What it does
A deterministic, provided-data marketing-measurement role. It reads only the values in the tool call, reaches no analytics platform, and invents nothing. Attribution first. Last-click, first-click, linear and time-decay assign DIFFERENT credit to the same touchpoints, and none of them measures causation — causation is a claim about what would have happened otherwise, and the data contains only what did. So it runs every convention and reports the SPREAD: a channel credited 31 percent under one rule and 9 under another was never measured at 31. It averages nothing, because a mean across four conventions is a fifth convention with no more justification, and it would hide the only quantity actually measured. A channel whose credit holds under every rule is reported robust, which is worth knowing precisely because it is rare. For retention it fits no curve and projects nothing. It reports, for every period, HOW MANY COHORTS could observe it — because period 12 is visible only to cohorts twelve periods old, so a curve may rest on twenty cohorts at the head and one at the tail while both print at the same weight. The last period every cohort reached is named as the honest end. For funnels it reports what a PERFECT fix at each step could be worth in final conversions, bounded by what follows it: rescuing people into a path that converts at four percent is worth four percent of them. It calls no drop a leak — an email confirmation and a payment page drop people by design, and whether a drop is a fault is a question about intent that counts cannot answer.</description> <parameter name="category_slug">marketing-advertising
Example prompts
- Here are my conversion paths — how much does each channel's credit change depending on the model?
- Last-click says paid search drove 60 percent of revenue. Is that a finding or a convention?
- I have 90 days of retention data. What can I honestly say about 12-month LTV?
