Please do not share this outside of the company without first talking to @Garrett Roach

1. A properly installed tracking tag on the stores website for a duration of time prior to attempting sales attribution
2. A sales report from the store containing at least: Home address, email address(s), phone number(s), and date of sale for each sale

A user uploads the sales report in the platform and maps the required fields for ingestion. In the future, we will ingest sales data daily from the stores CRM or DMS, these integrations are ongoing.

The matching process uses deterministic and probabilistic techniques to associate one or many website sessions with the offline sales data. The matching process uses a standard 90 day look back window in all of its website data queries.

1. Use email address and phone numbers, provided in the sales report, to match against tracked form submissions recorded from the website
2. Use email address and phone numbers, provided in the sales report, to match against anonymously identified users recorded on the website

1. Use address data, provided in the sales report, to match against location events recorded on the website.

Once a match is found for a given sales record, that match can be “enriched” using recorded data from that session. This process also has a combination of probabilistic and deterministic methods.

1. Finding other sessions from the same user, on the same device by matching unique user identifiers stored in first party cookies

1. Finding other sessions from the same user, by matching against that users IP address, provided the IP address is a broadband address and not a cellular address

At the completion of the matching and enrichment processes, the result is a sales record matched to one or many sessions on a given website. At this point, summary data can be provided for those sessions (campaign history, events, etc).

The probabilistic features of the methodology have been back-tested against the deterministic features and have shown an accuracy rate of 90% for the current model.

We are continuously working on the model to optimize accuracy and match rate.