The support metrics that actually predict revenue
In short
Four chat metrics predict revenue: median first-response time (segmented by hour), resolution on first reply, the pre-purchase vs. post-purchase split, and deflection accuracy measured at 24 hours. Volume metrics — messages handled, deflection rate on its own — do not predict anything, because they can be inflated without answering a single customer correctly.
Most sellers track chat metrics because the platform forces them to — Shopee's Chat Response Rate, TikTok Shop's response time badge. Treated as compliance boxes, they get gamed with canned "we'll get back to you" replies that satisfy the algorithm but not the customer.
Treated as a leading indicator, the same numbers tell a very different story: a shopper messaging before checkout is a warmer lead than one who adds to cart silently. How fast and how well you answer that message predicts whether the order happens at all.
The four numbers that matter
First response time (FRT) — median, not average. One 6-hour outlier at 3am shouldn't hide that your daytime FRT is 45 seconds. Segment by hour of day.
Resolution on first reply — did the customer get an answer that closed the loop, or did it open a back-and-forth that raises the odds they abandon the cart while waiting?
Pre-purchase vs. post-purchase split — a slow reply to "does this fit true to size" costs you a sale today; a slow reply to "where's my order" costs you a review tomorrow. They need different SLAs.
Deflection accuracy — the percentage of AI-handled conversations that did not need a human follow-up within 24 hours. Deflection volume alone is a vanity metric; accuracy is the one that protects your rating.
Why this compounds
Shopee and TikTok Shop both feed chat responsiveness into account health and, by extension, into organic reach and ad auction pricing. A store with sub-60-second replies isn't just closing more of today's chats — it's paying less for tomorrow's traffic.
The teams that win this aren't the ones with the most CS headcount. They're the ones who routed the 80% of repetitive questions (sizing, stock, shipping ETA, return policy) to something that never sleeps, and kept humans for the 20% that need judgment.
Frequently asked
- Should I use median or average first-response time?
- Median. A handful of overnight messages answered six hours later drags the average into uselessness while hiding the fact that your daytime replies are fast. Report the median, and segment it by hour of day so you can see exactly which shift is leaking sales.
- Is a high deflection rate a good sign?
- Not on its own. A bot that replies "please contact our support team" deflects perfectly and helps nobody. Measure deflection accuracy instead: of the conversations the AI handled, what share needed no human follow-up within 24 hours.
- Do chat metrics affect Shopee and TikTok Shop rankings?
- Both platforms feed chat responsiveness into seller account health, which influences organic visibility and how much you pay in the ad auction. Slow replies are therefore a traffic cost, not only a conversion cost.
Read in中文Bahasa Malaysia