Background: B2C side of the org, mostly first-party data plus some loyalty program imports. Match rate on deterministic rules (email + phone) is ~62% and we know we are leaving consolidations on the table.
Before I turn on probabilistic blending I want to make sure I'm not going to collapse household members onto a single profile. Has anyone run a side-by-side and trusted the output enough to ship it?
We ran the same comparison last quarter. Probabilistic was the right call for us but we capped the confidence threshold high — 0.95+ only. Caught the household collapse problem you're worried about by adding a 'shared-address-but-different-DOB' exclusion to the ruleset.
Worth running it in shadow mode for a week before flipping the switch — Data Cloud will compute the probabilistic match without committing it, and you can audit the proposed merges against your existing graph.