used

LLM-first advertising script system

A data pipeline for sourcing winner ads, extracting competitive patterns, and prompting a Scripter LLM for high-volume script generation—shipped as the SpyingAgent SaaS at spyingagent.com.

The model is tuned to convert competitor intelligence into scripts that align with brand voice, campaign goals, and audience context.

Technology stack

• Node.js — Custom orchestration and transformation layer that handles scraping, scoring, and dispatch logic.
• Facebook Ads Library Scraping — Automated retrieval of competitor ad data with consistent refresh and parsing controls.
• Spy Hero — Secondary intelligence source for deeper competitor creative trend detection.
• Upload Signal Intelligence — Additional source signals used to detect trending content velocity.
• Scoring Framework — Custom model combining upload frequency, impression trend, and freshness signals to identify winners.
• Slack API — Automated delivery channel for team-ready daily bundles by keyword and geography.