skōp - Discover top Creators to promote your app

skōp connects app developers with top social media Creators who drive real engagement. Find, vet, and collaborate with influencers who match your brand—boost downloads and visibility effortlessly.

Comment

Inspiration

Our team started with a simple but frustrating problem: how do indie developers promote their apps without burning through budgets? Most marketing advice felt anecdotal, influencer discovery was manual, and data about app markets was scattered across dozens of sources. We wanted to democratize intelligent insights: • Validate whether an app idea has market potential. • Recommend the right creators to promote it—based on data, not guesswork. That was the spark for our project: a two-step intelligence platform for app promotion.

What it does

Analyzes markets and recommends the best creators to promote an app—all in one click.

How we built it

Core Services Used

     **Cloud Run** – serverless hosting for backend
     **AI Studio (Gemini)** – LLM-powered content analysis
     **Agent Development Kit (ADK)** – AI agent management
     **Google Custom Search API** – discover creators
     **Cloud Storage & Firestore** – persistent data
     **Cloud Build** – CI/CD automation
     **Cloud Logging & Monitoring** – observability

We combined several moving parts into one workflow: Data Ingestion: Scraped/queried app store data (titles, installs, ratings) and pulled social creator profiles from TikTok, YouTube, and Instagram APIs. Vector Search: Converted app briefs and creator content into embeddings, enabling semantic similarity matching. Market Analysis Engine: Estimated market opportunity using app trends, CPI benchmarks, and sentiment analysis. Scoring Models: Market Attractiveness Score (MAS): MAS=w1cdotSize+w2cdotGrowth+w3(1−Competition)+w4frac1CPI+w5cdotSentimentMAS = w_1 \cdot Size + w_2 \cdot Growth + w_3 (1 - Competition) + w_4 \frac{1}{CPI} + w_5 \cdot SentimentMAS=w1​cdotSize+w2​cdotGrowth+w3​(1−Competition)+w4​frac1CPI+w5​cdotSentiment Creator Fit Score (CFS): CFS=alphaR+betaA+gammaE−epsilonFCFS = \alpha R + \beta A + \gamma E - \epsilon FCFS=alphaR+betaA+gammaE−epsilonF where RRR = Relevance, AAA = Audience Match, EEE = Engagement Quality, FFF = Fraud Risk. Frontend Dashboard: A clean React UI to show results—market summary plus Top-5 creators with rationale.

Challenges we ran into

Data Quality: Social APIs often return noisy or incomplete data. We had to normalize creators across platforms. Latency vs Depth: Running full analyses in real time sometimes took too long. We cached aggressively and built fallbacks. Fraud Detection: Spotting fake engagement is harder than we thought. Our anomaly detection model still has false positives. Team Bandwidth: Integrating ML, backend, and frontend in <48 hours meant tradeoffs—some features (like diversification optimizer) only run in batch mode.

Accomplishments that we're proud of

Built an end-to-end pipeline: input app concept → output Top-5 creators with rationale. Created a working demo dashboard that looks investor-ready. Learned how to blend market data, ML, and UX design into one coherent tool.

What we learned

Vector embeddings are game-changing for connecting seemingly different worlds (apps ↔ creators). Market sizing can be approximated surprisingly well by combining open datasets with proxy indicators like search trends. The value of explainability: users trust AI more when they see why a recommendation was made. Building on open APIs is fast, but hitting rate limits mid-hackathon is painful

What's next for skōp

Add more networks (Twitch, podcasts, Discord). Smarter fraud detection using graph analysis. Budget-aware optimization: simulate installs across multiple creators under spend constraints. First-party attribution integrations to measure real-world ROI.

Built With

  • adk
  • ai-studio
  • cloud-build
  • cloud-logging
  • cloud-run
  • firestore
Share this project:

Updates