Skip to content
flowss
Start freeChecking your session
Causal DAG templates

Recommender Exposure and Purchase (Causal DAG)

Why naive click-through attribution overstates a recommender: item popularity drives both what gets shown and what gets bought.

Causal DAGAI & PipelinesCausal DAG syntax docs
Template previewCausal DAG
Recommender exposure and purchase - popularity confoundingRItem shown on the home railBPurchase within 7 daysPItem popularityADJUST FORUUser propensity to buyADJUST FORCClick through to the product pageITime on the product pageSSeasonal demand shockAPrice and promotionExposureOutcomeConfounderMediatorUnobservedIdentifiable — adjust for `Item popularity` + `User propensity to buy`.Adjust for `Item popularity` + `User propensity to buy` — a minimal sufficient set. Conditioning on those blocks every backdoor path from `Item shown on the home …

Make it your own.

title "Recommender exposure and purchase - popularity confounding"
exposure R "Item shown on the home rail"
outcome B "Purchase within 7 days"
confounder P "Item popularity"
confounder U "User propensity to buy"
mediator C "Click through to the product page"
mediator I "Time on the product page"
unobserved S "Seasonal demand shock"
node A "Price and promotion"
R -> C
C -> I
I -> B
R -> B
P -> R
P -> B
U -> R
U -> B
A -> C
A -> B
S --- P
S --- B