Assisted conversion
Orders from shoppers who had an attributable agent interaction, compared with a documented baseline or holdout.
Engaged shoppers are self-selecting, so a raw cohort comparison does not prove causation.
Measurement, not a leaderboard
Benchmark ecommerce AI without manufacturing certainty.
Zubby AI does not currently publish a statistically reviewed, cross-customer benchmark dataset. This page documents the measurement framework merchants can use in their own store instead of presenting unsupported Zubby or competitor averages.
Recovered carts
Completed orders linked to a configured recovery touchpoint inside the selected attribution window.
Report the window, channel, discount cost, and duplicate-order handling with the result.
Answer quality
A reviewed sample of factuality, policy compliance, escalation behavior, and merchant feedback.
Automation rate alone can reward confident but incorrect answers.
Operating cost
Subscription, usage top-ups, discounts, provider costs, and human review time divided by attributable value.
Do not compare vendor sticker prices without normalizing included volume and workflow coverage.
Minimum evidence for a publishable claim
- Named population and sample size
- Exact metric and attribution window
- Comparison period or controlled holdout
- Treatment of refunds, discounts, and duplicates
- Median and distribution, not only an average
- Collection date, exclusions, and known limitations
Measure your own baseline first
Review the attribution methodology, define success before enabling a journey, and interpret dashboard data alongside your commerce platform’s order record.