Tool
The turn where the mask comes off
A support conversation drawn as a strip of turns. Underneath, the cost of only saying at that turn that there is an agent on the other side. Move the marker and see where it jumps.
Turn 9, customer: The last delivery arrived broken and I lost a day of work waiting for it
They have told something with emotional weight to what they believed was a person. Second jump.
What the agent would say here: I am sorry about the delivery. I should say I am an automated assistant, and a person on the team will see this.
The model, in the open
The curve is a relative scale from 0 to 100, built from what we see in our implementations: it says where the jumps are, not what a jump costs in money. The two jumps are the whole point of the tool, and they land right after the customer hands over personal details and right after they tell a problem. It is an illustrative model, not the measurement of any customer.
How to read it
- The curve is flat for the first six turns. Saying it there costs nothing, and it is exactly where almost nobody says it.
- The first jump is not where you would put it by eye: it is not when the case gets complicated, it is when the customer hands their ID and address to something they believed was a person.
- The second jump is emotional rather than about data. Once a problem has been told, finding out rewrites the whole conversation backwards.