Part 3 · Operate
Attribution
Which channel produced a conversation, and why one channel never moves another’s number.
01What it is
Two sources: tracked links that count clicks, and a classifier that reads what somebody said when asked how they found you. Each channel keeps its own posterior.
02Why it exists
Attribution across channels is where most tools quietly cheat, by crediting the last touch to whatever is easiest to measure.
Per-channel isolation is the rule that stops it. One channel’s results never move another’s multiplier, so a channel with 90 attempts and one conversation cannot bury a channel with 11 attempts and six.
03How to use it
- 01Read the per-channel table rather than the headline number.
- 02Look at the floor column. A channel below its floor has not earned a conclusion yet.
- 03Ask people how they found you and record the answer. Self-report is the strongest signal here.
04What good looks like
- You give it
- LinkedIn with 11 attempts and six conversations, cold email with 90 and one.
- You get
- LinkedIn leads the table, with the hedge stated because 11 attempts is below the floor for a confident claim.
- Why
- Each channel is ranked on its own posterior, and the hedge is attached to the sentence rather than hidden.
Derived from the rules in the code
05What weak looks like, and what it costs
- You give it
- Reading only the total conversation count.
- You get
- A number that hides which channel produced it.
- Why
- The total is the sum of channels with very different volumes and rates.
Derived from the rules in the code