Common wisdom says customers want fast, friendly support. That is true — but it is not the whole picture. When you look closely at what separates highly-rated support interactions from disappointing ones, more nuanced patterns emerge. Here is what the data shows.
Speed matters — but only for first contact
When shoppers describe what they want from support, speed consistently ranks first: they want a reply within 2 minutes. But here is the interesting finding: once contact is made, speed drops in importance. A customer who gets a first reply in 90 seconds but has to send three follow-up messages rates the interaction lower than one who waited 3 minutes for a single, complete answer.
The takeaway is that perceived responsiveness matters more than raw speed. A fast acknowledgement followed by a thorough answer beats a blizzard of quick but incomplete replies.
Clarity beats cleverness every time
Online shoppers do not want eloquent writing or thoughtful prose from a support interaction. They want a clear, direct answer to their specific question. "Your order ships tomorrow via PostNL, tracking number 3SVENT123456" dramatically outperforms "We are fully committed to timely delivery and your order is progressing through our fulfillment pipeline." Same information. Completely different impact.
This has significant implications for how AI systems should be configured. The temptation is to make AI sound sophisticated. The better instinct is to make it sound helpful — which means being specific, direct, and brief.
First contact resolution is the biggest satisfaction driver
The single strongest predictor of customer satisfaction is resolving the issue in one conversation without any follow-up required. Shoppers who had to send a second message — or worse, escalate to email after a chat — were twice as likely to be dissatisfied, even when their issue was ultimately resolved correctly.
This means front-loading responses is worth the extra effort. Instead of "let me check on that," take the 30 extra seconds to find the complete answer and deliver it all at once. The efficiency gain from doing fewer follow-ups outweighs the marginal extra time per first response.
Empathy makes bad news land better
Counterintuitively, some of the highest-rated support interactions involved delivering genuinely bad news — an item out of stock, a delayed delivery, a return denied outside the policy window. What separated the highly-rated from the poorly-rated ones was not the outcome, but the acknowledgement of the customer's situation.
"I completely understand how frustrating that is — here is what I can do to help" consistently outperforms "Unfortunately, the item is out of stock" even though both convey the same core information. Customers do not expect you to fix every problem. They do expect you to care that the problem exists.
What this means for AI and human support
AI excels at speed, clarity, and first contact resolution on well-defined questions. It does less well at empathy and nuanced judgment on complex situations. The optimal support setup uses AI to handle the fast, factual queries — product availability, delivery times, policy questions — while routing anything that requires human judgment or emotional sensitivity to an agent. That division of labor plays to both strengths and produces support interactions that customers consistently rate highly.