Sprint Engine / beseen.studio

```json { "plan_name": "AI Creative Video Analysis for SubSun", "estimated_duration_minutes": 25, "steps": [ { "id": "step_1", "type": "data_fetch", "name": "Fetch video creative ads from competitors", "depends_on": [], "params": { "table": "competitor_ads", "columns": ["id", "ad_archive_id", "competitor_page_id", "page_name", "primary_text", "headline", "description", "creative_type", "ad_delivery_start_time", "ad_delivery_stop_time", "first_seen_date", "last_seen_date", "is_currently_active", "eu_total_reach", "preview_url", "cta_type"], "filters": {"creative_type": "video"}, "limit": 5000 } }, { "id": "step_2", "type": "data_fetch", "name": "Fetch competitor page details", "depends_on": [], "params": { "table": "competitor_pages", "columns": ["page_id", "page_name", "category", "country_code"], "filters": {}, "limit": 100 } }, { "id": "step_3", "type": "data_fetch", "name": "Fetch video ad performance data", "depends_on": [], "params": { "table": "competitor_ad_daily", "columns": ["ad_archive_id", "check_date", "is_active", "eu_total_reach"], "filters": {}, "limit": 50000 } }, { "id": "step_4", "type": "python_compute", "name": "Filter and merge video ad dataset", "depends_on": ["step_1", "step_2", "step_3"], "params": { "description": "Merge video ads with page info and performance data, filter for fashion/sunglasses brands", "code": "import pandas as pd\nimport numpy as np\nfrom datetime import datetime, timedelta\n\n# Merge video ads with competitor pages\nvideo_ads_merged = competitor_ads.merge(\n competitor_pages, \n left_on='competitor_page_id', \n right_on='page_id', \n how='left',\n suffixes=('', '_page')\n)\n\n# Filter for fashion/accessory brands (assuming sunglasses brands are in this category)\nfashion_keywords = ['fashion', 'accessories', 'eyewear', 'sunglasses', 'lifestyle', 'apparel']\nvideo_ads_fashion = video_ads_merged[\n video_ads_merged['category'].str.lower().str.contains('|'.join(fashion_keywords), na=False) |\n video_ads_merged['page_name'].str.lower().str.contains('sunglasses|eyewear|glasses', na=False)\n]\n\n# Add performance data\nvideo_performance = competitor_ad_daily.groupby('ad_archive_id').agg({\n 'eu_total_reach': 'max',\n 'check_date': ['min', 'max'],\n 'is_active': 'sum'\n}).reset_index()\nvideo_performance.columns = ['ad_archive_id', 'max_reach', 'first_active_date', 'last_active_date', 'days_active']\n\n# Final merged dataset\nvideo_dataset = video_ads_fashion.merge(video_performance, on='ad_archive_id', how='left')\n\nprint(f\"Total video ads found: {len(competitor_ads)}\")\nprint(f\"Fashion/sunglasses video ads: {len(video_ads_fashion)}\")\nprint(f\"Video ads with performance data: {len(video_dataset)}\")\nprint(f\"Unique brands: {video_dataset['page_name'].nunique()}\")\nprint(\"\\nTop brands by video ad count:\")\nprint(video_dataset['page_name'].value_counts().head(10))" } }, { "id": "step_5", "type": "python_compute", "name": "Identify high-performing fun/creative video patterns", "depends_on": ["step_4"], "params": { "description": "Analyze video ad copy and headlines to identify fun, creative, AI-like content patterns", "code": "import re\nfrom collections import Counter\n\n# Define fun/creative keywords that might indicate AI or creative content\nfun_keywords = [\n 'ai', 'artificial intelligence', 'generated', 'create', 'creative', 'fun', 'cool', 'awesome', \n 'amazing', 'transform', 'magic', 'instant', 'viral', 'trending', 'filter', 'effect',\n 'style', 'vibe', 'mood', 'aesthetic', 'lens', 'perspective', 'vision', 'look'\n]\n\nemoji_pattern = re.compile(r'[\\U0001F600-\\U0001F64F\\U0001F300-\\U0001F5FF\\U0001F680-\\U0001F6FF\\U0001F1E0-\\U0001F1FF]+')\n\n# Score ads based on fun/creative content\nvideo_dataset['fun_score'] = 0\nvideo_dataset['has_emojis'] = False\nvideo_dataset['creative_indicators'] = ''\n\nfor idx, row in video_dataset.iterrows():\n text_content = str(row['primary_text']).lower() + ' ' + str(row['headline']).lower() + ' ' + str(row['description']).lower()\n \n # Count fun keywords\n fun_count = sum(1 for keyword in fun_keywords if keyword in text_content)\n video_dataset.at[idx, 'fun_score'] = fun_count\n \n # Check for emojis\n has_emojis = bool(emoji_pattern.search(str(row['primary_text']) + str(row['headline']) + str(row['description'])))\n video_dataset.at[idx, 'has_emojis'] = has_emojis\n \n # Identify creative indicators\n found_keywords = [kw for kw in fun_keywords if kw in text_content]\n video_dataset.at[idx, 'creative_indicators'] = ', '.join(found_keywords)\n\n# Filter for high-performing creative ads\nhigh_performance_creative = video_dataset[\n (video_dataset['fun_score'] > 0) | \n (video_dataset['has_emojis'] == True)\n].copy()\n\n# Sort by reach and fun score\nhigh_performance_creative['combined_score'] = (\n high_performance_creative['max_reach'].fillna(0) * 0.6 + \n high_performance_creative['fun_score'] * 1000 + \n high_performance_creative['has_emojis'].astype(int) * 500\n)\n\ntop_creative_ads = high_performance_creative.sort_values('combined_score', ascending=False).head(20)\n\nprint(f\"Creative/fun video ads found: {len(high_performance_creative)}\")\nprint(f\"\\nTop creative indicators:\")\nall_indicators = []\nfor indicators in high_performance_creative['creative_indicators']:\n if indicators:\n all_indicators.extend(indicators.split(', '))\nfor indicator, count in Counter(all_indicators).most_common(10):\n print(f\"- {indicator}: {count} ads\")\n\nprint(f\"\\nAds with emojis: {high_performance_creative['has_emojis'].sum()}\")\nprint(f\"Average fun score: {high_performance_creative['fun_score'].mean():.2f}\")" } }, { "id": "step_6", "type": "chart", "name": "Visualize creative video performance patterns", "depends_on": ["step_5"], "params": { "description": "Create charts showing relationship between creative elements and performance", "code": "import matplotlib.pyplot as plt\nimport seaborn as sns\n\n# Set dark theme\nplt.style.use('dark_background')\nfig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, figsize=(16, 12))\nfig.patch.set_facecolor('#0a0a0a')\n\n# 1. Fun score vs reach\nscatter_data = high_performance_creative[high_performance_creative['max_reach'].notna()]\nax1.scatter(scatter_data['fun_score'], scatter_data['max_reach'], \n c='#00ff88', alpha=0.6, s=50)\nax1.set_xlabel('Fun Score', color='#e5e5e5')\nax1.set_ylabel('Max Reach', color='#e5e5e5')\nax1.set_title('Creative Score vs Performance', color='#e5e5e5', fontweight='bold')\nax1.set_facecolor('#0a0a0a')\nax1.tick_params(colors='#e5e5e5')\n\n# 2. Emoji usage impact\nemoji_reach = high_performance_creative.groupby('has_emojis')['max_reach'].mean()\ncolors = ['#ff6b6b', '#00ff88']\nax2.bar(['No Emojis', 'With Emojis'], emoji_reach.values, color=colors, alpha=0.8)\nax2.set_ylabel('Average Max Reach', color='#e5e5e5')\nax2.set_title('Impact of Emoji Usage', color='#e5e5e5', fontweight='bold')\nax2.set_facecolor('#0a0a0a')\nax2.tick_params(colors='#e5e5e5')\n\n# 3. Top brands by creative video count\ntop_brands = high_performance_creative['page_name'].value_counts().head(8)\nax3.barh(range(len(top_brands)), top_brands.values, color='#4ecdc4', alpha=0.8)\nax3.set_yticks(range(len(top_brands)))\nax3.set_yticklabels(top_brands.index, color='#e5e5e5')\nax3.set_xlabel('Number of Creative Video Ads', color='#e5e5e5')\nax3.set_title('Most Active Creative Brands', color='#e5e5e5', fontweight='bold')\nax3.set_facecolor('#0a0a0a')\nax3.tick_params(colors='#e5e5e5')\n\n# 4. Creative trends over time\nvideo_dataset['month'] = pd.to_datetime(video_dataset['first_seen_date']).dt.to_period('M')\nmonthly_creative = video_dataset.groupby('month')['fun_score'].agg(['count', 'mean']).reset_index()\nmonthly_creative = monthly_creative.dropna()\nif len(monthly_creative) > 0:\n ax4.plot(range(len(monthly_creative)), monthly_creative['mean'], \n color='#ff9f43', linewidth=3, marker='o', markersize=6)\n ax4.set_xlabel('Time Period', color='#e5e5e5')\n ax4.set_ylabel('Average Fun Score', color='#e5e5e5')\n ax4.set_title('Creative Content Trends', color='#e5e5e5', fontweight='bold')\n ax4.set_facecolor('#0a0a0a')\n ax4.tick_params(colors='#e5e5e5')\n\nplt.tight_layout()\nsave_chart(fig, 'creative_video_analysis.png')" } }, { "id": "step_7", "type": "python_compute", "name": "Analyze successful video ad copy patterns", "depends_on": ["step_5"], "params": { "description": "Extract and analyze text patterns from top-performing creative video ads", "code": "import re\nfrom collections import defaultdict\n\n# Analyze top performing ads text patterns\ntop_ads_analysis = top_creative_ads[['page_name', 'primary_text', 'headline', 'description', \n 'max_reach', 'fun_score', 'has_emojis', 'creative_indicators']].copy()\n\n# Text length analysis\ntop_ads_analysis['primary_text_len'] = top_ads_analysis['primary_text'].astype(str).str.len()\ntop_ads_analysis['headline_len'] = top_ads_analysis['headline'].astype(str).str.len()\n\n# Extract common phrases (2-4 words)\nall_text = ' '.join(top_ads_analysis['primary_text'].astype(str).tolist())\nwords = re.findall(r'\\b\\w+\\b', all_text.lower())\n\n# Find bigrams and trigrams\nbigrams = [' '.join(words[i:i+2]) for i in range(len(words)-1)]\ntrigrams = [' '.join(words[i:i+3]) for i in range(len(words)-2)]\n\ncommon_bigrams = Counter(bigrams).most_common(15)\ncommon_trigrams = Counter(trigrams).most_common(10)\n\n# Analyze call-to-action patterns\ncta_patterns = top_ads_analysis['primary_text'].astype(str).str.extract(r'(shop now|buy now|get yours|discover|explore|try|experience|see more|learn more|click here)', flags=re.IGNORECASE)\n\n# Create examples for SubSun\nsubsun_examples = []\nfor idx, ad in top_ads_analysis.head(5).iterrows():\n example = {\n 'brand': ad['page_name'],\n 'original_text': ad['primary_text'],\n 'original_headline': ad['headline'],\n 'reach': ad['max_reach'],\n 'creative_elements': ad['creative_indicators']\n }\n subsun_examples.append(example)\n\nprint(\"=== TEXT PATTERN ANALYSIS ===\")\nprint(f\"Average primary text length: {top_ads_analysis['primary_text_len'].mean():.0f} characters\")\nprint(f\"Average headline length: {top_ads_analysis['headline_len'].mean():.0f} characters\")\n\nprint(\"\\n=== MOST COMMON BIGRAMS ===\")\nfor bigram, count in common_bigrams:\n if len(bigram.strip()) > 3: # Filter out short/meaningless pairs\n print(f\"{bigram}: {count}\")\n\nprint(\"\\n=== MOST COMMON TRIGRAMS ===\")\nfor trigram, count in common_trigrams:\n if len(trigram.strip()) > 5: # Filter out short phrases\n print(f\"{trigram}: {count}\")\n\nprint(\"\\n=== TOP PERFORMING CREATIVE ADS SAMPLE ===\")\nfor i, example in enumerate(subsun_examples, 1):\n print(f\"\\n{i}. {example['brand']} (Reach: {example['reach']})\")\n print(f\" Text: {example['original_text'][:100]}...\")\n print(f\" Headline: {example['original_headline']}\")\n print(f\" Creative elements: {example['creative_elements']}\")" } }, { "id": "step_8", "type": "python_compute", "name": "Generate SubSun AI creative video recommendations", "depends_on": ["step_7"], "params": { "description": "Create specific AI creative video concepts tailored for SubSun based on competitor analysis", "code": "# Generate SubSun-specific creative concepts based on successful patterns\n\n# Key insights from analysis\nsuccessful_patterns = {\n 'avg_text_length': int(top_ads_analysis['primary_text_len'].mean()),\n 'avg_headline_length': int(top_ads_analysis['headline_len'].mean()),\n 'top_creative_words': [word for word, count in Counter(all_indicators).most_common(5)],\n 'emoji_boost': high_performance_creative['has_emojis'].mean() * 100,\n 'top_brands': list(high_performance_creative['page_name'].value_counts().head(3).index)\n}\n\n# Create SubSun AI creative concepts\nsubsun_concepts = [\n {\n 'concept_name': 'AI Style Transform',\n 'video_concept': 'Show AI transforming ordinary moments into stylish ones with SubSun sunglasses',\n 'sample_copy': '✨ What if AI could instantly upgrade your style? Watch the magic happen with SubSun! From basic to iconic in seconds. Which vibe matches your mood? 🕶️ #AIStyle #SubSun',\n 'sample_headline': 'AI Meets Style. Your Perfect Look Awaits.',\n 'ai_elements': ['Style transformation', 'Mood detection', 'Instant makeover'],\n 'inspiration_from': 'Pattern of transformation themes in top ads'\n },\n {\n 'concept_name': 'Lens Perspective AI',\n 'video_concept': 'AI analyzes different perspectives/moods through SubSun lenses',\n 'sample_copy': 'See the world differently 🌍 Our AI reveals how SubSun changes your entire perspective. Confidence mode: ON. Adventure mode: ACTIVATED. What mode will you choose? 🔥',\n 'sample_headline': 'AI-Powered Vision. Endless Possibilities.',\n 'ai_elements': ['Perspective shift', 'Mood analysis', 'Vision enhancement'],\n 'inspiration_from': 'Vision/perspective themes common in successful ads'\n },\n {\n 'concept_name': 'Vibe Generator AI',\n 'video_concept': 'AI creates the perfect vibe/aesthetic for any situation with SubSun',\n 'sample_copy': 'Tell our AI your vibe and watch SubSun deliver ✨ Beach day? Business meeting? Date night? We have got the perfect pair. Your style, amplified by AI 🚀',\n 'sample_headline': 'AI Knows Your Vibe. SubSun Delivers.',\n 'ai_elements': ['Vibe matching', 'Situational styling', 'Personalization'],\n 'inspiration_from': 'Vibe/mood keywords frequently used in top performers'\n },\n {\n 'concept_name': 'Instant Cool AI',\n 'video_concept': 'AI detects coolness factor and shows how SubSun maximizes it',\n 'sample_copy': 'Cool level: Loading... 📊 AI analysis complete! SubSun sunglasses = +1000 cool points. Ready to level up your style game? The algorithm never lies 😎',\n 'sample_headline': 'AI Certified Cool. SubSun Approved.',\n 'ai_elements': ['Coolness measurement', 'Style scoring', 'Instant upgrade'],\n 'inspiration_from': 'Cool/awesome language patterns in high-reach ads'\n },\n {\n 'concept_name': 'Trend Prediction AI',\n 'video_concept': 'AI predicts what is trending and positions SubSun as the answer',\n 'sample_copy': 'AI predicts: SubSun styles will dominate 2024 📈 Get ahead of the trend before everyone else catches on. Future-proof your style today 🔮 #TrendAlert',\n 'sample_headline': 'AI Sees The Future. SubSun Leads The Trend.',\n 'ai_elements': ['Trend forecasting', 'Future styling', 'Early adoption'],\n 'inspiration_from': 'Trending/viral language from successful creative ads'\n }\n]\n\nprint(\"=== SUBSUN AI CREATIVE VIDEO CONCEPTS ===\")\nprint(f\"\\nBased on analysis of {len(high_performance_creative)} creative video ads from competitors\\n\")\n\nfor i, concept in enumerate(subsun_concepts, 1):\n print(f\"{i}. {concept['concept_name'].upper()}\")\n print(f\" Video Concept: {concept['video_concept']}\")\n print(f\" Sample Copy: {concept['sample_copy']}\")\n print(f\" Sample Headline: {concept['sample_headline']}\")\n print(f\" AI Elements: {', '.join(concept['ai_elements'])}\")\n print(f\" Inspiration: {concept['inspiration_from']}\")\n print()\n\nprint(\"=== KEY SUCCESS PATTERNS IDENTIFIED ===\")\nprint(f\"• Optimal text length: ~{successful_patterns['avg_text_length']} characters\")\nprint(f\"• Optimal headline length: ~{successful_patterns['avg_headline_length']} characters\")\nprint(f\"• Emoji usage increases reach by {successful_patterns['emoji_boost']:.0f}%\")\nprint(f\"• Top creative themes: {', '.join(successful_patterns['top_creative_words'])}\")\nprint(f\"• Leading creative brands: {', '.join(successful_patterns['top_brands'])}\")" } }, { "id": "step_9", "type": "python_compute", "name": "Create implementation roadmap and budget estimates", "depends_on": ["step_8"], "params": { "description": "Develop actionable implementation plan with timeline and resource requirements", "code": "# Create implementation roadmap for SubSun AI creative videos\n\nimplementation_plan = {\n 'phase_1': {\n 'name': 'Foundation & Testing (Weeks 1-2)',\n 'activities': [\n 'Set up AI video tools (RunwayML, Synthesia, or similar)',\n 'Create brand guidelines for AI content',\n 'Test 2-3 concepts with small budget ($500-1000)',\n 'A/B test against current video creative'\n ],\n 'budget_range': '$2000-4000',\n 'key_metrics': 'CTR, engagement rate, cost per click'\n },\n 'phase_2': {\n 'name': 'Scale & Optimize (Weeks 3-6)',\n 'activities': [\n 'Implement top-performing concept at full scale',\n 'Create variations of successful AI themes',\n 'Develop AI creative templates for quick iteration',\n 'Train team on AI video tools and best practices'\n ],\n 'budget_range': '$5000-10000',\n 'key_metrics': 'ROAS improvement, CPA reduction, reach expansion'\n },\n 'phase_3': {\n 'name': 'Innovation & Expansion (Weeks 7-12)',\n 'activities': [\n 'Experiment with advanced AI features',\n 'Create seasonal AI creative campaigns',\n 'Develop user-generated AI content campaigns',\n 'Integrate AI creative with influencer partnerships'\n ],\n 'budget_range': '$8000-15000',\n 'key_metrics': 'Brand awareness lift, customer acquisition, AOV impact'\n }\n}\n\n# Technical requirements and tools\nai_tools_recommendations = [\n {\n 'tool': 'RunwayML',\n 'use_case': 'AI video generation and editing',\n 'cost': '$15-35/month per user',\n 'pros': 'Easy to use, good for beginners, multiple AI models'\n },\n {\n 'tool': 'Synthesia',\n 'use_case': 'AI presenter videos',\n 'cost': '$30-90/month',\n 'pros': 'Professional AI avatars, multilingual support'\n },\n {\n 'tool': 'Pictory AI',\n 'use_case': 'Text-to-video conversion',\n 'cost': '$19-47/month',\n 'pros': 'Great for social media, automatic captions'\n },\n {\n 'tool': 'Luma AI',\n 'use_case': '3D product visualization',\n 'cost': 'Free tier available',\n 'pros': 'Perfect for sunglasses 3D modeling'\n }\n]\n\n# Success metrics and KPIs\nsuccess_metrics = {\n 'primary_kpis': [\n 'Video completion rate (target: >75%)',\n 'Click-through rate (target: +25% vs current)',\n 'Cost per acquisition (target: -15% vs current)',\n 'Return on ad spend (target: maintain or improve)'\n ],\n 'secondary_kpis': [\n 'Share rate and social engagement',\n 'Brand mention increase',\n 'Organic reach amplification',\n 'Customer lifetime value impact'\n ]\n}\n\nprint(\"=== SUBSUN AI CREATIVE IMPLEMENTATION ROADMAP ===\")\nprint()\nfor phase_key, phase in implementation_plan.items():\n print(f\"{phase['name'].upper()}\")\n print(f\"Budget: {phase['budget_range']}\")\n print(\"Activities:\")\n for activity in phase['activities']:\n print(f\" • {activity}\")\n print(f\"Key Metrics: {phase['key_metrics']}\")\n print()\n\nprint(\"=== RECOMMENDED AI TOOLS ===\")\nfor tool in ai_tools_recommendations:\n print(f\"{tool['tool']}: {tool['use_case']}\")\n print(f\" Cost: {tool['cost']}\")\n print(f\" Benefits: {tool['pros']}\")\n print()\n\nprint(\"=== SUCCESS METRICS TO TRACK ===\")\nprint(\"Primary KPIs:\")\nfor kpi in success_metrics['primary_kpis']:\n print(f\" • {kpi}\")\nprint(\"\\nSecondary KPIs:\")\nfor kpi in success_metrics['secondary_kpis']:\n print(f\" • {kpi}\")\n\ntotal_budget_low = 15000 # 2000+5000+8000\ntotal_budget_high = 29000 # 4000+10000+15000\nprint(f\"\\n=== TOTAL INVESTMENT ESTIMATE ===\")\nprint(f\"3-month program: ${total_budget_low:,} - ${total_budget_high:,}\")\nprint(f\"Expected ROI timeline: 4-8 weeks for initial results\")" } }, { "id": "step_10", "type": "llm_synthesis", "name": "Generate comprehensive strategic deliverable", "depends_on": ["step_9"], "params": { "prompt_template": "Create a comprehensive strategic deliverable for SubSun based on the AI creative video analysis. Structure it as an executive summary with key findings, specific creative concepts, implementation roadmap, and success metrics.\n\nKey data points to incorporate:\n- Total video ads analyzed: {len(competitor_ads)}\n- Creative/fun video ads found: {len(high_performance_creative)}\n- Top performing brands and patterns identified\n- 5 specific AI creative concepts developed for SubSun\n- 3-phase implementation plan with budget ranges\n- Recommended tools and success metrics\n\nMake this actionable and specific to SubSun's DTC sunglasses business, emphasizing ROAS, CPA, and AOV impact potential. Include specific examples and next steps.", "model": "claude-3-haiku" } } ] } ```

```json { "questions": [], "reasoning": "The brief is clear enough to proceed. SubSun wants to explore 'fun AI creative' and needs competitor video examples with similar vibes as inspiration. I have access to the competitor_ads table with 22k+ ads including creative_type and preview_url fields, which should contain video ads from competitors in the database. I can filter for video creatives, analyze their characteristics, and provide actionable examples. The only minor ambiguity is what constitutes 'fun AI creative' but I can infer this means playful, engaging video content that could be AI-generated or AI-enhanced, and identify similar styles from competitor data." } ```