Skip to contents

Collapses an mm_paths object into the aggregated "a > b > c" table that ChannelAttribution consumes. This package builds the journeys and fits the order-1 chain of markov_removal(); ChannelAttribution computes Markov removal effects on them in C++, for higher-order chains too.

Usage

as_channel_paths(paths, sep = " > ")

Arguments

paths

An mm_paths object from build_paths().

sep

Separator between channels, " > " by default. A channel name that contains sep is an error, since its path strings could not be split back into channels. ChannelAttribution's markov_model() has a sep argument of its own; give both the same value, as the example does.

Value

A data frame with columns path, total_conversions, total_null and, when the journey table carries values, total_conversion_value. One row per distinct channel sequence, ordered by descending conversions, then descending null journeys, then path in the C locale.

Details

The total_null column is why build_paths() keeps non-converting journeys by default. Markov removal effects compare the probability of reaching conversion with the probability of reaching the null state. Without null paths the transition matrix has no absorbing failure state, the baseline conversion probability is 1, and each channel's removal effect moves up or down depending on how the missing non-converting journeys differ from the converting ones.

Examples

data(mm_events)
paths <- build_paths(mm_events, id = "customer_id", channel = "channel",
                     timestamp = "timestamp", conversion = "conversion",
                     value = "value")

ca <- as_channel_paths(paths)
head(ca)
#>                    path total_conversions total_null total_conversion_value
#> 1                social                98         97               13267.94
#> 2           paid_search                81         73               10658.24
#> 3        organic_search                80         84                8997.63
#> 4               display                79        118                9628.17
#> 5 display > paid_search                72         90                9719.46
#> 6  social > paid_search                65         93                9576.09

# Hand off to ChannelAttribution, if it is installed, with the same
# separator on both sides
if (requireNamespace("ChannelAttribution", quietly = TRUE)) {
  ca_in <- as_channel_paths(paths, sep = ">")
  ChannelAttribution::markov_model(
    ca_in, var_path = "path",
    var_conv = "total_conversions", var_null = "total_null", sep = ">"
  )
}
#> 
#> Number of simulations: 100000 - Convergence reached: 1.90% < 5.00%
#> 
#> Percentage of simulated paths that successfully end before maximum number of steps (10) is reached: 99.11%
#> 
#> [1] "*** Install ChannelAttribution Pro for free running install_pro(). Visit https://channelattribution.io for more info. Set flg_pro=FALSE to hide this message."
#>      channel_name total_conversions
#> 1          social         450.44758
#> 2     paid_search         493.04727
#> 3  organic_search         417.42652
#> 4         display         464.31138
#> 5           email         356.17372
#> 6       affiliate         279.04056
#> 7         (blank)          82.17454
#> 8          (none)          75.87282
#> 9          direct          75.62075
#> 10      (missing)          46.88486