Start with one use case and one question
Pick a concrete outcome: reduce churn by identifying at-risk customers, personalize outreach using purchase history, or give service agents a complete view of each customer. Write down the question the data must answer and the decision it will inform. That defines which sources you need and which you can leave for later. A narrow first use case produces results in weeks and gives you a template for the next.
- Define the business question and the action that follows from the answer.
- Identify the minimum sources needed to answer it.
- Agree how you will measure whether it worked.
Connect sources and map them to a common model
Data arrives from different systems in different shapes. Ingest the sources you need, then map their fields to a common data model so that, for example, an email address from the CRM and one from the e-commerce system are understood as the same kind of thing. Mapping is detailed work that rewards care: errors here carry through to everything built on top. Document each mapping and who is responsible for each source.
Identity resolution: deciding who is the same person
A unified profile depends on deciding when records from different systems describe the same individual. Rules might match on email, phone, or loyalty number, with different levels of confidence. Start with strict rules that avoid wrongly merging people, then loosen as you review results. Monitor the outcomes: too many duplicates mean the rules are too tight, and incorrectly merged customers mean they are too loose.
Explore Salesforce solutions →Segments, insights and activation
With unified profiles you can build segments, such as customers who have not purchased in ninety days but opened recent emails, and calculate insights like lifetime value or engagement score. The value comes from activation: sending segments to marketing, service, or sales tools where they drive action. Plan the activation path early, because a segment nobody uses is only a report.
Governance, consent and cost awareness
Combining data raises obligations. Honor consent and preferences, restrict access to sensitive fields, and keep a record of where each piece of data came from. Involve your privacy officer early. Also keep an eye on cost: consumption-based pricing means that ingesting and processing everything has a price. Confirm the current pricing model with Salesforce and let the use case justify the data you bring in.
Implementation and integration services →Decisions to settle before you start
- First use case and owner. Name a business owner who cares about the outcome and will use the result.
- Matching rules. Agree which identifiers decide that two records are one person.
- Consent model. Define how consent travels with the data and who enforces it.
- Activation destination. Decide where segments will be used and who acts on them.
A realistic first 90 days
- Days 1 to 30. Select the use case, assess the quality of the sources, and design the data model mapping and consent rules.
- Days 31 to 60. Ingest the first sources, build identity resolution, and review matches with business users.
- Days 61 to 90. Create the first segments and insights, activate them in one channel, and measure the result.
Pitfalls to avoid
- Connecting everything first. More data does not mean more value. Bring in only what the use case needs.
- Ignoring data quality. Unified bad data is still bad data. Clean sources or build rules to handle them.
- Over-merging identities. Wrongly combining customers damages trust. Start strict and review.
- No path to action. Without activation, unified profiles are an expensive dashboard.
