Bayesian Inference is a method of statistical inference that can be used in the analysis of observed data from marketing tests. Bayesian updates start with a prior distribution (prior probable information about the environment) and a likelihood function (an expected distribution from which the samples are drawn). Then, given some observed data, the prior can be multiplied by the likelihood of the data to produce a posterior distribution of probabilities. At the core of all of this is Bayes’ Rule.
\[ P(A\ |\ Data) \sim P(Data\ |\ A) \cdot P(A)\] This package is intended to abstract the math of the conjugate prior update rules to provide 3 pieces of information for a user:
Select which piece of information you would like to calculate.
Metric | Function Call |
---|---|
All Below Metrics | calculate_all_metrics() |
Win Probability | estimate_win_prob() |
Value Remaining | estimate_value_remaining() |
Lift vs. Control | estimate_lift_vs_baseline() |
Win Probability vs. Baseline | estimate_win_prob_vs_baseline() |
If you would like to calculate all the metrics then use
calculate_all_metrics()
. This is a slightly more efficient
implementation since it only needs to sample from the posterior once for
all 4 calculations instead of once for each metric.
All of these functions require a very specific tibble format.
However, the same tibble can be used in all metric calculations. A
tibble is used here because it has the additional check that all column
lengths are the same. A tibble of this format can also conveniently be
created using dplyr’s group_by() %>% summarise()
sequence of functions.
The columns in the following table are required if there is an
X
in the box for the distribution. (Int columns can also be
dbl due to R coercian)
Distribution Type | option_name (char) | sum_impressions (int) | sum_clicks (int) | sum_sessions (int) | sum_conversions (dbl) | sum_revenue (dbl) | sum_cost (dbl) | sum_conversions_2 (dbl) | sum_revenue_2 (dbl) | sum_duration (dbl) | sum_page_views (int) |
---|---|---|---|---|---|---|---|---|---|---|---|
Conversion Rate | X | X | X | ||||||||
Response Rate | X | X | X | ||||||||
Click Through Rate (CTR) | X | X | X | ||||||||
Revenue Per Session | X | X | X | X | |||||||
Multi Revenue Per Session | X | X | X | X | X | X | |||||
Cost Per Activation (CPA) | X | X | X | X | |||||||
Total CM | X | X | X | X | X | X | |||||
CM Per Click | X | X | X | X | X | ||||||
Cost Per Click (CPC) | X | X | X | ||||||||
Session Duration | X | X | X | ||||||||
Page Views Per Session | X | X | X |
We will use the Conversion Rate distribution for this example so we need the columns option_name, sum_clicks, and sum_conversions.
raw_data_long_format <- tibble::tribble(
~option_name, ~clicks, ~conversions,
"A", 6, 3,
"A", 1, 0,
"B", 2, 1,
"A", 2, 0,
"A", 1, 0,
"B", 5, 2,
"A", 1, 0,
"B", 1, 1,
"B", 1, 0,
"A", 3, 1,
"B", 1, 0,
"A", 1, 1
)
raw_data_long_format %>%
dplyr::group_by(option_name) %>%
dplyr::summarise(sum_clicks = sum(clicks),
sum_conversions = sum(conversions))
#> # A tibble: 2 × 3
#> option_name sum_clicks sum_conversions
#> <chr> <dbl> <dbl>
#> 1 A 15 5
#> 2 B 10 4
This input dataframe can also be created manually if the aggregations are already done in an external program.
# Since this is a stochastic process with a random number generator,
# it is worth setting the seed to get consistent results.
set.seed(1776)
input_df <- tibble::tibble(
option_name = c("A", "B", "C"),
sum_clicks = c(1000, 1000, 1000),
sum_conversions = c(100, 120, 110)
)
input_df
#> # A tibble: 3 × 3
#> option_name sum_clicks sum_conversions
#> <chr> <dbl> <dbl>
#> 1 A 1000 100
#> 2 B 1000 120
#> 3 C 1000 110
One note: clicks or sessions must be greater than or equal to the number of conversions (this is a rate bound between 0 and 1).
input_df
is used in the following examples.
This function wraps all the below functions into one call.
estimate_all_values(input_df, distribution = "conversion_rate", wrt_option_lift = "A")
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> $`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.726 72.59%
#> 2 C 0.228 22.76%
#> 3 A 0.0465 4.65%
#>
#> $`Value Remaining`
#> 95%
#> 0.1328629
#>
#> $`Lift vs Baseline`
#> 30%
#> 0.1205052
#>
#> $`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.923 92.25%
#> 2 A 0.0775 7.75%
This produces a tibble with all the option names, the
win_prob_raw
so this can be used as a double, and a cleaned
string win_prob
where the decimal is represented as a
percent.
Value Remaining is a measure of loss. If B is selected as the current best option, we can estimate with 95% confidence (default), that an alternative option is not more than X% worse than the current expected best option.
estimate_value_remaining(input_df, distribution = "conversion_rate")
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> 95%
#> 0.1317859
This number can also be framed in absolute dollar terms (or percentage points in the case of a rate metric).
The metric
argument defaults to lift
which
produces a percent lift vs the baseline. Sometimes we may want to
understand this lift in absolute terms (especially when samples from the
posteriors could be negative, such as Contribution Margin (CM).)
This function is used to compare an individual option to the best option as opposed to the win probability of each option overall.
Samples can be directly collected from the posterior with the following function.
sample_from_posterior(input_df, distribution = "conversion_rate")
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> # A tibble: 150,000 × 3
#> option_name samples sample_id
#> <chr> <dbl> <int>
#> 1 A 0.0991 1
#> 2 A 0.109 2
#> 3 A 0.106 3
#> 4 A 0.0923 4
#> 5 A 0.102 5
#> 6 A 0.113 6
#> 7 A 0.103 7
#> 8 A 0.102 8
#> 9 A 0.0875 9
#> 10 A 0.0924 10
#> # ℹ 149,990 more rows
(input_df_rps <- tibble::tibble(
option_name = c("A", "B", "C"),
sum_sessions = c(1000, 1000, 1000),
sum_conversions = c(100, 120, 110),
sum_revenue = c(900, 1200, 1150)
))
#> # A tibble: 3 × 4
#> option_name sum_sessions sum_conversions sum_revenue
#> <chr> <dbl> <dbl> <dbl>
#> 1 A 1000 100 900
#> 2 B 1000 120 1200
#> 3 C 1000 110 1150
estimate_all_values(input_df_rps, distribution = "rev_per_session", wrt_option_lift = "A")
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> $`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.580 57.99%
#> 2 C 0.396 39.62%
#> 3 A 0.0239 2.39%
#>
#> $`Value Remaining`
#> 95%
#> 0.2942552
#>
#> $`Lift vs Baseline`
#> 30%
#> 0.209918
#>
#> $`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.942 94.17%
#> 2 A 0.0583 5.83%
You may want to pass alternate priors to a distribution. Only do this if you are making an informed decision.
Beta - alpha0, beta0
Gamma - k0, theta0 (k01, theta01 if alternate Gamma priors are required)
Dirichlet - alpha_00 (none), alpha_01 (first conversion type), alpha_02 (alternate conversion type)
# You can also pass priors for just the Beta distribution and not the Gamma distribution.
new_priors <- list(alpha0 = 2, beta0 = 10, k0 = 3, theta0 = 10000)
estimate_all_values(input_df_rps, distribution = "rev_per_session", wrt_option_lift = "A", priors = new_priors)
#> $`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.580 58%
#> 2 C 0.398 39.76%
#> 3 A 0.0225 2.25%
#>
#> $`Value Remaining`
#> 95%
#> 0.291834
#>
#> $`Lift vs Baseline`
#> 30%
#> 0.2100943
#>
#> $`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.941 94.14%
#> 2 A 0.0586 5.86%
You may want to evaluate the results of a test in multiple different distributions.
(input_df_all <- tibble::tibble(
option_name = c("A", "B", "C"),
sum_impressions = c(10000, 9000, 11000),
sum_sessions = c(1000, 1000, 1000),
sum_conversions = c(100, 120, 110),
sum_revenue = c(900, 1200, 1150),
sum_cost = c(10, 50, 30),
sum_conversions_2 = c(10, 8, 20),
sum_revenue_2 = c(10, 16, 15)
) %>%
dplyr::mutate(sum_clicks = sum_sessions)) # Clicks are the same as Sessions
#> # A tibble: 3 × 9
#> option_name sum_impressions sum_sessions sum_conversions sum_revenue sum_cost
#> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 A 10000 1000 100 900 10
#> 2 B 9000 1000 120 1200 50
#> 3 C 11000 1000 110 1150 30
#> # ℹ 3 more variables: sum_conversions_2 <dbl>, sum_revenue_2 <dbl>,
#> # sum_clicks <dbl>
distributions <- c("conversion_rate", "response_rate", "ctr", "rev_per_session", "multi_rev_per_session", "cpa", "total_cm", "cm_per_click", "cpc")
# Purrr map allows us to apply a function to each element of a list. (Similar to a for loop)
purrr::map(distributions,
~ estimate_all_values(input_df_all,
distribution = .x,
wrt_option_lift = "A",
metric = "absolute")
)
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> Using default priors.
#> [[1]]
#> [[1]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.724 72.41%
#> 2 C 0.229 22.93%
#> 3 A 0.0466 4.66%
#>
#> [[1]]$`Value Remaining`
#> 95%
#> 0.01443064
#>
#> [[1]]$`Lift vs Baseline`
#> 30%
#> 0.01252419
#>
#> [[1]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.922 92.22%
#> 2 A 0.0778 7.78%
#>
#>
#> [[2]]
#> [[2]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.726 72.57%
#> 2 C 0.229 22.89%
#> 3 A 0.0454 4.54%
#>
#> [[2]]$`Value Remaining`
#> 95%
#> 0.01414997
#>
#> [[2]]$`Lift vs Baseline`
#> 30%
#> 0.01266849
#>
#> [[2]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.924 92.39%
#> 2 A 0.0761 7.61%
#>
#>
#> [[3]]
#> [[3]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.993 99.34%
#> 2 A 0.0066 0.66%
#> 3 C 0 0%
#>
#> [[3]]$`Value Remaining`
#> 95%
#> 0
#>
#> [[3]]$`Lift vs Baseline`
#> 30%
#> 0.008780786
#>
#> [[3]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.993 99.34%
#> 2 A 0.0066 0.66%
#>
#>
#> [[4]]
#> [[4]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.578 57.77%
#> 2 C 0.398 39.81%
#> 3 A 0.0241 2.41%
#>
#> [[4]]$`Value Remaining`
#> 95%
#> 0.3055693
#>
#> [[4]]$`Lift vs Baseline`
#> 30%
#> 0.1960611
#>
#> [[4]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.938 93.8%
#> 2 A 0.0620 6.2%
#>
#>
#> [[5]]
#> [[5]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.584 58.4%
#> 2 C 0.394 39.38%
#> 3 A 0.0222 2.22%
#>
#> [[5]]$`Value Remaining`
#> 95%
#> 0.3012504
#>
#> [[5]]$`Lift vs Baseline`
#> 30%
#> 0.2060909
#>
#> [[5]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.944 94.41%
#> 2 A 0.0559 5.59%
#>
#>
#> [[6]]
#> [[6]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 A 1 100%
#> 2 B 0 0%
#> 3 C 0 0%
#>
#> [[6]]$`Value Remaining`
#> 95%
#> 0
#>
#> [[6]]$`Lift vs Baseline`
#> 30%
#> 0
#>
#> [[6]]$`Win Probability vs Baseline`
#> # A tibble: 1 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 A 1 100%
#>
#>
#> [[7]]
#> [[7]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.534 53.38%
#> 2 C 0.429 42.9%
#> 3 A 0.0373 3.73%
#>
#> [[7]]$`Value Remaining`
#> 95%
#> 336.4168
#>
#> [[7]]$`Lift vs Baseline`
#> 30%
#> 155.9331
#>
#> [[7]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.906 90.55%
#> 2 A 0.0945 9.45%
#>
#>
#> [[8]]
#> [[8]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.534 53.44%
#> 2 C 0.432 43.18%
#> 3 A 0.0337 3.37%
#>
#> [[8]]$`Value Remaining`
#> 95%
#> 0.3244191
#>
#> [[8]]$`Lift vs Baseline`
#> 30%
#> 0.1576408
#>
#> [[8]]$`Win Probability vs Baseline`
#> # A tibble: 2 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 B 0.911 91.11%
#> 2 A 0.0889 8.89%
#>
#>
#> [[9]]
#> [[9]]$`Win Probability`
#> # A tibble: 3 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 A 1 100%
#> 2 B 0 0%
#> 3 C 0 0%
#>
#> [[9]]$`Value Remaining`
#> 95%
#> 0
#>
#> [[9]]$`Lift vs Baseline`
#> 30%
#> 0
#>
#> [[9]]$`Win Probability vs Baseline`
#> # A tibble: 1 × 3
#> option_name win_prob_raw win_prob
#> <chr> <dbl> <chr>
#> 1 A 1 100%