OBLYTECH Logo
OBLYTECH Logo

ServiceNow / Practical guide

ServiceNow CMDB Reconciliation Rules for Duplicate Control

ServiceNow CMDB reconciliation rules decide how incoming information identifies an existing configuration item, which source can update each attribute, and what happens when two sources disagree. Without that design, a CMDB can accumulate duplicate records or lose useful values through unplanned overwrites.

For CMDB owners investigating duplicate or conflicting configuration item records..

ServiceNow CMDB reconciliation rulesServiceNow duplicate CIsCMDB source authority

Guide map

Where duplicate CMDB records come from

Define what makes a configuration item the same item

Two records describe the same configuration item only when their identifying attributes match on purpose, such as serial number or host name, not by coincidence of a shared label.

Give each source a clear authority boundary

Decide in advance which discovery source wins when two feeds disagree about a host name or serial number, and write that rule into the reconciliation settings.

Set precedence for conflicting attributes

When two sources disagree about a field, the record should follow a written rule for which source wins, and that rule should be easy to find.

The practical framework

Design ServiceNow CMDB reconciliation rules around identity, source authority, precedence, exceptions, duplicates, ownership, and review evidence.

The core task is not to make every source equal. It is to define identity, authority, precedence, exception handling, and review evidence in terms that the platform and the operating teams can apply consistently.

Define what makes a configuration item the same item

Start with identity. Decide which attributes, or combination of attributes, distinguish one item from another in each class. The answer may differ between a server, application, business service, or other configuration item type. A value that looks unique in a spreadsheet may not stay unique across environments or sources.

Document the identity rule with examples of valid matches, missing values, changed values, and conflicting values. Include the response when identity cannot be established. Creating a new record by default may be safer than merging two items incorrectly, but it also creates a duplicate-control obligation.

The ServiceNow CMDB data quality route is the relevant internal destination when identity rules need to become repeatable process controls.

Give each source a clear authority boundary

List the sources that contribute data and the attributes each source is expected to own. One source may be authoritative for discovery details while another maintains a business owner or service relationship. Authority should be assigned by field or data domain when a single source cannot maintain the whole record.

A source authority decision needs an owner, a failure path, and a review point. If the source stops updating, the CMDB should not silently present old data as current. If a team changes the source, someone must know which reconciliation assumptions need to be revisited.

Set precedence for conflicting attributes

Conflicts are normal when multiple systems describe the same item. Define which source wins for each attribute, what evidence supports that choice, and whether the update should be accepted automatically or sent for review. Precedence without an exception path can turn a simple rule into a hidden data-loss mechanism.

The ServiceNow integration strategy route provides an internal path when reconciliation depends on source mapping, data movement, or integration boundaries.

Do not use a global “latest update wins” rule as a substitute for authority. The latest value may be a correction, a partial record, or a value from a source that was never intended to own the field.

Handle exceptions instead of hiding them

Create explicit outcomes for missing identity, duplicate suspicion, conflicting ownership, stale source data, unexpected class, and unsupported relationship. Record the reason, owner, next action, and evidence. Exceptions that disappear into a log cannot improve the model.

Use a small set of categories that teams can understand and act on. The purpose is not to create a perfect taxonomy; it is to make the boundary between automatic reconciliation and human review visible.

Review duplicate risk as an operating process

Reconciliation is not finished when the rule is configured. Review duplicate candidates, rejected matches, stale records, and unusual update patterns. Check whether the rules behave differently after a source schema, discovery scope, integration, or class definition changes.

The ServiceNow continuous improvement route is the appropriate internal destination when the rules need recurring review rather than a one-time correction.

Common questions

Questions teams ask about reconciliation rules

Can one source own every CMDB field?

Only if it can maintain every field accurately and continuously. In many environments, different sources are better suited to different data domains. Explicit field ownership is easier to review than an assumption that one source is always authoritative.

What should happen to a suspected duplicate?

Preserve the evidence, stop automatic merging until identity is clear, assign an owner, and decide whether to merge, retire, correct, or retain the records. The decision should be traceable to the identity rule and the service impact.

The research-grounded boundary

Reconciliation rules hold up when a later reviewer can see why two records were matched, which source was allowed to change each field, and who approved any exception. Test the rules against real duplicate and conflict cases before relying on them, and keep the decision record beside the rule so the reasoning survives staff changes.

What the reconciliation engine decides on every incoming record

When data from discovery, an integration or a manual import reaches the CMDB, the platform has to decide whether that data describes an existing configuration item or a new one. The identification and reconciliation logic works through identifier attributes in a defined order, and the result is either an update to a matched record or the creation of a new one. The person who owns the CMDB needs to understand that order, because most duplicates come from identifiers that were never specific enough to separate similar items, not from the platform making a random choice.

A useful starting exercise is to take one class, such as a virtual server, and list the identifier attributes each source supplies for it. Serial numbers may be blank on virtual machines. Host names may be reused after a rebuild. IP addresses change with DHCP leases. Each attribute has a failure mode, and the rule set should be designed around those failure modes rather than around the attributes that happen to be populated.

Source authority, field by field

A reconciliation design answers one question for each attribute: which source is allowed to write it. The table below shows the kind of decision a CMDB owner should make for a few common attributes. The sources named are typical, and the actual answers depend on the tools in your environment.

AttributeLikely authoritative sourceSource that should not overwrite itWhat to do on conflict
Operating system versionDiscovery from the deviceA manually entered inventory spreadsheetAccept the discovered value and log the spreadsheet difference for the asset team
Business ownerThe service ownership registerDiscovery, which can see a login but not the accountable personHold the discovered value as a suggestion and route it to the service owner
Support groupThe service catalog or support modelAny integration that copies a group from a ticketing tool without reviewKeep the existing value, record the proposed change, and review it
Installation statusThe change or deployment record that created itA scan that happens to run during a rebuildDefer the update until the change is marked complete

Two principles sit behind the table. A source should own only the attributes it can maintain accurately and continuously. And a conflict between sources should be resolved by the rule for that attribute, not by which update arrived last. The CMDB ownership model post deals with who is accountable for a data domain, what quality means and which decision follows when a record is wrong.

Building the duplicate control sequence

Duplicate control is a process as much as a rule. The sequence below is one practical approach, and each step leaves a record that can be reviewed later.

  1. List the classes in scope and name an owner for each one who can judge whether two records describe the same thing.
  2. Write the identifier rule for each class, including what the rule does when an identifier is blank or ambiguous.
  3. Decide whether an ambiguous match creates a new record, holds the update, or sends it to review. Creating a new record is usually safer than a wrong merge, but it adds a duplicate to clean up later.
  4. Define the duplicate candidate report and the person who works through it, with a standing time for the work.
  5. Test the rules with real data from a copy of the environment, including records that are known to be the same and records that are known to be different.
  6. Switch the rules on for one class at a time and watch the candidate queue before moving to the next.

Failure modes that appear after the rules are live

Several problems show up only after the rules run on live data. Over-broad identifiers match records that should stay separate, typically when a host name or a generic model value is shared by many items. Under-specific identifiers leave duplicates in place, especially when one source records the fully qualified name and another records the short name. Lower-authority sources can update fields they were never meant to own, because the rule for that field was never written. And a class change, such as a new subtype introduced by an application team, can alter which identifiers are used without anyone noticing.

Each of these produces a symptom that looks like a data quality problem. A reconciliation review should ask first whether the rule is doing what it was designed to do.

Measures that show the control is holding

A duplicate control is working when the queue of suspected duplicates is worked and shrinking, when the number of manual merges falls because the rules are right, and when the age of the oldest unreviewed candidate stays within a limit the CMDB owner has set. A single count of duplicates removed says little, because it can rise when the rules are finally catching real problems. Look at the trend and at the reasons the candidates were rejected, since repeated rejection of the same pattern points to a rule that needs changing.

When rules are the wrong tool

Reconciliation rules cannot repair a source that is wrong from the start. If a discovery tool scans the wrong network range or an integration populates the same serial number on every record, the rules will faithfully reconcile bad data. In that situation the fix belongs upstream, in the source configuration or the import mapping. The CMDB data quality rules post covers which data quality rules are worth operating. Similarly, a class that has no owner and no agreed purpose should not be reconciled at all until someone is accountable for it. Turning on rules for an unowned class simply moves the disagreement into the update history.

Questions CMDB owners ask about reconciliation

Should the rules prefer the most recent update?

Not as a general policy. Recency can reflect a correction, but it can equally reflect a partial record or an integration that was never meant to write that field. Use recency only for attributes where it has been tested and agreed.

How often should the identifier rules be reviewed?

Review them when a source changes, when discovery scope changes, when a class is redefined and after any incident where a record was wrongly merged or wrongly separated. A fixed calendar review is useful as a backstop, but the triggers matter more.

Can a merged record be split again?

Sometimes, but not cleanly. Before any merge goes live, record the source identifiers and the fields each source contributed, so a later split can be done from the evidence rather than from memory.

Who decides between two sources that both claim a field?

The owner of the field, named in the CMDB ownership model. If that owner is not clear, the disagreement is a governance gap and should be escalated, not settled by the engine.

Next step

When duplicate records are interfering with impact analysis or audit work, the CMDB reconciliation and duplicate control page is the relevant route.

A practical next step is to test your current rules against a sample of known duplicates from your own CMDB before changing them. Talk through your reconciliation design with OBLYTECH.

A practical next step

Need a clearer ServiceNow operating model?

OBLYTECH can discuss the questions raised by this guide and the next review step. See the ServiceNow consulting services page for the commercial route.

Useful starting pointBring the current question, scope, ownership concerns, and evidence needs into the first review conversation.

ServiceNow guides on this topic

Guides from the Oblytech ServiceNow library that explain the practice behind this page.

More from the blog