robyn_crossmmm: Cross-MMM Budget Optimization

In laresbernardo/lares: Analytics & Machine Learning Sidekick

Cross-MMM Budget Optimization

Description

Given a list of recreated Robyn models, this function optimizes budget allocation across MMMs with respective constraints by maximizing incremental revenue/conversions. This method assumes each model is independent and can be compared given its spends were cleanly and properly split.

Usage

robyn_crossmmm(
  models,
  initial_budgets = NULL,
  start_dates = NULL,
  end_dates = NULL,
  budget_constr_low = 0.5,
  budget_constr_up = 1.5,
  channel_constr_low = budget_constr_low,
  channel_constr_up = budget_constr_up,
  cores = NULL,
  quiet = FALSE,
  ...
)

## S3 method for class 'robyn_crossmmm'
print(x, ...)

Arguments

models	Lists. Recreated Robyn models with robyn_recreate().
initial_budgets	Numeric vector. Default will use the total spends per model for the specified or total date range. Must be length 1 or same as models.
start_dates, end_dates	Character vector. Start and end dates for each specific model. You can specify a single date and will be used in all models. Default empty value will assume you want all available data and date range. Must be length 1 or same as models.
budget_constr_low, budget_constr_up	Numeric vector. Relative minimum and maximum budgets to consider based on initial_budgets. By default it'll consider 50 Must be length 1 or same as models.
channel_constr_low, channel_constr_up	Numeric vector. Relative minimum and maximum budgets to consider per channel. By default it'll consider 50 Must be length 1 or same as models.
cores	Integer. How many cores to use for parallel computations? Set to 1 to not use this option. Default will the minimum between 10 cores and all available cores - 1.
quiet	Boolean. Keep quiet? If not, informative messages will be printed.
...	Additional parameters to be passed to internal functions.
x	Object to print: robyn_crossmmm() output.

Value

Invisible vector with results by letter.

Examples

# You must have Robyn installed and some models stored as JSON files
## Not run: 
# Import and recreate the models
files <- c("BrandA.json", "BrandB.json", "BrandC.json", "BrandN.json")
models <- lapply(files, function(x) Robyn::robyn_recreate(x))
names(models) <- gsub("\\.json", "", files)

# Calculate cross-brand optimal allocation
res <- robyn_crossmmm(
  models,
  cores = 10,
  start_dates = "2023-01-01",
  end_dates = "2023-12-01"
)
print(res)
res$summary

## End(Not run)

laresbernardo/lares documentation built on Oct. 23, 2024, 12:05 p.m.