Evaluates an event against a scenario and returns the outcome with its reasoning.

​
Request

​
string
required
The scenario that will evaluate it. Its live version is used.
​
string
required
The table name in your data model — trigger_object is validated against that schema.
​
object
required
The event itself. object_id and updated_at are required.
​
boolean
default:"false"
When on, each rule’s result and score is included in the response.
​
boolean
default:"false"
When on, the decision is also published as a decision.created webhook.

​
Response

​
Outcomes

Unknown is not an error: it means the scenario took no interest in this event. Handle it explicitly in your code — folding it into Approve means that the day your trigger condition narrows, all your traffic gets silently approved.

​
Rule results

An outcome of hit means the rule fired and contributed its score; no_hit means it was evaluated and did not fire. Including the rules that did not fire is deliberate — it is usually where the answer to “why does this rule never do anything” lives.

​
Latency

execution_time_ms is the time that decision spent in the engine; it excludes network time. Tens of milliseconds in a typical scenario, driven mostly by rule count and the width of your velocity windows.

​
Where the time went

One total says a decision was slow without saying which part of it was — and those parts slow down for unrelated reasons: the rules got more expensive, an aggregate is scanning more rows than it used to, the database is under load, or the broker is slow to acknowledge. A call asking for with_rule_execution_details also gets stage_timings: each stage, how many times it ran, and how long it took in total.
Repeated stages are aggregated rather than listed, because that is where the sentence worth reading is: “18 queries, 240 ms” explains a slow decision, where eighteen four-millisecond rows are the same fact spelled out in a way nobody adds up.
The stages do not sum to execution_time_ms. They nest and overlap, and an unmeasured gap is work nobody has named a stage yet rather than idle time. Do not turn this into a pie chart.
This is the answer for an integrator asking about one slow payment from yesterday, holding a correlation id and nothing else — no percentile dashboard can be asked about a single transaction.