Insights
Every conversation is read, not sampled. These are the patterns that came out of 16,904 of them — each one with the figures it rests on, so you can decide whether you believe it before you act on it.
If every open insight were acted on
Spend the insights say isn't earning
Regenerated nightly from the full transcript set
Sizing is a third of every conversation, and it's one question
Sizing is the stated blocker on 30.7% of replies — the largest single reason by some way. The agent handles it well: those conversations recover at 47.5%, against 20.3% across the account. That is the problem. It is answering the same question 104 times a week because the product page does not.
Conversations about fit
2,697 (30.7%)
Recovery once answered
47.5%
Account recovery rate
20.3%
Median time to a reply
34m
Do this: Put the fit note on the tee template above the size selector, then watch whether the objection's share drops.
Failed payments recover better than anything else we do
A shopper whose card declined recovers at 79.5% — far and away the best outcome on the board, because nothing is wrong except the checkout. They are also the smallest group we message. Every one of these is an order that was already decided.
Recovery rate
79.5%
Account recovery rate
20.3%
Reply rate overall
51.8%
Do this: Trigger these in minutes rather than at the 55-minute mark — a declined card is a live session, not an abandoned one.
The free-shipping gap is smaller than the shipping charge
When shipping cost is the blocker, the median gap to the $75 threshold is $10.01 — less than a sticker pack, and less than the shipping they were quoted. The agent already reaches for that framing and it lands: those carts recover at 40.1% with no discount at all. It only reaches for it when the shopper says the word first.
Shipping-cost objections
1,071 (12.2%)
Median gap to free shipping
$10.01
Recovery when the gap is named
40.1%
Discount released
$0
Do this: Let the agent name the gap proactively when it is under $15, rather than waiting to be asked.
Most escalations are a lookup, not a decision
216 conversations went to a person in the last 30 days. Delivery guarantees and bulk orders are most of them, and only bulk genuinely needs a human — a ships-by date is a lookup the agent could do itself if it could read the fulfilment cutoff.
Escalations, 30 days
216
Share of replies
13.5%
Median time to a human
9m
Do this: Give shippingEstimate the real ships-by cutoff and the delivery-guarantee escalations disappear.
Codes are going to shoppers who never raised price
22.9% of recovered carts carried a code, and on 27.1% of those the shopper never raised price — they raised sizing or shipping and took the code because it was offered. The guardrail already says price only. The model is reading it as permissive.
Recovered carts with a code
22.9%
Of those, no price objection
27.1%
Margin released, 30 days
$2,079
Do this: Tighten the offer step to price objections only, and re-measure the sizing path's recovery rate after two weeks.
A tenth of replies are people telling us they're not buying
10.6% of everyone who answers says some version of "just looking", and those recover at 12.3% — effectively nothing. The agent takes the no and stops, which is right. The cost is not the send, it is the 1.68% of messaged shoppers who ask us to stop entirely: 52 numbers in the last 30 days that no flow can ever reach again.
Share of replies
10.6%
Recovery rate
12.3%
Opt-outs, 30 days
52
Opt-out rate
1.68%
Do this: Find a signal that separates them before the send — time on the checkout page, or whether they got as far as the address.