R/mgrrel_matrix.R

In wpa: Tools for Analysing and Visualising Viva Insights Data

# --------------------------------------------------------------------------------------------
# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License. See LICENSE.txt in the project root for license information.
# --------------------------------------------------------------------------------------------

#' @title Manager Relationship 2x2 Matrix
#'
#' @description
#' Generate the Manager-Relationship 2x2 matrix, returning a 'ggplot' object by
#' default. Additional options available to return a "wide" or "long" summary
#' table.
#'
#' @author Lucas Hogner <lucas.hogner@@microsoft.com>
#'
#' @param data Standard Person Query data to pass through. Accepts a data frame.
#' @param hrvar HR Variable by which to split metrics. Accepts a character
#'   vector, e.g. "Organization". Defaults to `NULL`.
#' @param mingroup Numeric value setting the privacy threshold / minimum group
#'   size. Defaults to 5.
#'
#' @param return String specifying what to return. This must be one of the
#'   following strings:
#'   - `"plot"`
#'   - `"table"`
#'   - `"data"`
#'
#' See `Value` for more information.
#'
#' @param plot_colors Pass a character vector of length 4 containing HEX codes
#'   to specify colors to use in plotting.
#' @param threshold
#' Specify a numeric value to determine threshold (in minutes) for 1:1 manager hours.
#' Defaults to 15.
#'
#' @return
#' A different output is returned depending on the value passed to the `return`
#' argument:
#'   - `"plot"`: ggplot object. When `NULL` is passed to `hrvar`, a two-by-two
#'   grid where the size of the grid represents total percentage of employees is
#'   returned. Otherwise, a horizontal stacked bar plot is returned.
#'   - `"table"`: data frame. A summary table is returned.
#'   - `"data"`: data frame. A long table grouped at the `PersonId` level with
#'   the following columns:
#'     - `PersonId`
#'     - HR variable supplied to `hrvar`
#'     - `CoattendanceRate`
#'     - `Meeting_hours_with_manager_1_on_1`
#'     - `mgr1on1`
#'     - `Type`
#'
#' @import dplyr
#' @import reshape2
#' @import ggplot2
#' @importFrom scales percent
#'
#' @family Visualization
#' @family Managerial Relations
#'
#' @examples
#' # Return matrix
#' mgrrel_matrix(sq_data)
#'
#' # Return stacked bar plot
#' mgrrel_matrix(sq_data, hrvar = "Organization")
#'
#' ## Visualize coaching style types
#' # Ensure dplyr is loaded
#' library(dplyr)
#'
#' # Extract PersonId and Coaching Type
#' match_df <-
#'   sq_data %>%
#'   mgrrel_matrix(return = "data") %>%
#'   select(PersonId, Type)
#'
#' # Join and visualize baseline
#' sq_data %>%
#'   left_join(match_df, by = "PersonId") %>%
#'   keymetrics_scan(hrvar = "Type",
#'                   return = "plot")
#'
#' @export
mgrrel_matrix <- function(data,
                          hrvar = NULL,
                          mingroup = 5,
                          return = "plot",
                          plot_colors = c("#fe7f4f", "#b4d5dd", "#facebc", "#fcf0eb"),
                          threshold = 15){

  ## Add dummy "Total" column if hrvar = NULL
  if(is.null(hrvar)){

      data <- mutate(data, Total = "Total")
      hrvar <- "Total"

    }


  ## Check inputs
  required_variables <- c("Date",
                          hrvar,
                          "PersonId",
                          "Meeting_hours_with_manager",
                          "Meeting_hours",
                          "Meeting_hours_with_manager_1_on_1")

  ## Error message if variables are not present
  ## Nothing happens if all present
  data %>%
    check_inputs(requirements = required_variables)

  ## Create a Person Weekly Average
  data1 <-
    data %>%
    # Coattendance Rate with Manager
    mutate(coattendman_rate =
             (Meeting_hours_with_manager / Meeting_hours) %>%
             tidyr::replace_na(replace = 0) %>% # Replace NAs with 0s
             ifelse(!is.finite(.), 0, .)) %>% # Replace Inf with 0s
    group_by(PersonId, !!sym(hrvar)) %>%
    summarise(
      Meeting_hours_with_manager = mean(Meeting_hours_with_manager, na.rm = TRUE),
      Meeting_hours = mean(Meeting_hours, na.rm = TRUE),
      Meeting_hours_with_manager_1_on_1 =
        mean(Meeting_hours_with_manager_1_on_1, na.rm = TRUE),
      coattendman_rate = mean(coattendman_rate, na.rm = TRUE),
      Employee_Count = n_distinct(PersonId),
      .groups = "drop"
    )

  ## Threshold
  thres_low_chr <- paste("<", threshold, "min")
  thres_top_chr <- paste(">=", threshold, "min")


  ## Create key variables
  data2 <-
    data1 %>%
    mutate(coattendande = ifelse(coattendman_rate < 0.5, "<50%", ">=50%"),
           mgr1on1 = ifelse(Meeting_hours_with_manager_1_on_1 * 60 < threshold,
                            thres_low_chr,
                            thres_top_chr))

  ## Filter mingroup
  valid_orgs <- NULL

  valid_orgs <- # String with valid organizations
    data2 %>%
    hrvar_count(hrvar = hrvar,
                return = "table") %>%
    filter(n > mingroup) %>%
    pull(!!sym(hrvar))

  ## Grouping variable split
  if(hrvar == "Total"){

    chart <-
      data2 %>%
      count(mgr1on1, coattendande) %>%
      mutate(perc = n / sum(n)) %>% # Calculate percentages
      mutate(xmin = ifelse(mgr1on1 == thres_low_chr, -sqrt(perc), 0),
             xmax = ifelse(mgr1on1 == thres_top_chr, sqrt(perc), 0),
             ymin = ifelse(coattendande == "<50%", -sqrt(perc), 0),
             ymax = ifelse(coattendande == ">=50%", sqrt(perc), 0),
             mgrRel = case_when(mgr1on1 == thres_low_chr & coattendande == "<50%" ~ "Under-coached",
                                mgr1on1 == thres_low_chr & coattendande == ">=50%" ~ "Co-attending",
                                mgr1on1 == thres_top_chr & coattendande == ">=50%" ~ "Highly managed",
                                TRUE ~ "Coaching")) %>%
      mutate_at("mgrRel", ~as.factor(.))

    clean_tb <-
      chart %>%
      select(mgrRel, n, perc) %>%
      group_by(mgrRel) %>%
      summarise_all(~sum(., na.rm = TRUE))

  } else if(hrvar != "Total"){

    chart <-
      data2 %>%
      count(!!sym(hrvar), mgr1on1, coattendande) %>%
      group_by(!!sym(hrvar)) %>%
      mutate(perc = n / sum(n)) %>% # Calculate percentages
      mutate(xmin = ifelse(mgr1on1 == thres_low_chr, -sqrt(perc), 0),
             xmax = ifelse(mgr1on1 == thres_top_chr, sqrt(perc), 0),
             ymin = ifelse(coattendande == "<50%", -sqrt(perc), 0),
             ymax = ifelse(coattendande == ">=50%", sqrt(perc), 0),
             mgrRel = case_when(mgr1on1 == thres_low_chr & coattendande == "<50%" ~ "Under-coached",
                                mgr1on1 == thres_low_chr & coattendande == ">=50%" ~ "Co-attending",
                                mgr1on1 == thres_top_chr & coattendande == ">=50%" ~ "Highly managed",
                                TRUE ~ "Coaching")) %>%
      ungroup() %>%
      mutate_at("mgrRel", ~as.factor(.)) %>%
      filter(!!sym(hrvar) %in% valid_orgs)

    clean_tb <-
      chart %>%
      select(mgrRel, !!sym(hrvar), n, perc) %>%
      group_by(mgrRel, !!sym(hrvar)) %>%
      summarise_all(~sum(., na.rm = TRUE))

  }

  ## Sort colours out
  # Legacy variable names
  myColors <- plot_colors
  names(myColors) <- levels(chart$mgrRel)

  ## Show stacked bar chart if multiple groups

  if(hrvar == "Total"){

    plot <-
      chart %>%
      ggplot() +
      geom_rect(aes(xmin=xmin, xmax=xmax, ymin=ymin, ymax=ymax, fill = mgrRel), color = "white") +
      scale_fill_manual(name = "mgrRel", values = myColors) +
      geom_text(aes(x = xmin + 0.5*sqrt(perc),
                    y = ymin + 0.5*sqrt(perc),
                    label = scales::percent(perc, accuracy = 1))) +
      coord_equal() +
      scale_x_continuous(breaks = c(-max(abs(chart$xmin),abs(chart$xmax))/2,max(abs(chart$xmin),abs(chart$xmax))/2),
                         labels = c(thres_low_chr, thres_top_chr),
                         limits = c(-max(abs(chart$xmin), abs(chart$xmax)), max(abs(chart$xmin), abs(chart$xmax)))) +
      scale_y_continuous(breaks = c(-max(abs(chart$ymin), abs(chart$ymax))/2, max(abs(chart$ymin), abs(chart$ymax))/2),
                         labels = c("<50%", ">=50%"),
                         limits = c(-max(abs(chart$ymin), abs(chart$ymax)), max(abs(chart$ymin), abs(chart$ymax)))) +
      theme_wpa_basic() +
      labs(
        x = "Weekly 1:1 time with manager",
        y = "Employee and manager coattendance",
        caption = extract_date_range(data, return = "text")
        )

  } else if(hrvar != "Total"){

    plot <-
      chart %>%
      mutate(Fill = case_when(mgrRel == "Co-attending" ~ rgb2hex(68,151,169),
                              mgrRel == "Coaching" ~ rgb2hex(95,190,212),
                              mgrRel == "Highly managed" ~ rgb2hex(49,97,124),
                              mgrRel == "Under-coached" ~ rgb2hex(89,89,89))) %>%
      ggplot(aes(x = !!sym(hrvar), y = perc, group = mgrRel, fill = Fill)) +
      geom_bar(position = "stack", stat = "identity") +
      geom_text(aes(label = paste(round(perc * 100), "%")),
                position = position_stack(vjust = 0.5),
                color = "#FFFFFF",
                fontface = "bold") +
      scale_fill_identity(name = "Coaching styles",
                          breaks = c(rgb2hex(68,151,169),
                                     rgb2hex(95,190,212),
                                     rgb2hex(49,97,124),
                                     rgb2hex(89,89,89)),
                          labels = c("Co-attending",
                                     "Coaching",
                                     "Highly managed",
                                     "Under-coached"),
                          guide = "legend") +
      scale_y_continuous(labels = scales::percent) +
      coord_flip() +
      theme_wpa_basic() +
      labs(title = "Distribution of types of \nmanager-direct relationship across organizations",
           subtitle = "Based on manager 1:1 time and percentage of overall time spent with managers",
           y = "Percentage")
  }


  if(return == "plot"){

    plot +
      labs(title = "Distribution of types of \nmanager-direct relationship",
           subtitle = "Based on manager 1:1 time and percentage of\noverall time spent with managers")

  } else if(return == "table"){

    clean_tb %>%
      as_tibble() %>%
      return()

  } else if(return == "chartdata"){

    chart

  } else if(return == "debug"){

    data2

  } else if(return == "data"){


    data2 %>%
      mutate(Type =
               case_when(mgr1on1 == thres_low_chr & coattendande == "<50%" ~ "Under-coached",
                         mgr1on1 == thres_low_chr & coattendande == ">=50%" ~ "Co-attending",
                         mgr1on1 == thres_top_chr & coattendande == ">=50%" ~ "Highly managed",
                         TRUE ~ "Coaching")) %>%
      select(PersonId,
             !!sym(hrvar),
             CoattendanceRate = "coattendman_rate",
             Meeting_hours_with_manager_1_on_1,
             mgr1on1,
             Type)

  } else {

    stop("Please enter a valid input for `return`.")

  }
}