In creating this package, which is meant for a wide range of use cases, we had to take opinionated stances on a few different questions we came across during development. We've consolidated significant choices we made here, and will continue to update as the package evolves.
We chose to build off of the layer of attribution occurring in the Klaviyo-only dbt package. This means that we refer to last_touch_campaign_id and last_touch_flow_id (a coalescing of the Klaviyo-provided campaign_id/flow_id and those attributed by the package itself) when attributing Shopify orders to Klaviyo campaigns and flows.
This package performs last-touch attribution only, primarily in line with Klaviyo's internal attribution model. It would've been fairly straightforward to also perform first-touch attribution here, but given the nature of email marketing (ie its place at the bottom of the funnel), this didn't seem very valuable. First-touch (and multi-touch) attribution would make more sense with ad sources and other kinds of marketing channels.
Note that the package's attribution model diverges from Klaviyo's own slightly. In Klaviyo, open/click email events do trigger the attribution window to open, but the timeframe of the window actually starts when the customer received the email. We decided that starting the timeframe from the open/click event would be more appropriate. Thus, you may see more generous attribution metrics from the package than from the Klaviyo UI.
Theoretically, emails should be unique within shopify.customers and klaviyo.person. However, we have seen cases where this is not true, namely when a bot has created a bunch of accounts under the same email and/or name. Thus, because this package relies on email to join across platforms, we have rolled up these tables to the email-grain. We have also incorporated warnings on the shopify__customers and klaviyo__persons models so that you are notified if an email is not unique (and that you may have a bot problem!)
In the daily_customer_customer_campaign_flow model, refund metrics are tied to the day the order was placed, not when it was refunded.