Skip to content
Optima docs
ConsoleSettings

Identity

Define the rulesets that recognise several profiles as one person, and test them on real profiles.

The Identity screen is where you tell Optima how to recognise that several profiles belong to the same person, for example one created from a web form and one imported from your shop. You write rulesets made of rules, switch one on, and Optima groups matching profiles into an identity cluster without changing the original profiles. Open it under Settings → Workspace → Identity. The tabs are Rulesets, Resolve and Clusters. For the ideas behind it, see Identity.

Who uses it

Integration specialists. Changing rulesets needs the schema_editor role, or another role your administrator has given access to this area — see Licences and roles. Looking up a cluster needs only read access to profiles.

The Identity screen on the Rulesets tab with one active ruleset, its merge policy and row actions

Fields

Rulesets tab

The list shows each ruleset with its entity type, merge policy and status. A filter above it narrows the list to one entity type. New ruleset opens the dialog.

FieldMeaningValid valuesDefault
NameA name for the ruleset.TextEmpty
Entity typeThe kind of profile it applies to. Fixed once created.An existing entityNone
DescriptionA note for colleagues.TextEmpty
Merge policyWhich values the combined view shows when members disagree.Latest wins, Source priority, Root wins, AdditiveLatest wins

After you save a new ruleset you land on its page, where its rules are added: a ruleset without rules matches nothing.

Ruleset page

The page shows the entity type, merge policy and status, and the list of Rules with Priority, Name, Match type, Matchers, Confidence and status. New rule opens the rule dialog, which has a Builder and a JSON (advanced) view.

FieldMeaningValid valuesDefault
NameA name for the rule.TextEmpty
PriorityRules run in priority order, lowest first.A number
Match typeHow certain the match is.Deterministic, ProbabilisticDeterministic
LogicShown for probabilistic rules: how the matchers combine.All matchers must pass, Weighted thresholdAll matchers must pass
Min scoreShown with a weighted threshold: the summed weight of the matchers that fire must reach this.0 to 1
ConfidenceStamped on the links this rule creates.0 to 1
Min match countAt least this many matchers must fire.1 or more
ActiveWhether the rule is used.On or offOn
MatchersOne or more comparisons, added with Add matcher (at least one is required).See below

Each matcher has a Field type (Identifier or Attribute), a Field, a Compare mode (Exact, Normalized, Domain only, Fuzzy), a Weight, Required, and for fuzzy matching a Fuzzy threshold between 0 and 1. A matcher marked Required must fire, or the rule does not. The Matcher reference panel in the dialog summarises the compare modes and when a rule fires.

Resolve tab

Runs the active ruleset for one profile, so you can see what it would do.

FieldMeaningValid valuesDefault
Entity typeThe kind of profile.An existing entityNone
Profile IDThe profile to resolve.A profile idEmpty
…or ad-hoc identifiersIdentifier values to resolve without an existing profile.Key and value pairsNone
Dry run (do not persist)Show the cluster the rules would form without storing anything.On or off

Clusters tab

Cluster inspector looks up a cluster by its Cluster ID and lists the members with their Confidence; the root member is marked Root.

Actions

  • Activate or Deactivate a ruleset. Only an active ruleset is used.
  • Run re-resolution queues every profile of the entity type to be matched again. A confirmation warns that "This may create new merges", and afterwards the screen reports how many profiles were queued.
  • Edit rules opens the ruleset page; Edit changes its name, description and merge policy.
  • Resolve on the Resolve tab shows the Resulting cluster with its members and the Edges (the rule behind each link and its confidence).
  • To find and merge profiles that share an identifier value, use Dedup identifiers.

Examples

Match on email. Create a ruleset customer identity for customer. Add a rule same email, Deterministic, Priority 10, with one matcher: Identifier email, Compare Normalized. Activate the ruleset, then use Resolve with a dry run on a known profile to check the cluster.

Looser match on name and city. Add a second rule, Probabilistic, Priority 50, Logic Weighted threshold with Min score 0.8, and matchers on last_name (weight 0.5, Fuzzy) and city (weight 0.5, Exact).

Common problems

  • Nothing matches. The ruleset is inactive, has no rules, or its matchers need an attribute that is missing on one of the profiles.
  • Too many profiles merged. A fuzzy or weighted rule is too loose. Raise the threshold or Min score, or mark the strong matcher Required.
  • "At least one matcher is required." Add a matcher before saving the rule.
  • Re-resolution seems to do nothing. It runs in the background; check the Clusters tab after a short while.

On this page