Data Cloud Calculated Insights and Segmentation
Turn unified profiles into audiences people can use. We define metrics as calculated insights, build segments on them and test that activations arrive correctly.
A unified profile is only useful once it answers questions such as "who has spent most in the last year?" or "which customers have an open case and a renewal next month?". In Data Cloud, those answers come from calculated insights, which compute metrics across your data model, and segments, which use profile attributes and insights to define audiences for activation.
GeekMidst designs and builds both. The aim is a small, well-defined set of metrics that marketing, service and analytics teams agree on, and segments that refresh on a sensible schedule and arrive in the destination with the counts and fields you expect.
Where teams usually struggle
- The same metric (lifetime value, last purchase date, engagement score) is defined differently in every segment.
- Insight results do not match CRM or finance reports, and nobody can explain why.
- Segments are built directly on raw events, so they are slow and hard to maintain.
- Activated audiences in Marketing Cloud or ad platforms do not match the segment count in Data Cloud.
- Refresh schedules are set by default rather than by how fresh the audience needs to be.
What we build
Metric catalogue
Agreed definitions for metrics like spend, order count, recency and engagement, with owners and the source fields behind them.
→Calculated insights
SQL or builder-based insights at the unified individual level, with dimensions and measures chosen for reuse across segments.
→Segment design
Reusable segment patterns, nested logic and exclusions such as consent, suppression and recent-contact rules.
→Activation set-up
Activation targets, attributes passed with each member and checks in the destination system.
→Troubleshooting
Diagnosis of empty segments, mismatched counts and failed activations in existing setups.
→An illustrative calculated insight
The example below shows the shape of a customer spend insight at unified individual level. It assumes a custom Orders__dlm object with customer_id__c, order_id__c and order_total__c fields mapped from an order source, where customer_id__c holds the same ID as the source Individual record that identity resolution links. Unified individual and identity link object and field API names depend on your org and ruleset, so check them in Data Explorer before adapting this. It is a sketch, not production code.
SELECT
UnifiedIndividual__dlm.ssot__Id__c AS customer_id__c,
COUNT(Orders__dlm.order_id__c) AS order_count__c,
SUM(Orders__dlm.order_total__c) AS lifetime_spend__c
FROM Orders__dlm
JOIN IndividualIdentityLink__dlm
ON Orders__dlm.customer_id__c = IndividualIdentityLink__dlm.SourceRecordId__c
JOIN UnifiedIndividual__dlm
ON IndividualIdentityLink__dlm.UnifiedRecordId__c = UnifiedIndividual__dlm.ssot__Id__c
GROUP BY UnifiedIndividual__dlm.ssot__Id__cHow we deliver it
- 01
Agree the questions
Start from the audiences and decisions teams need, then list the metrics they depend on. Most programmes need fewer insights than first expected.
- 02
Define each metric
Write the definition in plain words, including grain, time window, currency, returns and cancellations, and the source of truth to reconcile against.
- 03
Build and reconcile
Create insights, run them and compare results with CRM or finance figures for a sample of customers. Differences are explained or fixed before use.
- 04
Design segments
Build segments on insights and profile attributes, with consent and suppression exclusions applied consistently.
- 05
Plan refresh and activation
Choose refresh frequency per segment based on how fresh the audience needs to be and the consumption it creates, then configure activation targets.
- 06
Test end to end
Confirm member counts and attribute values in the destination, and document what each segment is for and who owns it.
What you receive
- A metric catalogue with definitions, owners and reconciliation notes.
- Calculated insights and segments configured in your org, with naming conventions.
- An activation map showing which segments go where, with which attributes.
- A short run book for common issues: empty segments, stale insights and count mismatches.
Dependencies and risks
This work assumes identity resolution is in a reasonable state. If unified profiles are wrong, every insight built on them will be wrong too, so we may recommend an identity resolution review first. Segment refresh and insight processing both contribute to consumption, and processing schedules and limits vary, so we check current limits for your org in Salesforce Help during design rather than assume them.
Questions we hear
Should a metric be a calculated insight or a segment filter?
Why do activated counts differ from the segment count?
Can marketers build segments themselves?
How often should segments refresh?
Need audiences you can trust?
Tell us which audiences you are trying to build and where they need to go, and we will suggest how to structure the insights behind them.
