Revenue leaks after the first message.

Across the measured price-answer conversations, that share contained a price but no recorded invitation to book, choose a time or take another next step. This is a process gap visible in the captured thread — not proof that every person left, would have purchased, or had no later contact elsewhere.

• 2–19%were redirected from chat to a phone call; the captured thread does not show whether they later called
• 6–16%show an in-chat booking but no recorded downstream outcome in the source data
• Measuredfrom price-answer conversations in the pre-Llucky clinic inboxes

Ads with a similar cost per conversation can produce very different business outcomes. When we linked each paid conversation to the outcome recorded by the business, cost per marked booking differed by five to thirty-one times inside the same account. This is an allocation gap we can diagnose — not a saving we claim to have produced already.

• 8–22%of message-optimised campaign spend went to ads with zero business-marked bookings
• 9–41%of paid conversations were classified as cold clicks across the three accounts
• 1 in 1,092across 1,092 conversations classified as cold clicks, we observed one objectively recorded contact event
• One case: 61%of one clinic's $944 pre-period spend led into an inbox classified as dead — not a pooled benchmark
• $1.25–2.08measured cost of one paid conversation across the three accounts

That share of messages comes in at night and on weekends — at every business in the sample. Later production cases show that this demand can be commercially relevant: 35% of recorded bookings in one clinic and 50% of HOT enquiries in another began outside working hours.

• One clinic: 35%of recorded bookings began with a message outside working hours
• Another clinic: 50%of enquiries later classified as HOT began outside working hours

Across the pre-Llucky clinic inboxes, average time to a first substantive answer ranged from 2.6 to 18.2 hours, while the per-clinic medians ranged from 11 minutes to 5.4 hours. In the measured post-Llucky comparison windows on the same inboxes, the median was 12–16 seconds. That is a scoped observation from those cases, not a universal speed promise or proof that faster replies caused a commercial outcome.

• 3–29%received no substantive answer after automatic templates were excluded — a per-clinic range, not a pooled rate
• 11–42 hwas the p90 range across clinics — the slowest tenth of answers
• 11 min – 5.4 hwas the measured range of per-clinic medians before Llucky, among enquiries with a substantive reply
• 5–8%asked for a price and no numeric price appeared later in the captured session under the audit rules

That many apparent first replies in the conversations we reviewed were automatic templates. A speed report can therefore look fast while the customer's actual question is still unanswered. Our response-time calculation excludes those templates and starts at the first substantive answer.

• 5 / 5measured clinics were running at least one automatic template
• Excludedautomatic templates are not counted as a substantive answer in this report