The Analytics page tells you whether the AI is doing its job, and where the next 5% of efficiency hides. Here's what each metric actually means.
AI Resolution Rate
The share of messages the AI handled end-to-end without escalating to you. A healthy store sits at 85–95%. Below 70% usually means: brand voice is too vague, FAQs are missing, or the confidence threshold is set too high.
Average Response Time
Measured from the moment the customer message lands to the moment Taskra sends the reply (or flags it for you). Autopilot replies typically land in 1–4 seconds.
CSAT (Customer Satisfaction)
When Taskra detects a thank-you or positive sentiment in the reply chain, we treat the conversation as positive CSAT. It's a directional signal — not survey-accurate, but useful for trend detection.
Channel Split
How your traffic divides between Email and on-store Chat. Sellers usually find chat skews higher to pre-purchase questions, email to post-purchase. Use this to decide where to invest replies first.
Top Question Categories
A live list of what customers ask most often. Every question becomes an opportunity to either:
- Fix the underlying issue (e.g. unclear shipping timeline on product pages)
- Add a custom FAQ in AI Training so the AI nails the answer every time
- Update the policy the question is about
Hourly Activity
When your customers actually message — peaks usually align with traffic, but not always. A late-evening spike often means people browsing on phones in bed. Make sure your AI is on autopilot during those windows.
Sentiment Breakdown
Positive / Neutral / Negative split across all incoming messages. A creeping share of negative sentiment is the earliest signal of an operational issue — sometimes hours before it shows up in returns or chargebacks.
Open Analytics at the start of each week, switch to "This Week" view and look for any metric that's moved more than 10% from the previous week. That's where to focus your attention.