Your corporate decisions are either manual, report-enabled, semi-automated, or fully automated....
Review timing (during or after?)
In-Flight vs. Post-Implementation reviews: model build and refinement
Data-driven (or data-enabled) decision-making hinges on reliability and accuracy.
For simplicity in this article, we will call those inputs "models".
In practice, they could include data, rules, models, flows, transformations, and a range of other items.
- In-flight reviews happen during the project or build
- Post-implementation reviews happen after deployment.
In this article, we detail the benefits of both approaches for model build and refinement.
In-Flight Reviews
Auditors conduct these during the active development and deployment of models.
Post-Implementation
Auditors conduct these after your team has deployed your model.
-
Maturity. Data quality, and processes, can improve over time. Post-implementation reviews work with a more mature dataset and refined process. This may be more accurate and comprehensive than what was available during development.
-
Resource capacity. Conducting audits during projects can put strain on project resources during development. This becomes easier (in theory) post deployment.
A Hybrid Approach
A hybrid approach combines the advantages of both.
-
Emphasise In-Flight: a heavier emphasis on in-flight reviews during the active development and deployment phases.
-
Lightweight Post-Implementation: focus on the most critical items, rather than exhaustive reviews. Include items that have shifted since implementation.
-
Adjust and Refine: apply insights gained from post-implementation assessments to refine in-flight processes for future projects.
Summing up
The choice between in-flight and post-implementation is not an "either-or" scenario.