Automate lending decisions in one decision flow
Rules sit in one place, data calls in the application, scorecards in a spreadsheet and history in logs. Decisimo joins the decision in one flow and runs it through one endpoint.
Run the whole decision through one endpoint
A decision flow joins data sources, rule sets, decision tables, scorecards, functions and models in order. Each step reads earlier results by path in one JSON document. The application sends one document to the flow's endpoint and gets the whole document back with every step's output. See what a rule set and a decision table look like.
Call bureau and fraud data during the decision
One data source calls several providers in parallel, including a credit bureau, device and fraud data, and a merchant profile. A failed call carries its error and HTTP code, so a rule decides what a missing answer means. A repeat lookup can come from a cache in the lender’s own database. Connect a provider.
Use scores, models and reasons in each decision
A scorecard puts each value in a bin, and a test run shows the bin and its contribution. A model step can call a model over HTTP, run an uploaded Python, R or Julia model, or call an LLM. A decision table row writes a reason in plain language with its result.
Keep every decision in the lender’s database
Once the lender connects a decision archive, every decision is written to a table in its own PostgreSQL database with its input and output. An outcome history beside it records what happened afterwards.
Judge a change on decisions already made
A backtest runs a changed release over decisions the production endpoints already made and lists every decision that would change, with application performance where outcomes are recorded. Every save creates a revision. Nothing runs until a release is deployed to an endpoint, which can carry an approval sequence. See how to set up a backtest on production data.
Questions we get asked
What is a decision engine?
A decision engine combines data sources, rules, scorecards and models to produce a decision. Decisimo runs them in one decision flow, called through an endpoint that returns the full JSON document with each output.
How does a decision engine differ from a rules engine?
A rules engine applies rule sets and decision tables. A decision engine joins those with data sources, scorecards and models in a decision flow. See what a rule set and decision table look like.
How does the lender’s system call a decision engine?
Each decision flow has its own API endpoint. The lender’s application sends one JSON document and receives the full document back with every output from the flow.
What happens when a bureau or fraud provider does not answer?
The failed call carries its error and HTTP code in the output. A rule reads those values like any other input, so the policy decides what the missing result means.
Can a decision engine run the lender’s machine learning models?
A model step can call a model over HTTP or run an uploaded Python, R or Julia model. It can also call an LLM as part of the decision flow.
How is a decision change tested before customers see it?
A backtest runs a changed release over decisions already made by production endpoints and lists the decisions that would change. Saved revisions and test cases can also be compared and rerun before deployment.
Start with one decision flow
After signing up free, create a decision flow and add the logic that runs your lending decision.