What gets counted automatically
- time to first reply on every enquiry
- whether price was quoted, and on which message by count
- the number of touches from the rep's side
- whether a next step with a date was agreed
- whether bank details went out immediately after agreement
- thread length and message count
- the language the rep replied in
What gives the most
The share of conversations where price was quoted, and the message it was quoted on. Two numbers almost nobody counts, and the fastest ones to put right.
Measurements across projects showed price coming up in roughly half of conversations, on average at the eighth to eleventh message. Once that became a rule and a metric, both numbers moved within a month.
Where the machine gets it wrong
- calls a question unanswered when the answer was a voice note
- misses the context when the customer asked to be called back later
- counts technical messages as touches
- does not understand irony or a polite refusal
So every finding gets reread by a person before it becomes a complaint about a rep. Without that the read loses the team's trust quickly and people stop reacting to it.
How to set it up
- Connect the sources: the phone system and the messaging channels.
- Define the list of metrics that get counted.
- Produce a short daily summary and a weekly review.
- Introduce one rule: a finding gets verified before it becomes a task.