Data unification in Dynamics 365 Customer Insights
Table of Contents
Introduction
Customer records rarely live in one place. The same person appears in your CRM, eCommerce platform, and loyalty program under slightly different names, emails, and phone numbers, making it harder to keep customer data aligned.
Left unresolved, those duplicates split one customer into several profiles, which skews segments, reporting, and personalization. Customer Insights Data consolidates them into a single profile, enabling downstream analytics and marketing to rely on it.
This guide walks through the current unification workflow, from selecting customer data and deduplication to matching rules and the unified data view. It also covers normalization, fuzzy matching, exceptions, and best practices for accurate results so you can reduce duplicate records and protect downstream analytics and marketing.
What is Data Unification in Dynamics 365 Customer Insights?
Data unification combines customer data from several sources into a single master dataset. Customer Insights – Data removes duplicate rows, matches records across tables, and assigns each unified profile a unique Customer ID.
Microsoft renamed parts of the product, so older articles use outdated terms. Dynamics 365 Customer Insights is now Customer Insights – Data, and entities are now called tables. The old Map, Match, Merge steps map to the current workflow like this:
| Older term | Current step in Customer Insights - Data |
|---|---|
|
Map (select entities) |
Customer data |
|
Not covered in older guides |
Deduplication rules |
|
Match |
Matching rules |
|
Merge |
Unified data view |
What Do You Need Before You Unify Data?
Ingest your customer data sources first, and confirm each one refreshed successfully. Include only customer profile tables, such as contacts, loyalty members, and eCommerce accounts.
Exclude activity data such as purchases or invoices, which have a one-to-many relationship with the customer.
- Review ingested tables and remove unwanted rows before you start.
- Identify the table with the most complete, reliable customer data. It becomes your primary table.
- Identify fields likely to be unique to each customer, such as email, phone, or a name-and-address combination.
- Check Data > Data sources for failed or stale refreshes. A failed refresh is a common cause of the Match failed error.
You may also like: Real-Time Data with Dynamics 365 Customer Insights
How Do You Select Customer Data for Unification?
In the Customer data step, choose the tables and columns to include, confirm the primary key for each table, and map common fields such as email and address. Leave out any column you don’t need for matching or for the final profile.
Customer Insights suggests field mappings automatically, so review them rather than accepting them blindly. Columns with a purpose unique to your business, such as an internal ID, can stay unmapped.
Example: A retailer unifies three tables: CRM Contacts, eCommerce Contacts, and Loyalty Customers. Its invoice tables are activity data. The team adds them as activities after unification instead of unifying them.
Not Sure Your Unification Rules Are Working?
Get a review of your match order, rules, and unique record counts, with clear recommendations to improve match accuracy.
Request a ConsultationHow Does Deduplication Work Before Matching?
Deduplication identifies duplicate rows for the same customer in a single table and selects a single row to represent that customer. Each table is deduplicated separately, before any cross-table matching begins.
Every deduplication rule runs against every row, and rules that share a record chain together. Suppose Rule 1 (name and phone) matches records 1 and 2, and Rule 2 (name and email) matches records 2 and 3. Records 1, 2, and 3 become one group.
Avoid matching on name alone, as it can group different people who share a name.
Choose the winner row with merge preferences:
- Most filled: Keeps the row with the most populated columns. This is the default.
- Most recent: Keeps the newest row. It requires a date or numeric field.
- Least recent: Keeps the oldest row. It requires a date or numeric field.
- Advanced: Sets a policy per column, such as the most recent email with the most complete address.
The system keeps the non-winner rows as alternate rows. During matching, an old phone number on an alternate row can still link the winner to a record in another table.
How Do Matching Rules Work Across Tables?
Matching rules connect records between tables. You set a match order that starts with the primary table, then define rules for each table pair. Each rule holds one or more conditions, and every condition must be true for records to match.
- Primary table: The basis of your unified profiles. Choose the most complete table that shares the most columns with others. A table can only match against tables ranked above it.
- Include all records: Keeps unmatched records (singletons) as unique customers.
- Rule order: Rules act as OR conditions. The system tries rule 1, then moves to rule 2 only if rule 1 finds no match.
- Column choice: Match on columns unique to a customer, such as email or phone. Avoid activity columns such as a purchase ID.
- Example: With CRM Contacts as the primary table, match eCommerce Contacts using two rules. Rule 1 matches Email exactly. Rule 2 matches Phone exactly using Type – Phone normalization, plus Last name using Type – Name normalization at High (80%) precision.
How Do Normalization, Fuzzy Matching, And Precision Work?
Normalization standardizes values for comparison without changing the stored data. Precision sets how close two values must be to match: Exact (100%), High (80%), Medium (60%), Low (30%), or a custom percentage. Any setting below Exact turns on fuzzy matching.
The most useful normalization options for enterprise data:
| Option | What it does |
|---|---|
|
Type – Phone |
Standardizes formats, so +01 425.555.1212 matches 1 (425) 555-1212. A number with a country code won’t match one without. |
|
Type – Name |
Resolves over 500 common name variations and titles, such as debby to deborah. |
|
Type – Address |
Standardizes street, northwest, and similar terms. |
|
Type – Organization |
Removes noise words such as co, corp, and ltd. |
|
Text to lower case, Symbols, Whitespace, Unicode to ASCII |
Cleans formatting differences. |
|
Alias mapping |
Treats custom string pairs as identical, such as MSFT and Microsoft. |
|
Custom bypass |
Ignores values that should never match, such as a dummy email. |
Fuzzy matching measures edit distance, the number of added, deleted, or changed characters between two strings. The score equals base length minus edit distance, divided by base length.
For example, [email protected] and [email protected] differ by one character, which scores about 95% and passes a High (80%) threshold.
Use fuzzy matching for typos, not name variants. Handle Suzzie versus Suzanne with Type – Name normalization or alias mapping. Fuzzy matching applies only to string columns and runs slower than exact matching, so test whether the extra matches justify the extra run time.
How Do Exceptions and Custom Matching Handle Edge Cases?
Exceptions stop a rule from matching records that meet specific criteria, which corrects false positives. Custom matching overrides default logic with lists you upload, so you can force or block individual matches.
For example, a rule on last name, city, and date of birth would merge twins. Add an exception that blocks the match when first names differ.
Custom match supports four options:
| Option | Purpose | Example |
|---|---|---|
|
Always match |
Forces two primary keys to match |
Match key 12345 to key 54321 |
|
Never match |
Blocks two primary keys from matching |
Never match key 12345 to key 54321 |
|
Bypass |
Ignores values during matching |
Ignore 11111 and Unknown |
|
Alias mapping |
Treats values as equal |
Joe equals Joseph |
To apply them, download the template for your option, fill in table names and key values, and ingest the file as a new data source. Table names are case sensitive, and the file name can’t contain spaces. Then select the template tables in the Custom pane.
For Bypass and Alias mapping, also select Custom bypass or Alias mapping in the rule’s normalization list. Custom matching on a table requires at least one deduplication rule for that table.
Get One Customer View Across Every System
AlphaBOLD's Dynamics 365 team designs deduplication and matching rules, prepares source data, and implements Customer Insights - Data.
Talk To an ExpertHow Do You Configure the Unified Data View (Merge)?
The Unified data view step controls which customer columns appear in the unified profile. You can keep columns separate, combine them into one, rename them, reorder them, or exclude them. Combined columns pick a winner value by Importance (the default), Most recent, or Least recent.
Combine related fields into a group when they must stay together. Without a group, the system selects each field’s winner separately and skips null values.
A null second address line in your top-ranked source can pull Suite 950 from another source, producing an address that exists nowhere. Grouping the address fields prevents that mix.
This step also controls three further settings:
- Clusters: Group related profiles, such as a household. The system creates a household automatically when the Customer table includes Person. Last Name and Location. Address.
- Customer ID: The system generates a unique GUID per profile. Keep the default. If you must configure it, use stable fields such as a government-issued ID, and avoid phone, email, or address.
- Output: The run produces the Customer table of unified customer profiles.
How Do You Review and Validate Unification Results?
Run unification, then check the rule statistics on the Deduplication rules and Matching rules pages. If a new rule leaves the Unique records count unchanged, an earlier rule already found those matches. Remove rules that add no value.
- Go to Settings > System > Status and select Match to see how long deduplication and matching took per run.
- Open the Tables page to inspect system-generated output tables for each step.
- Use the Deduplication output tables to see why records were grouped. They include the matched rule and the returned score.
- If a run fails, refresh the data sources, confirm source changes were ingested, check for missing files, and rerun.
What Are the Best Practices for Data Unification?
More rules don’t mean better results. Every added rule increases runtime and can introduce false matches. Start small and add complexity only when results justify it. These practices keep unification accurate and quick:
- Start with a primary table and a small rule set: Compare each table to your most complete table first. Add rules one at a time and keep only those that improve results.
- Deduplicate every table first: One row per customer gives matching a clean base, and alternate rows still help link records across tables.
- Normalize before you go fuzzy: Standardize phones, names, and addresses so exact matching catches more records before you lower precision.
- Pair every fuzzy condition with an exact one: Exact conditions narrow the comparison set, which keeps fuzzy matching accurate and faster.
- Avoid heavily repeated values: Default form values and dummy emails create false matches. Use data profiling to find them and custom bypasses to ignore them.
You may also like: Dynamics 365 Customer Insights – Journeys Best Practices
Ready To Unify Your Customer Data?
Share your data sources and goals, and AlphaBOLD will outline a unification approach that fits your environment.
Talk To an ExpertConclusion
Data unification in Dynamics 365 Customer Insights – Data turns scattered customer records into one profile per customer. Deduplication cleans each table, matching rules connect them, and the unified data view controls what the final profile contains.
Accuracy comes from disciplined rule design, not rule volume. Start with a primary table, normalize before you use fuzzy matching, and validate results after every change. Done well, unification gives your teams a customer profile they can rely on for segmentation, reporting, and personalization.
FAQs
Explore Recent Blog Posts
