uflow
Glossary

Automated credit glossary: 30 terms every team should share

From decision engine to champion/challenger: 30 automated credit terms defined in two lines, so business, risk and technology speak the same language.

Updated July 2026 · 7 min read

In short

Thirty automated credit terms — from decision engine to champion/challenger — defined in two or three lines each, so business, risk and technology use the same vocabulary.

Half the arguments between business, risk and technology are vocabulary misunderstandings. This glossary defines, in two or three lines each, the terms that come up every day in a credit automation project — the same ones we use throughout the rest of these guides.

Written byMariano Sokal · COO at uFlow

COO at uFlow. Works with banks, fintechs and retailers across Latin America on how credit policies are governed: who decides, how a change is controlled, and what evidence remains for audit and the regulator.

The engine and the policies

  • Decision engine: a platform that executes decision policies — rules, calculations, data-source calls and models — automatically, traceably, and in real time or in bulk.
  • Policy (decision policy): the complete flow that assesses an application: which data it queries, which rules it applies and what outcome it produces. In uFlow it is designed visually with nodes.
  • Node: each piece of a policy — a credit bureau call, a calculation, a matrix, a decision. Nodes connect to form the flow.
  • Variable: each piece of data moving through the policy: what comes in (ID document, income), what sources return (bureau score) and what calculations generate.
  • No-code: an approach where the policy is built and modified visually, without programming — the risk team operates without depending on development.
  • Versioning: the record of every state of a policy. It lets you know which rules were running on any given date and roll back a failed change.
  • Rollback: immediate reactivation of a previous policy version when a new change produces unwanted results.
  • Subpolicy (call): a policy invoked from another one to reuse common logic — for example, an identity module shared across several products.

Data and sources

  • Credit bureau: an external source of credit history (debts, arrears, inquiries). Regional examples: Veraz, Experian, TransUnion, DataCrédito, Equifax.
  • Waterfall (data cascade): a design that queries sources in stages, from the cheapest to the most expensive, and only while the decision still needs them.
  • Concurrent calls: running several sources in parallel to reduce the total latency of the decision.
  • Re-query window (provider cache): the number of days during which the engine reuses a provider response it already paid for instead of querying again.
  • Raw response: the provider's original response exactly as it arrived, unprocessed — evidence for audit and debugging.
  • Alternative data: non-credit information with predictive power — bank transactions, telco, utilities, behavior with your own business.
  • Open finance: a framework in which the customer authorizes sharing their financial data between institutions via APIs — the formal route to transaction data.
  • Thin file: an applicant with little or no credit history, where the bureau contributes little signal.

Models and risk

  • Score: a number summarizing the risk of a profile. It can come from the bureau or from the institution's own model.
  • Credit scoring model: a statistical or machine learning model that estimates the probability of default from applicant variables.
  • PD (probability of default): the estimated probability that a customer defaults within a given horizon — an input to pricing and provisioning.
  • Champion/challenger testing: a technique that runs the live policy (champion) and a candidate policy (challenger) over shares of real traffic to compare outcomes before adopting the change.
  • Model risk: the risk that a model decides badly — through bias, deterioration or misuse. It is governed with validation, monitoring and documentation.
  • Drift: the gradual loss of a model's predictive power when reality shifts away from the data it was trained on.
  • Bias: systematic distortion of a model or policy against a segment, inherited from historical data or from proxy variables.
  • Explainability: the ability to answer, for every decision, why it was made — with the rules, data and version that produced it.

Operations and governance

  • Transaction: the record of one policy execution: inputs, path taken, version applied and outcome. It is the unit of decision traceability.
  • Decision traceability: the ability to reconstruct any past decision end to end from its transaction.
  • Batch (bulk processing): running a policy over thousands of cases at once — portfolio re-scoring, pre-approval campaigns, backtesting.
  • Pre-approval: a credit offer calculated in bulk across the customer base before customers ask for it, ready for the commercial campaign.
  • Webhook: an automatic notification from the engine to another system when a transaction finishes — the basis of asynchronous flows.
  • API key: the credential a system uses to authenticate against the engine in order to execute policies or retrieve results.
  • Segregation of duties: separation of permissions between whoever designs, whoever approves and whoever deploys — nobody controls the full cycle alone.
  • Early-stage delinquency: the first days past due, where preventive collections work has the greatest impact on recovery.

Technical documentation

How this is implemented in the engine, step by step.

Frequently asked questions

Common questions

Are a decision engine and a rules engine the same thing?+

A classic rules engine executes if/then conditions. A modern decision engine adds what real operations need: orchestration of data sources, ML models, versioning, traceability per transaction, bulk processing and controlled experimentation.

Who is this glossary for?+

For the cross-functional teams driving automation: risk, product, technology and compliance. Shared definitions shorten discussions — and each term points to a guide where it is developed in depth.

Would this work for your decisioning process?

Transform your credit assessment process with the decision engine.