Skip to contents

Applies media_transform() to each channel column of a plan, each with its own carryover and saturation, and returns the plan with those columns replaced. It is the step between a table of spend and the regressors a model is fitted on, and the step simulate_plan() and allocate_budget() repeat for every plan they evaluate.

Usage

transform_plan(plan, spec, by = NULL)

Arguments

plan

A data frame with one row per period and a numeric column of spend for each channel. Other columns, such as a date or a geography, may be present.

spec

A list with one entry per channel, named after the channel's column in plan. Each entry is a list with elements adstock and saturation, in the form media_transform() takes them, and optionally order. An element left out takes media_transform()'s default, so an entry without adstock gets geometric carryover at decay = 0.5; write adstock = NULL for none.

by

Optional grouping, so that carryover runs within each group, such as a geography, and never from one into the next: column names of plan, or a vector or data frame of vectors with one element or row per row of plan, as for media_transform().

Value

plan, with each column named in spec replaced by its transformed series. The rows keep their order and every other column is unchanged.

Details

A spec entry may hold only adstock, saturation and order. Any other element name is an error, so a misspelt saturaton cannot quietly leave a channel unsaturated. The grouping is set once, by by, for every channel, and a by inside an entry's adstock list is an error.

A character by whose elements all name columns of plan is read as those columns; any other by is a grouping vector, or a data frame of them. simulate_plan(), reallocate() and allocate_budget() read by the same way.

An error or warning from one channel's transform is reported with the channel's name, followed by the message media_transform() gave.

Examples

data(mm_weekly)
north <- mm_weekly[mm_weekly$geo == "north", ]
truth <- attr(mm_weekly, "truth")
channels <- c("tv", "video", "search", "social", "display")

# One entry per channel, here with the parameters the data was simulated
# from.
spec <- lapply(stats::setNames(channels, channels), function(ch) list(
  adstock = list(decay = truth$decay[[ch]]),
  saturation = list(half_max = truth$half_max[[ch]],
                    shape = truth$shape[[ch]])
))

media <- transform_plan(north, spec)
round(head(media[channels], 4), 4)
#>       tv  video search social display
#> 1 0.0000 0.1663 0.3435 0.3175  0.0000
#> 2 0.0000 0.1225 0.5001 0.4202  0.2958
#> 3 0.0441 0.0890 0.5501 0.2459  0.1877
#> 4 0.0343 0.0640 0.5598 0.3838  0.3349

# A panel: carryover runs within each geography
panel <- transform_plan(mm_weekly, spec, by = "geo")
nrow(panel) == nrow(mm_weekly)
#> [1] TRUE