–40%
CPA after rollout
Project overview
Control center for Yandex.Direct and Metrica
Yandex Audit Platform is an enterprise-grade SaaS for marketers, agencies, and in-house performance teams. It blends traditional advertising controls with generative AI—from instant campaign audits to complex “generate an A/B test, launch it, and disable the losers in a week” scenarios.
The AI agent is the nucleus: it understands screen context, campaign IDs, account structure, API status, and triggers the right functions. For users it feels like an “intelligent teammate” living inside the UI.
The platform merges data from Direct, Metrica, Webmaster, Business, and Audiences to deliver a panoramic ROI/ROMI view. Celery workers and a custom engine handle 50K+ search-query rows, build CPA/CPC distributions, surface anomalies, and produce a plan of action.
The outcome: a shift from manual Excel audits to an automated optimization loop that becomes the central decision-making hub for marketing.
Capabilities
Interfaces that work while you sleep
01
Unified dashboard
Accounts, campaigns, status, budgets, balance, and API quotas—visualized in one glass panel. This is the command center for audits and optimization playbooks.
DashboardRealtime
Summary screen
02
AI assistant with Tool Calling
Floating agent that understands context (campaign view / A/B / semantic analysis) and executes complex commands via natural language.
AI AgentNLU
Tool Calling
03
Deep query audits
Excel exports are ingested, distributions are calculated, wasteful n-grams detected, visualizations built, and negative keywords suggested with projected savings.
N-gramAnomaly Detection
Financial control
04
Campaign ops + A/B
Flexible control over status/budget, creative auto-generation, and a “loser killer” that automatically pauses underperforming ads.
AB testsAutopilot
A/B center
05
Analytics & time targeting
Demand heatmaps, end-to-end analytics for Metrica + Direct, segmentation by device and geo, plus smart bid-adjustment guidance.
HeatmapROI
Time targeting
06
Semantics, audiences, reputation
Shared Sets with negative keywords, RFM segments from Yandex Audiences, review monitoring, and AI replies in Yandex Business—all integrated into a single data bus.
RFMShared sets
RFM + reputation
Inside the platform
Modules that consolidate entire ad universes
The platform spans the full cycle—from raw log ingestion to AI-agent actions. Each module is its own service, yet all are synced through one data bus.
- Neural auditors that score traffic quality and search phrases.
- Task orchestrator: Celery workers digest backups and Excel exports.
- Realtime KPI showcase with SLA and API-quota alerts.
- Tool Calling agent that launches scenarios without human involvement.
Workflow scenarios
How the agent guides the marketer
Diagnostics
Upload the search-query log → compute CPA/CTR stats, detect outliers, and plot traffic maps.
Waste detection
The n-gram engine flags junk patterns, recommends negative keywords, and forecasts savings.
Optimization & tests
The AI assistant generates creatives, launches A/B experiments, and monitors the “loser killer.”
KPI control
End-to-end ROI/ROMI analytics, time heatmaps, PDF/HTML reporting, and an action log for compliance.
Value
What the client gets
Budget optimization. AI discovers leakages and builds negative keyword sets, saving 15–40% of spend without losing leads.
Execution speed. Routine tasks like “audit the campaign / find problematic keywords” take 2–5 minutes instead of an hour.
Enhanced A/B testing. Creative generation plus the auto-killer deliver +20–35% CTR.
End-to-end analytics. Funnel from impression to revenue with segmentation by device, geo, RFM, and true ROMI.
Reputation under control. Automated replies to Yandex Business reviews and unified rating monitoring.
Junior-friendly. Natural language makes advanced workflows accessible to junior PPC specialists.
Technology
Stack & integrations
Backend: Python 3.11, FastAPI, SQLAlchemy, Pydantic, Celery + Redis, Pandas, NumPy, SlowAPI, Alembic.
Frontend: HTML5, CSS3, Vanilla JS (ES6+), Chart.js, NProgress, WebSockets.
AI & ML: OpenAI GPT-4o / GPT-4o-mini, Responses API, LangChain patterns, custom n-gram engine, Z-score & IQR analysis.
Yandex integrations: Direct API, Metrica API, Audiences, Webmaster, Business, OAuth, JSON-RPC + REST.
Infrastructure: Docker Compose, Nginx, PostgreSQL/SQLite, Redis, python-dotenv, pre-commit.
Security: JWT, rate limiting, retry with backoff, input validation, structured logging.
Architecture
Structure of the Yandex Audit Platform project
The auditYandex tree shows where agents, APIs, analytics, and infrastructure live—saving onboarding hours and helping the team locate modules quickly.
Results
Implementation metrics
Economics: CPA –25…–40% through smarter negative keywords and budget rebalancing; 15–30% spend savings.
Operations: Campaign audits take 5–10 minutes instead of 4–6 hours; 50K query rows processed with no performance drop.
Strategy: Shift to proactive management—the system flags risks before spend is wasted and scales across dozens of accounts.
Team: Fewer human errors, faster learning thanks to AI tips and knowledge bases, transparent action logs.
Want your own AI marketing center?
We’ll build a platform that runs your advertising, data, and processes—as polished and bulletproof as this case.