Automate real-time credit decisions from edit to endpoint

Policy changes move through separate tools, email and engineering. Decisimo takes each change from edit to a live endpoint in one product.

Start free See the stages

From an edit to the live endpoint

  1. Build the decision flow

  2. Test the parts

  3. Run regression cases

  4. Assess impact

  5. Sign-off

  6. Deploy the release

  7. Run live decisions

One decision flow, built from shared parts

  • Shared logicThe portal lists every flow that uses a shared rule set
  • FunctionsBuild a mathematical formula in the editor, with no code.
  • Model servicesModels trained in Python, R or Julia are served over HTTP.

Rules and tables

Decision canvas connects Affordability, Knock-out rules, Risk Scorecard and Score Cutoff from Start to End
Consumer Lending Decision runs Start to End through Affordability, Knock-out rules, Risk Scorecard and Score Cutoff.

See what a policy change moves

  • Test each partRun each artifact with chosen inputs and check table row overlap.
  • Regression suiteStudio runs become cases, then suites rerun after edits.
  • Impact scenariosCompare both policies on the same rows and keep changed decisions.

A change goes live as a release

  • Release comparisonCompare rules, rows and predictors, then draft release notes.
  • Sign-off sequenceNamed roles approve in order where the endpoint has sign-off.
  • Release contentsStores each flow step as compiled, with its source revision

Each live request returns a decision and reason

  • REST endpointSend one JSON document and receive JSON in return.
  • Response detailThe response can list what ran and the rules that fired.
  • Decision recordsInputs and outputs can be saved to your Postgres and browsed.
Score Cutoff decision table showing score ranges that return decline, refer or approve with reason codes
Score Cutoff returns decline, refer or approve, each with a reason code, in this demo case.

Where the endpoint runs and how it calls data

  • Granted regionEurope, the United States, Singapore or the Gulf.
  • Data sourcesSources run in parallel. Failure leaves values absent with an error flag.

Connect a data source

When a decision takes days

Evidence arriving over days or waiting on a person belongs in a case, not one request.

Decisions that wait

Questions we get asked

What is real-time decisioning?

Real-time decisioning applies a policy to each incoming request and returns a decision. Decisimo runs the released policy at an endpoint for each request.

Can I run my own scoring model at the endpoint?

A model trained in Python, R or Julia can be uploaded and served over HTTP. Formula models and tree ensembles also run inside the engine.

Test a policy change before release

Start free