Baseline global behavior parameters
These fields are used by every marketplace in baseline version 1.0.0. Their values are visible in simulations/baseline/configs/default.json; none is presented as measured Amazon shopper behavior.
weekly_traffic_multipliers
Meaning. A repeating list of relative traffic levels, one for each position in a seven-day cycle.
Example. A value of 1.18 produces 18 percent more expected impressions and path users than a neutral value of 1.00, before other scales and noise.
Why it makes sense. Shopping and media use often vary by day of week. A repeating cycle is a transparent way to represent seasonality.
Relationship. It is multiplied by daily_traffic_trend. Both performance impressions and path users share the result.
Confidence. Medium for the existence of weekly seasonality; low for the configured pattern.
daily_traffic_trend
Meaning. A fixed proportional traffic increase for each day after the first.
Example. With 0.0025, day 20 receives:
That is 5 percent above the first day's trend level.
Why it makes sense. Campaign reach can gradually change as budgets, delivery, awareness, or platform learning accumulate. A linear trend is easy to inspect.
Limitation. Real growth may plateau or reverse; the current slope is assumed.
Confidence. Low.
conversion_probability_daily_noise_standard_deviation
Meaning. The spread of one shared daily random change to conversion log odds for a marketplace.
Example. 0.12 allows some days to be better or worse than the configured response even when touchpoints are unchanged.
Why it makes sense. Weather, payday timing, promotions, news, and unrecorded demand can influence daily buying. Modeling unexplained variation prevents identical days.
Relationship. The regional version can scale this spread using the Gini coefficient; the baseline uses it unchanged.
Confidence. High that unobserved daily variation exists; low for 0.12.
performance_volume_noise_standard_deviation
Meaning. The spread of random multiplicative variation in impressions.
Example. A normal draw is passed through
Why it makes sense. Ad delivery varies with auctions and available inventory. Multiplicative noise reflects proportional variation and cannot create negative impressions.
Relationship. Applies after base_impressions, traffic_multiplier, and the seasonal multiplier.
Confidence. Medium for the method; low for 0.10.
path_audience_noise_standard_deviation
Meaning. The spread of random multiplicative variation in the users assigned to each path.
Example. 0.08 lets a path configured with 620 users vary from day to day while remaining positive.
Why it makes sense. The number of people following one exact path is not constant even when total campaign traffic is stable.
Confidence. Medium for the method; low for 0.08.
performance_revenue_noise_standard_deviation
Meaning. Small multiplicative variation in sales per simulated unit in the performance table.
Example. 0.025 represents modest effects from discounts, tax treatment, bundle composition, or reporting variation.
Why it makes sense. A real order value is rarely identical for every purchase. Positive multiplicative noise avoids impossible negative sales.
Relationship. Separate from path_revenue_noise_standard_deviation, because the Amazon Ads diagnostic report and Amazon Marketing Cloud outcome report are distinct measurements.
Confidence. Low.
path_revenue_noise_standard_deviation
Meaning. Small multiplicative variation in revenue per purchase for path outcomes.
Example. 0.018 lets path revenue differ slightly from a fixed list price.
Why it makes sense. Aggregated path revenue can include different basket compositions. The separate noise also prevents readers from treating the two reports as one accounting ledger.
Confidence. Low.
additional_unit_probability
Meaning. Probability that a diagnostic purchase contains one extra unit.
Example. 0.14 gives each purchase a 14 percent chance of adding a second unit, so unitsSold may exceed purchases.
Why it makes sense. A purchase is an order-like event, while units count products. One order can contain multiple units.
Confidence. High for the distinction; low for 14 percent.
repeat_purchase_probability
Meaning. Probability that one converted path user contributes one additional purchase during the observation.
Example. 0.09 lets purchase_count exceed converted_users without ever making converted_users exceed users.
Why it makes sense. Unique buyers and transactions are different counts. A buyer can make more than one purchase.
Confidence. High for the distinction; low for 9 percent.