How every vendor performs on the two axes that decide the category: automation rate and conversational competence. Toggle between pre-sale and post-sale to switch the job, then read the full ranking below, scored by a composite of automation, answer quality, and speed.
Up and to the right is better on both axes.
| # | Vendor | Composite | Automation | Quality | Mean | N |
|---|
Blind LLM-judge quality (/100) for each conversation type, Gorgias against the ranked field. Baked from judged transcripts in the ranking window.
Gorgias leads every Shopping Assistant intent by 15 to 19.2 points, strongest on budget compare.
Gorgias leads every Support Agent intent by 3.8 to 24.9 points, strongest on order tracking.
The field drops 28 points from its easiest conversation to its hardest. Gorgias drops 17.
One point per week, pooled across that week's captures. A week with fewer than 10 conversations stays blank, so a thin sample can't draw a spike. Gorgias and up to three providers you pick are in color; the rest of the field is grey. Ignores the date window (shows every week).
Best vendor per vertical in this lane, over the selected date window.
Every conversation follows the same arc. T1 is a single direct question. T2 to T7 push discovery and multi-turn reasoning. T8 to T10 force the transactional close: cart, shipping, and policy. Each curve shows where an agent holds up, and where it breaks down.
No cleanly-timed turns for this vendor in the current window.
Shopping-discovery conversations that must end in a real product recommendation.
Shopping-only lower-bound count from transcript evidence: product cards, product links, and named product recommendations. Generic categories do not count; card-only UI that is not serialized into the transcript can be undercounted.
| Vendor | Conversations | Avg products / conv | Bar | Conv with products | Max |
|---|
Beyond answering: does it use quick replies (suggested chips), render product cards ("pull cards" with price/links), surface reviews, and complete the sale itself? ● yes · ● no · ● not yet tested
| # | Vendor | Quick replies | Product cards | Reviews | Completes sale (no handoff) | Score |
|---|
The full test inventory: which live storefronts each vendor is benchmarked on, the widget technology driving each, and where measurement stands per lane. Ignores the date window (shows the whole program). ✓ measured · ◌ engaged, not enough timed answers · ✕ no answer (dead/gated widget) · … awaiting first capture
| Vendor / store | Website | Widget tech | Locale | 🛍️ Shopping | 🎧 Support |
|---|