bayesnet_attribution We have 3 channels that drive a conversion (C1, C2, C3) all drive Conv C1 also drives (C2, C3) We can estimate most conditional probabilites like P(conv|C1, C2, C3) etc. We populate the BN and then get the channel effect as $P(conv|C1) - P(conv|\neg C1)$ 🦇