Case Study: How a $5M E-commerce Business Deflected 52% of Support Tickets

This case study covers a real deployment of Sierra at a $5M/year e-commerce business selling home goods. The business was handling 800-1,000 support tickets per month with a 5-person team, and response times were stretching to 6+ hours during peak periods. Within 90 days of deploying Sierra, they achieved 52% ticket deflection and dramatically improved response times.

The challenge

The business was growing 40% year-over-year, but support staffing wasn't keeping pace. The 5-person support team was overwhelmed:

• 800-1,000 tickets per month, growing 10% monthly
• Average response time: 6.4 hours (including queue time)
• CSAT score: 78 (below industry average of 82)
• Support costs: $8,500/month in labor
• Team was stressed, turnover was increasing

The business needed to scale support without proportional headcount increases.

The solution

After evaluating several options (Intercom Fin, Zendesk AI, Ada), the business chose Sierra for its purpose-built support focus and strong Shopify integration.

Implementation timeline

• Week 1: Connected Sierra to Shopify, Gorgias (help desk), and knowledge base. Configured basic responses.
• Week 2: Trained Sierra on product catalog, return policy, shipping procedures. Tested with historical tickets.
• Week 3: Soft launch to 25% of tickets. Monitored performance and refined.
• Week 4: Expanded to 50% of tickets.
• Week 5-6: Expanded to 100% of tickets with human backup.

What Sierra handles

• Order status inquiries (30% of tickets)
• Shipping questions and tracking (20%)
• Return and exchange requests (15%)
• Product questions — sizing, materials, compatibility (12%)
• Billing inquiries (8%)

Human agents handle: complex complaints, VIP customers, product defects, social media escalations.

The results (90 days)

Detailed metrics

• Ticket deflection: 52% (432 of 830 monthly tickets resolved by Sierra)
• Resolution time: 2.1 minutes average (vs 6.4 hours for human-handled tickets)
• CSAT: 81 for Sierra-resolved tickets (vs 80 for human-resolved)
• Support costs: Reduced from $8,500/month to $4,300/month (labor) + $399/month (Sierra) = $4,699/month total
• Team satisfaction: Support team reported less stress, more time for complex work

What worked well

• Shopify integration was seamless. Sierra could look up orders, process refunds, and check inventory without custom integration work.
• Knowledge base quality mattered. Tickets that Sierra couldn't resolve were almost always cases where the knowledge base lacked relevant information.
• Gradual rollout built confidence. Starting at 25% and expanding let the team catch issues early.
• Escalation worked well. Sierra correctly identified when to hand off to humans, and handoffs included full context.

What didn't work well

• Complex return scenarios. Multi-item returns with partial refunds confused Sierra initially. Required additional training.
• Angry customers. Sierra's polite-but-not-empathetic tone was wrong for upset customers. Added sentiment-based escalation.
• Product launch periods. When new products launched, Sierra didn't have enough information to answer questions. Required rapid knowledge base updates.

ROI analysis

Annual savings: $50,400 (labor savings) - $4,788 (Sierra cost) = $45,612 net savings

ROI: 954% in year one

Payback period: 2.6 weeks

Beyond the direct savings, the business saw improved CSAT, faster response times, and reduced team stress — all of which contribute to long-term business value.

Lessons for other businesses

• Invest in your knowledge base first. Sierra is only as good as the information it has access to.
• Plan for gradual rollout. Don't go live with 100% of tickets on day one.
• Monitor sentiment. Escalate angry customers to humans immediately.
• Update knowledge base for product launches. Agents need current information.
• Set realistic expectations. 50% deflection is achievable; 100% is not.

Next steps

See our Sierra review for full capabilities, and our guide to building a support agent for implementation details.