ROI Case File No.362 | 'VirtuShoes' Biannual Photography Ordeal'

Chapter 1: Biannual Photography as Ritual — A 3 Million Yen Burden

The day after resolving Bravura Retail's PPM case, a consultation arrived regarding image production efficiency. Volume 29, "The Pursuit of Reproducibility," Case 362 tells the story of discovering optimal solutions through customer segmentation.

"Detective, our photography is hell. Twice a year, we photograph 120 new shoe models. Model arrangement, studio booking, prop preparation. All manual. And the cost is 3 million yen per session. 6 million yen annually. However, 70% of photographed images are barely used."

Miki Takahashi, Marketing Director of VirtuShoes, a Harajuku native, visited 221B Baker Street with an exhausted expression. In her hands, she held a portfolio of beautiful shoe photographs taken by professional photographers, contrasting sharply with a cost-effectiveness analysis report titled "Image Production Cost Analysis 2024-2025."

"We operate a ladies' shoe brand. 45 employees. Annual revenue of 1.2 billion yen. 12 physical stores, proprietary EC, mall-type EC. However, image production is inefficient."

VirtuShoes' Current State: - Founded: 2015 (Ladies' shoe brand) - Employees: 45 - Annual Revenue: 1.2 billion yen - Sales Channels: 12 physical stores, proprietary EC, Rakuten/Amazon - Issues: High photography costs, massive work hours, low image utilization

Deep anxiety resonated in Takahashi's voice.

"The photography breakdown is as follows. Studio rental: 120,000 yen/day × 3 days = 360,000 yen. Professional photographer: 80,000 yen/day × 3 days = 240,000 yen. Models: 50,000 yen/day × 3 days × 2 people = 300,000 yen. Hair and makeup: 30,000 yen/day × 3 days = 90,000 yen. Stylist: 40,000 yen/day × 3 days = 120,000 yen. Props and backgrounds: 300,000 yen. Image retouching: 120 images × 8,000 yen = 960,000 yen. Total: 2.37 million yen. Including direction and preparation work hours, it exceeds 3 million yen."

Photography Schedule Reality:

Preparation Period (2 weeks before photography): - Create list of 120 new shoe models (4 hours) - Arrange models, studio, and photographer (8 hours) - Plan photography concept and create materials (16 hours) - Procure props and costumes (12 hours) Total preparation hours: 40 hours

Photography Period (3 days): - Day 1: Casual line 40 models (8 hours) - Day 2: Business line 40 models (8 hours) - Day 3: Party line 40 models (8 hours) Total photography hours: 24 hours (3 staff = 72 hours)

Post-processing Period (1 week after photography): - Image selection (16 hours) - Retouching instructions (8 hours) - Final confirmation (4 hours) Total post-processing hours: 28 hours

Total hours: 140 hours (17.5 days) Twice annually = 280 hours/year

Takahashi sighed deeply.

"There's another problem. Low utilization of photographed images. We photograph 120 models, but only 30 models are frequently used on SNS. Only 20 models are bestsellers on EC sites. Only 15 models are used for in-store POP. In other words, 70% of photographed images are barely used. Cost-effectiveness is terrible."

Chapter 2: Uniform Photography for All Models — Customer Purchasing Behavior Invisible

"Ms. Takahashi, do you believe photographing all shoes uniformly will resonate with all customers?"

Takahashi showed a puzzled expression at my question.

"Eh, isn't that so? I thought customers would be satisfied if all models were beautifully photographed."

Current Understanding (Uniform Photography Model): - Expectation: Customer satisfaction improves through high-quality photography of all models - Problem: Customer segments and purchasing behavior remain invisible

I explained the importance of optimizing image strategy through customer segmentation.

"The problem is the idea of 'photographing all shoes uniformly.' RFM—Recency, Frequency, Monetary. Most recent purchase date, purchase frequency, purchase amount. We segment customers along these three axes and establish optimal image strategies for each segment. And by utilizing AI image generation, we achieve reproducible efficiency."

⬜️ ChatGPT | Catalyst of Conception

"Don't photograph all models uniformly. Segment customers through RFM and establish optimal image strategy"

🟧 Claude | Alchemist of Narrative

"Images are always 'the first point of contact that captures customers' hearts.' Who sees what is everything"

🟦 Gemini | Compass of Reason

"Classify customers through RFM. Premium customers, dormant customers, new customers. Change image strategy by segment"

The three members began analysis. Gemini developed the "RFM Matrix" on the whiteboard.

RFM's Three Axes: 1. Recency (Most Recent Purchase Date): Days elapsed since last purchase 2. Frequency (Purchase Frequency): Number of purchases in past year 3. Monetary (Purchase Amount): Total purchase amount in past year

"Ms. Takahashi, let's first analyze customer data through RFM."

Chapter 3: Phase 1 — Segmenting Customers Through RFM

Step 1: Data Collection (1 week)

Analysis Target: - Proprietary EC members: 8,500 - Rakuten/Amazon purchasers: 12,000 (email address acquired) - Total: 20,500

Collected Data: - Customer ID - Most recent purchase date - Number of purchases in past year - Total purchase amount in past year

Step 2: RFM Scoring (1 week)

Recency (Most Recent Purchase Date): - R5: Purchased within 0-30 days (Score 5) - R4: Purchased within 31-90 days (Score 4) - R3: Purchased within 91-180 days (Score 3) - R2: Purchased within 181-365 days (Score 2) - R1: No purchase for 365+ days (Score 1)

Frequency (Purchase Frequency): - F5: 6+ purchases per year (Score 5) - F4: 4-5 purchases per year (Score 4) - F3: 2-3 purchases per year (Score 3) - F2: 1 purchase per year (Score 2) - F1: Less than 1 purchase per year (Score 1)

Monetary (Purchase Amount): - M5: 200,000+ yen per year (Score 5) - M4: 100,000-200,000 yen per year (Score 4) - M3: 50,000-100,000 yen per year (Score 3) - M2: 20,000-50,000 yen per year (Score 2) - M1: Under 20,000 yen per year (Score 1)

Step 3: Customer Segment Classification (1 week)

Classify into 5 tiers based on total RFM score:

Segment 1: Premium Customers (RFM total 13-15) - Number: 1,230 (6%) - Characteristics: Recent purchase, high frequency, high amount - Average annual purchase: 180,000 yen - Total purchase amount: 221.4 million yen (55% of total)

Segment 2: Promising Customers (RFM total 10-12) - Number: 2,460 (12%) - Characteristics: Relatively recent purchase, medium frequency, medium amount - Average annual purchase: 70,000 yen - Total purchase amount: 172.2 million yen (43% of total)

Segment 3: General Customers (RFM total 7-9) - Number: 4,100 (20%) - Characteristics: Time elapsed since purchase, low frequency, low amount - Average annual purchase: 8,000 yen - Total purchase amount: 32.8 million yen (8% of total)

Segment 4: Dormant Customers (RFM total 4-6) - Number: 6,150 (30%) - Characteristics: Long period without purchase, past purchases exist - Average annual purchase: 1,000 yen - Total purchase amount: 6.15 million yen (1.5% of total)

Segment 5: New Customers (RFM total 3 or below) - Number: 6,560 (32%) - Characteristics: Only initial purchase or no purchase - Average annual purchase: 500 yen - Total purchase amount: 3.28 million yen (0.8% of total)

Critical Discovery: - Premium customers (6%) account for 55% of sales - Promising customers (12%) account for 43% of sales - In other words, top 18% accounts for 98% of sales

Chapter 4: Phase 2 — Segment-Specific Image Strategy and AI Utilization

Step 4: Establish Segment-Specific Image Strategy (2 weeks)

Segment 1: Premium Customers (1,230) - Image strategy: Professional photography, high quality, exclusive early release - Target models: 15 new party line models - Distribution channels: SNS (Instagram), member-exclusive email - Photography frequency: Twice yearly (biannually) - Photography cost: 15 models × 25,000 yen = 375,000 yen/session × 2 = 750,000 yen/year

Segment 2: Promising Customers (2,460) - Image strategy: AI image generation, high-quality style, regular distribution - Target models: 20 popular business line models - Distribution channels: SNS (Instagram), EC product pages - Generation frequency: Twice monthly - AI generation cost: 20 models × 500 yen/image × 24 times = 240,000 yen/year

Segment 3: General Customers (4,100) - Image strategy: AI image generation, standard quality, seasonal distribution - Target models: 30 standard casual line models - Distribution channels: EC product pages, in-store POP - Generation frequency: Seasonally (4 times yearly) - AI generation cost: 30 models × 500 yen/image × 4 times = 60,000 yen/year

Segment 4: Dormant Customers (6,150) - Image strategy: AI image generation, simple, repurchase promotion - Target models: 10 related products based on past purchase history - Distribution channels: Retargeting ads, email - Generation frequency: Twice yearly - AI generation cost: 10 models × 500 yen/image × 2 times = 10,000 yen/year

Segment 5: New Customers (6,560) - Image strategy: Reuse existing images, minimize cost - Target models: 5 bestseller models - Distribution channels: Advertising, landing pages - Generation frequency: None (use existing images) - Cost: 0 yen

Step 5: Introduce AI Image Generation (Month 1-2)

Select AI Image Generation Tool: - Tools: Midjourney + Stable Diffusion - Use: Background replacement, model composition, scene generation for shoes

AI Image Generation Flow:

Base on Professional Photography Images (Premium Customers): 1. Professionally photograph 15 new models in studio (white background, product only) 2. Photography cost: 15 models × 5,000 yen = 75,000 yen 3. Generate backgrounds, models, scenes with AI - Pattern 1: Cafe scene - Pattern 2: Office scene - Pattern 3: Party scene 4. Generate 3 patterns per model 5. AI generation cost: 15 models × 3 patterns × 500 yen = 22,500 yen 6. Total: 75,000 + 22,500 = 97,500 yen

Remake Existing Images (Promising/General Customers): 1. Change backgrounds and colors based on past photography 2. AI generation cost: 50 models × 500 yen = 25,000 yen

Month 3: Measure Effects

KPI 1: Photography Cost - Before: 3 million yen × 2 times/year = 6 million yen/year - After: 750,000 yen (professional photography for premium customers) + 310,000 yen (AI generation) = 1.06 million yen/year - Reduction rate: 82% - Amount saved: 4.94 million yen/year

KPI 2: Production Work Hours - Before: 280 hours/year (17.5 days × 2 times) - After: 48 hours/year (professional photography 1 time 2 days + AI generation work 1 day) - Reduction rate: 83% - Time saved: 232 hours/year

KPI 3: Image Utilization Rate - Before: 30% (only 36 of 120 models frequently used) - After: 95% (76 of 80 models used, segment-specific optimization) - Improvement: +65 points

KPI 4: Premium Customer Engagement Rate (Instagram) - Before: 3.2% (likes/comments rate) - After: 7.8% (priority distribution, high-quality images) - Improvement: +4.6 points

Annual Effects:

Photography Cost Reduction: - 4.94 million yen/year

Labor Cost Reduction Through Work Hour Reduction: - 232 hours × 3,500 yen = 812,000 yen/year

Sales Increase Through Improved Engagement: - Premium customer SNS-driven purchase increase: average +15% - 1,230 people × 180,000 yen × 15% = 33.21 million yen/year

Total Annual Effect: - 4.94 million + 812,000 + 33.21 million = 38.96 million yen/year

Investment: - AI image generation tools: 360,000 yen annually - System development: 800,000 yen initial

ROI: - (38.96 million - 1.16 million) / 1.16 million × 100 = 3,259% - Payback period: 1.16 million ÷ 38.96 million = 0.03 years (11 days)

Chapter 5: Detective's Diagnosis — Discovering Optimal Solutions Through Customer Segmentation

That night, I contemplated the essence of RFM.

VirtuShoes held a uniform strategy of "photographing all shoes uniformly." However, customer purchasing behavior is diverse.

By classifying customers into 5 segments through RFM analysis, a critical discovery emerged. Premium customers (6%) account for 55% of sales. For this 6%, provide high-quality professional photography; for other segments, optimize efficiency through AI image generation.

Through segment-specific image strategy, we reduced photography costs by 82% (6 million yen → 1.06 million yen), work hours by 83% (280 hours → 48 hours), and improved image utilization by 65 points (30% → 95%). And sales increased by 33.21 million yen through improved premium customer engagement.

Annual effect: 38.96 million yen, ROI: 3,259%, payback period: 11 days.

"Don't photograph all models uniformly. Segment customers through RFM—Recency, Frequency, Monetary—and establish optimal image strategy. High quality for premium customers, AI generation efficiency for others. Segment-specific optimization creates reproducible efficiency."

The next case will also depict the moment of discovering optimal solutions through customer segmentation.

"RFM—Recency, Frequency, Monetary. Segment customers and establish optimal strategy. Don't treat everything uniformly. Segment-specific optimization realizes true efficiency"—From the Detective's Notes

Related Files