Case Study: Shoe Store Increased Sales 34% with AI Recommendations

StrideFit (name changed for confidentiality) is a D2C footwear brand based in Bangalore selling athletic and casual shoes primarily to young professionals and fitness enthusiasts.

Their product range includes:

• Running shoes (15 styles, multiple colors each)
• Training shoes (12 styles)
• Casual sneakers (20 styles)
• Walking shoes (8 styles)
• Accessories (insoles, socks, shoe care products)

Pre-optimization metrics (October 2024):

• Monthly revenue: ₹32-38 lakh
• Monthly traffic: 45,000 visitors
• Conversion rate: 1.8%
• Average order value: ₹2,450
• Product catalog: 180 SKUs (55 base styles × variants)
• 8-12 new products added monthly

On the surface, these are respectable numbers. But the founder, Arjun, knew they were leaving money on the table.

Understanding baseline metrics is the first step to optimization

The Problem

StrideFit had installed a popular product recommendation app six months earlier. The app promised to show "related products" and "frequently bought together" suggestions.

The reality was different.

Issue 1: Manual Setup Was Crushing Them

The app required manual configuration for every product. For each of their 55 styles (each with 3-5 color variants), someone had to:

1. Open the product in the app dashboard
2. Search and select 4-5 related products
3. Search and select 3-4 upsell products
4. Search and select 3-4 cross-sell products
5. Save and move to the next product

Arjun's team calculated: 4 minutes per product × 180 products = 12 hours of initial setup.

They completed it over two weekends. It was exhausting, but done.

Then the real problem emerged.

Issue 2: New Products Broke Everything

StrideFit launched 8-12 new products every month—new colors, seasonal styles, collaborations.

Each new product needed:

• Its own recommendations configured (4 minutes)
• To be added to existing products' recommendations (where relevant)

The second part is what killed them.

When they added "Running Shoe – Neon Green," it should appear as a related product on their other running shoe pages. But the app did not do this automatically. Someone had to manually go to each existing running shoe product and add the new color to recommendations.

With 15 running shoe styles, that meant editing 15 products every time they added one new color.

Arjun's team fell behind. New products sat without recommendations for weeks. Existing products never showed new arrivals as related items. The recommendation sections became stale.

After three months, only 40% of products had up-to-date recommendations.

Manual product recommendation setup becomes unsustainable at scale

Issue 3: The Recommendations Were Not That Good

Even when configured, the recommendations were mediocre.

The app matched products based on collections and tags. If two shoes shared a "Running" tag, the app considered them related.

But tags are blunt instruments. "Running Shoe – Road Racing" and "Running Shoe – Trail Heavy-Duty" are both tagged "Running" but target completely different customers. Road racers want lightweight speed; trail runners want durability and grip.

The app showed them as related products. Customers ignored these irrelevant suggestions.

Arjun watched session recordings. Customers would scroll past the "Related Products" section without engaging. The recommendation widgets generated clicks on only 3-4% of views—far below the 12-15% industry benchmark.

The Real Cost

The problems were costing them:

Time Cost:

• 2-3 hours/week maintaining recommendations
• 130 hours/year = ₹65,000 in labor (at ₹500/hour)

Opportunity Cost:

• Stale recommendations reduced discovery
• New products invisible to existing customers
• Lower average order value from poor cross-sells
• Estimated revenue loss: ₹4-6 lakh/year

Psychological Cost:

• Team frustration with manual work
• Feeling like they were constantly behind
• Technology that created work instead of solving it

Arjun knew there had to be a better way.

The Solution

In late October 2024, Arjun discovered AI CrossSell, Upsell & Related by Scalefront.

The promise seemed too good: "Zero manual setup. AI automatically generates recommendations. New products sync automatically."

He was skeptical. But the app had a free tier, so the risk was zero.

He installed it on a Friday afternoon.

AI-powered recommendations eliminate manual configuration

Installation: 2 Minutes

Literally 2 minutes:

1. Install from Shopify App Store
2. Click "Sync Products"
3. Enable the widget on product pages
4. Done

The AI immediately began analyzing their catalog.

How the AI Works

The app analyzes five data points for every product:

1. Product Name: "Men's CloudRun Pro Running Shoes – Lightweight Road Racing"
2. Description: Full description including features, use case, materials
3. Images: Visual analysis to understand style, color, design elements
4. Categories/Collections: Running shoes, men's footwear, road racing
5. Tags: Running, lightweight, road racing, neutral cushioning

From this, the AI builds a "product understanding"—essentially a detailed profile of what each product is, who it is for, and how it relates to other products.

For the CloudRun Pro example above, the AI understands:

• Category: Men's running shoes, road racing subcategory
• Attributes: Lightweight, neutral cushioning, responsive, designed for speed
• Use case: Road racing, training runs, performance-focused runners
• Style: Technical athletic, not casual
• Price tier: Premium (₹4,299)

With this understanding, the AI generates three types of recommendations:

Similar Products (Related):

• Other lightweight running shoes
• Other road racing shoes
• Same shoe in different colors

Upsells:

• More premium models with advanced features
• Same shoe with upgraded cushioning technology

Cross-Sells:

• Performance running socks
• Insoles for runners
• Shoe care products
• Running accessories

All of this happened automatically. No manual configuration required.

AI analyzes multiple product attributes to generate intelligent recommendations

The Critical Feature: Automatic New Product Sync

When StrideFit adds a new product, the AI:

1. Analyzes the new product immediately
2. Generates recommendations for it (pulling from existing catalog)
3. Updates existing products to include the new product where relevant

That third step is what changes everything.

Add "CloudRun Pro – Electric Blue" and the AI automatically:

• Creates recommendations for the new color
• Adds it to the related products on other CloudRun Pro colors
• Adds it to similar lightweight running shoes
• Shows it in the "new arrivals" widget

Zero manual work. Instant integration across the entire store.

Manual Override Capability

While the AI works automatically, the app also allows manual adjustments.

For key products, Arjun's team could:

• Pin specific products to always show first
• Exclude certain products from recommendations
• Add specific cross-sells the AI missed

This hybrid approach—automatic baseline with manual overrides—gave them control without requiring constant maintenance.

Implementation Timeline

Week 1: Testing and Validation

Arjun installed the app on a Friday but did not enable it publicly. First, he wanted to verify the recommendations made sense.

He spot-checked 20 products across different categories:

• Running shoes: Recommendations were accurate. Similar styles, appropriate cross-sells.
• Casual sneakers: Good matches. The AI distinguished between sporty casual and fashion casual.
• Walking shoes: Relevant suggestions focused on comfort features.
• Accessories: Appropriately paired with relevant footwear types.

The recommendations were better than their manually configured ones.

Thorough testing ensures AI recommendations meet quality standards

Week 2: Soft Launch

They enabled the AI-powered widget on 50% of product pages (A/B test) while keeping the old app on the other 50%.

Early metrics looked promising:

• AI widget click-through rate: 11.2%
• Old widget click-through rate: 3.8%
• AI recommendations added to cart: 8.7%
• Old recommendations added to cart: 2.4%

The AI was outperforming 3:1.

Week 3: Full Rollout

They replaced the old app entirely and enabled AI recommendations across:

• All product pages
• Cart page ("Frequently bought together")
• Collection pages ("Similar styles")
• Homepage ("Recommended for you")

They also uninstalled the old recommendation app, saving ₹3,500/month.

Week 4: New Product Test

The real test came when they launched a new collection: 6 new training shoe styles in 4 colors each (24 new SKUs).

With the old app, this would have required:

• 24 products × 4 minutes = 96 minutes of new product configuration
• Plus manually adding each to existing products' recommendations
• Total time: 3-4 hours

With the AI app: Zero time. The products synced automatically overnight.

Arjun checked the next morning. All 24 new products had relevant recommendations. Existing training shoes now showed the new styles. It just worked.

The Results

We tracked results for 8 weeks post-implementation (November-December 2024).

Primary Metrics

Metric	Before	After	Change
Conversion Rate	1.8%	2.2%	+22%
Average Order Value	₹2,450	₹3,215	+31%
Add-to-Cart from Recommendations	2.4%	8.7%	+263%
Products per Order	1.12	1.38	+23%
Monthly Revenue	₹35.2L (avg)	₹47.1L (avg)	+34%

34% revenue increase from AI-powered product recommendations

Time Savings

Task	Before	After	Time Saved
Initial setup	12 hours	2 minutes	12 hours
Weekly maintenance	2-3 hours	0 hours	2.5 hours/week
New product integration	45 min/product	0 minutes	~9 hours/month
Annual time savings	—	—	140 hours

At ₹500/hour, that is ₹70,000/year in labor savings alone.

Customer Behavior Changes

Session recording analysis showed changed browsing patterns:

Before:

• Customers viewed 3.2 products per session
• 62% viewed only the product they clicked from ads
• "Related Products" section ignored in 92% of sessions

After:

• Customers viewed 4.8 products per session
• 48% explored 3+ products via recommendations
• "Related Products" clicked in 11.2% of sessions

Customers were discovering more products, engaging longer, and buying more.

Improved product discovery leads to higher engagement and conversions

Specific Example: The Cross-Sell Win

One significant win was accessories.

Previously, only 8% of shoe purchases included accessories (socks, insoles, care products). The old app rarely suggested accessories because setting up those cross-sells for every shoe was tedious.

The AI automatically identified logical accessory pairings:

• Running shoes → performance socks + insoles
• Casual sneakers → no-show socks + protector spray
• Walking shoes → comfort insoles

Accessory attachment rate increased from 8% to 19%.

Average accessory add-on value: ₹380.

With ~800 monthly shoe orders, that is an additional 88 accessories sold per 100 orders × 8 = ₹66,000/month in accessory revenue.

Annual impact from accessories alone: ₹7.9 lakh.

What Made It Work

Looking back, Arjun identified key factors in the success:

1. Zero Configuration Removed Friction

The biggest win was not setting up recommendations. Just install and it works. This meant:

• Fast implementation (days, not weeks)
• No ongoing maintenance burden
• Team could focus on marketing and operations

2. Automatic New Product Sync Was Game-Changing

This single feature saved 9+ hours/month and ensured new products were instantly discoverable across the store.

3. The Recommendations Were Actually Relevant

Unlike tag-based or collection-based matching, the AI understood product attributes. "Lightweight running shoe" recommended other lightweight running shoes—not just any shoe tagged "Running."

Relevant recommendations drive higher engagement and conversion

4. Multiple Placement Options

Having recommendations on product pages, cart, collections, and homepage created multiple discovery touchpoints. Customers who ignored them on one page engaged on another.

5. The Price Was Right

Free tier for testing, then very affordable compared to saving ₹3,500/month from the old app plus ₹70k/year in labor.

Challenges and Learnings

Not everything was perfect. Here is what they learned:

Challenge 1: Product Data Quality Matters

The AI works by analyzing product information. Products with detailed descriptions and accurate tags got better recommendations than products with minimal data.

Solution: They spent a few hours improving product descriptions for key items. Adding details like "neutral cushioning," "trail durability," "fashion-forward design" helped the AI make better matches.

Lesson: If you want AI to understand your products, give it good data.

Challenge 2: Some Manual Overrides Were Needed

For 5-10% of products, the automatic recommendations needed tweaking. For example:

• Limited edition shoes should upsell to other limited editions (AI initially showed standard models)
• Kids' shoes should only cross-sell kids' accessories (AI occasionally suggested adult socks)

Solution: Manual override feature let them pin specific products or exclude irrelevant ones.

Lesson: AI baseline + manual refinement is more powerful than either alone.

Manual overrides for edge cases ensure perfect recommendations

Challenge 3: Widget Styling Required Adjustment

The default widget worked but did not perfectly match their theme's aesthetics initially.

Solution: Support team helped customize colors, fonts, and layout to match their brand.

Lesson: Plan for minor customization to make widgets feel native to your store.

Challenge 4: Monitoring Performance Still Required

While the AI ran automatically, they still needed to monitor:

• Which recommendations performed best
• Whether new categories needed data improvements
• If seasonal products were being handled correctly

Solution: Weekly 10-minute check of the analytics dashboard.

Lesson: "Automatic" does not mean "unattended." You still monitor results.

ROI Analysis

Let us look at the financial impact:

Direct Revenue Increase:

• Pre-implementation: ₹35.2L/month average
• Post-implementation: ₹47.1L/month average
• Monthly increase: ₹11.9L
• Annual increase: ₹1.43 crore

Cost Savings:

• Eliminated previous app: ₹3,500/month = ₹42,000/year
• Labor time saved: 140 hours × ₹500 = ₹70,000/year
• Total annual savings: ₹1.12 lakh

Total Annual Impact:

• Revenue increase: ₹1.43 crore
• Cost savings: ₹1.12 lakh
• Combined benefit: ₹1.44 crore

Investment:

• App cost: Free tier (they later upgraded to premium for advanced features: minimal monthly cost)
• Setup time: 2 minutes
• Customization: Included in app support

The ROI is absurd. Near-zero investment, ₹1.44 crore annual impact.

Exceptional ROI from AI-powered recommendations

Key Takeaways

If you run a product-based ecommerce store, StrideFit's experience offers lessons:

1. Manual Recommendation Setup Does Not Scale

If you have 50+ products, manually configuring recommendations is impractical. The initial setup takes forever, and maintenance becomes impossible when adding products regularly.

2. Collection/Tag-Based Matching Is Not Enough

Simple rule-based systems ("show products from same collection") generate weak recommendations. Customers ignore them. You need semantic understanding of what products actually are.

3. Automatic New Product Integration Is Critical

The ability to add products and have them automatically integrated into recommendations across your store is worth its weight in gold. This single feature saved StrideFit 9+ hours/month.

4. AI Can Outperform Manual Configuration

Even with humans carefully selecting recommendations, the AI performed better—because it analyzed every product consistently and identified patterns humans miss at scale.

5. Recommendations Impact More Than Just AOV

Beyond average order value, good recommendations:

• Increase engagement (more products viewed per session)
• Reduce bounce rates (customers explore instead of leaving)
• Surface new products (customers discover recent launches)
• Improve conversion rates (relevant suggestions feel helpful)

6. Implementation Speed Matters

The faster you can implement and start seeing results, the sooner you can iterate and optimize. Waiting weeks for manual setup delays revenue improvement.

Conclusion

StrideFit's results are not unusual. They are typical for stores moving from manual or rule-based recommendations to AI-powered automatic matching.

The combination of:

• Zero configuration required
• Automatic new product sync
• Semantic understanding of products
• Multiple placement options

...creates a recommendation experience that customers actually engage with.

For StrideFit, this translated to 34% revenue increase and 140 hours/year in time savings.

If you are currently:

• Manually configuring product recommendations
• Using collection/tag-based matching
• Spending hours on new product setup
• Seeing low engagement on recommendation widgets

...you are likely experiencing the same problems StrideFit had.

The solution is AI-powered recommendations that actually work.

Ready to replicate these results in your store?

Try our AI CrossSell, Upsell & Related app and let AI do the heavy lifting:

• Zero manual setup
• Automatic new product sync
• Free to install and test

Or if you need custom ecommerce solutions beyond product recommendations, we build custom Shopify apps and optimizations tailored to your specific business.

Schedule a consultation with our team to discuss your store's optimization strategy.