There is no defensible universal split — it is a function of motion, ACV and market maturity. The useful version of the question is what each source costs and converts at, measured on cohorts, so the mix becomes a capital-allocation decision rather than a target imported from a benchmark.
Before comparing to any published mix, check the denominator and the panel: whether partner and event-sourced pipeline is counted as inbound varies by company, which alone can move the split by twenty points. Most benchmark disagreements in this area are definitional.
Every claim above comes from work LeanScale published. These are the sources.