Stage I
Frame the question
Define exactly what is being measured and which decision the answer is meant to inform. Clarify the question before choosing data or a method.
Two questions guide the work. Which parts of the process can be handled more efficiently, and what can the resulting information tell the decision-maker?
Automation can collect, reconcile, and prepare information when the source systems allow it. Economic analysis then uses that evidence to examine costs, productivity, performance, and the trade-offs behind the decision.
The analysis follows the same basic sequence from framing through interpretation. Smaller questions may require less work at each stage, while larger ones may require several rounds within a stage.
Define exactly what is being measured and which decision the answer is meant to inform. Clarify the question before choosing data or a method.
Ask what outcome would have been expected without the decision or event being studied. That comparison helps determine whether the design should use difference-in-differences, an event study, synthetic control, or another method suited to the setting and available data.
Identify the relevant sources, document how they are cleaned and combined, and define the treatment, comparison, timing, and outcomes in reproducible code where the setting allows it.
Estimate the effect together with its uncertainty and sensitivity to reasonable alternative assumptions, samples, and specifications. Those checks are part of interpreting the result.
Translate the result into the decision it is meant to inform. The final form may be a short memo, a reporting update, or another concise output, with the technical work available underneath for review.
Once a workflow is in place, agents can continue collecting and preparing information on a schedule when the source systems permit it. That creates a more consistent evidence base for the analytical stages above without turning the automated output itself into the conclusion.
Method choice depends on the structure of the question, the comparison available, and the quality of the data. The options below are part of the working toolkit and are selected according to those conditions.
Additional methods can include marginal and opportunity-cost analysis, structural modeling, instrumental variables, regression discontinuity, bootstrap inference, and Bayesian updating when the question and data support them.
An estimate should be open to examination. The supporting work can document the inputs, design choices, alternative specifications, and assumptions that materially affect the result so the conclusion can be challenged and checked.
The goal is reproducibility. Another analyst with the relevant training should be able to follow the documentation, understand the choices that were made, and reproduce the result when the data and permissions allow it.
Analysis can run in R, Python, or Stata depending on the task. Code, data-processing steps, and reporting logic can be version-controlled and documented so the work can be reviewed and handed off.
When the engagement and data permissions allow it, replication materials can accompany the final analysis so the client team can update the work as new data becomes available.
See how the analysis fits into Business & Sports work, or return to the broader consulting capabilities.