A conversation that simply stops teaches the customer one durable lesson: don't bother next time.
We cite no vendor's survey and publish no performance number of our own, because none has been measured with real customers yet. What follows is what we will report, defined before we have a result to flatter.
Of the conversations that needed a person, how many reached one.
Every vendor reports the share handled without a human. Nobody reports what happened to the rest, and the rest is where a customer decides whether to stay.
Of the promises made to customers, how many reached an ending before their deadline.
“Someone will call you back” is a liability the moment it is said. Counting them is the only way to know whether the company keeps its word.
From the handoff to a specific person holding the conversation.
A queue is not an owner. The gap between “escalated” and “someone has it” is where conversations go quiet.
How often a customer had to say the same thing twice to reach an answer.
The promise on the front page is that they never start over. This is the number that would prove it or break it.
Not results, properties. Each of these is a fact about how the system is built, and each one you can watch working rather than take on trust.
2
endings a conversation may have
Resolved, or a named person holding it. There is no third state to close a record into.
4
checks every outbound contact passes
Consent, calling window, suppression, frequency, in that order, before a single dial.
0
ways to put a release live untested
The gate compares the compiled version against the tested one. Arithmetic, not a convention.
1
record per customer, every channel
An opt-out said once, anywhere, stops every channel at once.
Two rules, and we would rather be held to them than to a benchmark. The first: no number about Roux's performance appears anywhere on this site, because none has been measured with real customers. Where a metric has no value, it says so plainly rather than borrowing a proxy.
The second: we do not cite other vendors' research. Their surveys are written to sell their own product, and quoting a competitor's study to argue against that competitor is a trick, not an argument. What we claim instead are mechanisms, things the product does by construction, that you can watch working in a thirty-minute demo.
When the first cohort produces real numbers, they will appear here with their sample, their period and their caveats, including the ones that are unflattering. A metric that only gets published when it looks good is marketing, not measurement.