Baseline control parameters
These parameters control one complete simulation run. They do not use the additional national economic fields described in Regional Parameters.
seed
Meaning. The starting state of Python's pseudorandom number generator.
Example. 20260729 reproduces the same sequence of traffic noise, clicks, purchases, and revenue draws when every other resolved parameter is unchanged.
Why it makes sense. A synthetic experiment must be repeatable before two algorithms can be compared fairly. The seed controls repeatability; it is not a claim about shopper behavior.
Related parameters. Every parameter that controls a random distribution, especially conversion_probability_daily_noise_standard_deviation.
Confidence. High for reproducibility.
report_start_date
Meaning. The first included calendar date.
Example. 2026-07-01 begins the baseline run on July 1, 2026.
Why it makes sense. Advertising tables are time-bounded reports. An explicit start prevents accidental mixing of periods.
Related parameter. report_end_date.
Confidence. High as a design control.
report_end_date
Meaning. The last included calendar date.
Example. July 1 through July 31 creates 31 daily observations for each marketplace-touchpoint or marketplace-path combination.
Why it makes sense. Inclusive boundaries make the number of generated days auditable.
Confidence. High as a design control.
advertiser_id
Meaning. A synthetic account identifier written into both Amazon-like model-input tables.
Example. adv_synthetic_skincare lets the two reports refer to the same simulated advertiser without impersonating a real company.
Why it makes sense. Stable identifiers allow joins and validation. The synthetic word helps prevent provenance confusion.
Confidence. High as a data-contract choice.
base_product_price
Meaning. Product price in the simulator's reference currency before local-currency conversion.
Example. 125.0 represents a premium skincare product priced at 125 reference-currency units. In Canada it is multiplied by price_multiplier.
Why it makes sense. Purchase counts need a price to become revenue. Price also makes regional currency conversion testable.
Limitation. The repository does not show that 125 is an observed Amazon product price.
Confidence. Low; campaign assumption.
baseline_conversion_log_odds
Meaning. Purchase tendency before simulated touchpoint effects and daily noise.
Example. -4.0 becomes:
That is about a 1.8 percent starting probability.
Why it makes sense. Adding effects in log odds lets the final logistic response remain between zero and one. It also represents diminishing probability gains as conversion becomes more likely.
Related parameters. conversion_log_odds_effect and adjacent_synergy_log_odds.
Limitation. -4.0 is not calibrated to an Amazon experiment.
Confidence. Medium for the method; low for the value.