Customer Insights Data is positioned as the customer data platform for the Microsoft stack. The product works. The implementations often do not, because teams underestimate the unification step.
The match rule order matters
Match rules in Customer Insights run sequentially, and each rule produces a confidence score. If your first rule is Email exact match and your second is Phone normalized match, a profile that matches on both gets the email score, not the higher of the two. Order rules from highest precision to lowest, and add an explicit Custom ID exact match first if you have a CRM contact ID.
Normalization is non-obvious
The Normalize transform is your friend, but it does not handle international phone formats well. A US number +15551234567 will not match (555) 123-4567 unless you explicitly configure the normalization. Build a Power Query step that strips all non-digits and prefixes the country code before the match phase.
Conflation is irreversible without effort
Two profiles get merged. Three months later you discover one of them was a different customer with a coincident email. The fix requires re-ingesting the source data with a discriminating attribute and rerunning unification. There is no unmerge button. Always ingest a stable source ID and use it as a tiebreaker in match rules.
Activity timeline ingestion limits
Activities (orders, tickets, web events) ingest into the unified profile timeline. The default retention is 12 months. High-volume sources (web clickstream) will exceed this and silently drop older data. If you need longer history, project activities into a custom Dataverse table and surface them via a related-data tab instead.
Segments vs measures
Segments are the customer set (“VIP customers in California”). Measures are the metric (“Average order value last 90 days”). Confusing them is the fastest way to wrong dashboards. A segment used as a measure denominator is almost always a bug.
Activation push to D365 Sales
The reverse-ETL push from CIData into Dynamics 365 Sales is great when it works. The failure mode is a measure that does not exist on the destination contact, silently dropping. Validate every measure exists as a column on the destination table before scheduling the push.
Customer Insights Data -> Activations -> Dynamics 365 Sales
-> Map: VIP_Score (CI Data) -> new_vipscore (Contact, Decimal)What to do this week
Run the “match analysis” report for your top match rule. If more than 5% of matched profiles share an email but have a different last name, your rule is over-conflating. Add a name distance check before approving the merge.
