ClickLossScenario

Economics scenarios · Implementation: mlsysim/engine/scenario_definitions/economics.py.

Construct this scenario with keyword inputs. The fields below show the types and defaults declared by this class.

A silent click-through regression from ctr_before to ctr_after at request_rate, caught after detection_time; each click is worth value_per_click.

Notes

Equation: Exposed requests are rate*time; lost clicks are exposed_requests*(ctr_before-ctr_after).

Assumptions: Constant request rate, click-through rates, and value per click during the delay.

Excludes: Seasonality, click dependence, recovery effects, and causal revenue validation.

References

Derivation: At constant request rate, the observation window contains rate*time requests. Multiplying by the absolute click-through decrease yields expected lost clicks, and multiplying by value_per_click yields the assumed monetary loss.

Examples

>>> from mlsysim import Scenarios
>>> s = Scenarios.DelayedAlertClickLoss
>>> s.loss.to("USD").magnitude >= 0
True

Attributes

Name	Description
ctr_after
ctr_before
ctr_drop	Absolute click-through probability decrease, measured in probability points.
detection_time
loss	Expected lost revenue in USD at the supplied value per click.
loss_rate	Expected revenue loss rate in USD/hour during the regression.
lost_clicks	Expected lost-click count under constant request rate and click-through decrease.
request_rate
requests	Request count exposed during the configured detection delay.
value_per_click

request_rate

Type: Quantity.

Declared default: required.

detection_time

Type: Quantity.

Declared default: required.

ctr_before

Type: float.

Declared default: required.

ctr_after

Type: float.

Declared default: required.

value_per_click

Type: Quantity.

Declared default: required.

requests

engine.scenario_types.ClickLossScenario.requests

Request count exposed during the configured detection delay.

Equation: (request_rate * detection_time).to('').magnitude.

ctr_drop

engine.scenario_types.ClickLossScenario.ctr_drop

Absolute click-through probability decrease, measured in probability points.

Equation: ctr_before - ctr_after.

lost_clicks

engine.scenario_types.ClickLossScenario.lost_clicks

Expected lost-click count under constant request rate and click-through decrease.

Equation: requests * ctr_drop.

loss

engine.scenario_types.ClickLossScenario.loss

Expected lost revenue in USD at the supplied value per click.

Equation: (lost_clicks * value_per_click).to('USD').

loss_rate

engine.scenario_types.ClickLossScenario.loss_rate

Expected revenue loss rate in USD/hour during the regression.

Equation: (loss / detection_time).to('USD/hour').