Situation
The store sold items priced between $35–$115, targeting women aged 28–42, with an average order value of $65. Cart abandonment was at 71% — above the premium fashion segment benchmark (65%). Recovery emails converted at 4.2%, below the industry benchmark of 6%.
The problem wasn’t price. It was hesitation. Customers reached checkout, stopped at the address or shipping field, and left.
Decision
The team tested three approaches in parallel:
- Discount pop-up — triggered when exit intent was detected
- Human chat — agent available from 9am to 10pm
- AI conversational agent — active 24h, trained with store FAQ and product data
The agent was configured with a specific rule: it only started a conversation if the user had been idle on the same checkout step for more than 12 seconds. No aggressive pop-ups on page entry.
Result
After 90 days:
- Abandonment dropped from 71% to 47% (34% reduction)
- The agent handled 4,300 conversations in the period
- 68% of conversations were about delivery time or return policy
- Post-conversation conversion: 31% (vs. 4.2% from recovery emails)
- Positive ROI on day 47
The discount pop-up converted more in the short term, but trained customers to wait for discounts. It was discontinued after 30 days.
Learnings
The time-based trigger (12 seconds of inactivity) was the most valuable insight. Users who leave immediately don’t want to talk — they either want to buy or they don’t. Users who pause are the ones with doubts.
The second finding: 68% of questions were about shipping and returns. The store updated the product page to surface this information more prominently, and the agent’s conversation volume dropped by 22% — a sign that the root cause had been addressed.